<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:media="http://search.yahoo.com/mrss/">
  <channel>
    <title>Matheus Mierzwa</title>
    <link>https://www.mmlo.app/en</link>
    <description>AI Enablement Lead &amp; Frontend Tech Lead na InBot. Transformo processos manuais em skills, agentes e sistemas internos de IA, com React, TypeScript e Convex.</description>
    <language>en</language>
    <lastBuildDate>Sun, 08 Nov 2026 00:03:52 GMT</lastBuildDate>
    <atom:link href="https://www.mmlo.app/en/rss.xml" rel="self" type="application/rss+xml"/>
    <item>
      <title>Tacit Knowledge Became a Company Asset</title>
      <link>https://www.mmlo.app/en/blog/saber-tacito-ativo-da-empresa</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/saber-tacito-ativo-da-empresa</guid>
      <description>Who Keeps the Value?</description>
      <content:encoded><![CDATA[<p class="intro">In 1911, Frederick Taylor wrote that the knowledge built up in a trade was "the principal asset" of every worker. His proposal was to take that asset away from the worker and hand it to management. In 2026, we do the same thing under another name: we take the way a person works and turn it into a <em>skill</em> that any colleague, or any AI agent, can run. I know this from the inside, because part of my job as head of AI Enablement is exactly that. In this post I go through what Taylor did with the factory worker's knowledge, why the skill repeats that move and goes further, who keeps the value of what gets taught, what this does to junior, mid-level and senior roles, and what we can demand from companies.</p>

<h2>What Taylor Did with the Worker's Knowledge</h2>

<p>At the beginning of the 20th century, the person who knew how to do the work in a factory was the worker. The foreman knew less. Taylor himself admits it in <em>The Principles of Scientific Management</em>: foremen and superintendents knew, "better than any one else", that their own knowledge fell far short of the combined knowledge of all the men under them.</p>

<p>That knowledge had a name in the book. Taylor called it "rule-of-thumb": practical know-how, passed from person to person, that nobody had ever written down. And he wrote the sentence I borrowed for the title of this post:</p>

<blockquote><p><strong>"This mass of rule-of-thumb or traditional knowledge may be said to be the principal asset or possession of every tradesman."</strong></p></blockquote>

<p>His solution was to change who owned that asset. Management would take on the burden of gathering all the traditional knowledge of the workers and of "classifying, tabulating, and reducing this knowledge to rules, laws, and formulae". Each worker would then receive, at least one day in advance, complete written instructions describing the task and the way to do it. Taylor put it in one line: "In the past the man has been first; in the future the system must be first."</p>

<h3>Knowledge you can't put into words</h3>

<p>Decades later, the philosopher Michael Polanyi gave a name to what Taylor wanted to capture. In <em>The Tacit Dimension</em>, from 1966, he starts from a simple sentence: <strong>"we can know more than we can tell"</strong>. That is <em>tacit knowledge</em>. The mechanic who listens to an engine and knows what is wrong, the support agent who can tell from a customer's tone that they are about to cancel, the developer who looks at a log and already suspects the cache.</p>

<p>That knowledge protected a lot of people for a long time. The economist David Autor called it "Polanyi's Paradox": because our tacit knowledge goes beyond what we can explain, it was hard to write a program that did the same thing. What didn't fit into a rule didn't fit into a machine. Generative AI, and the skill written alongside it, go straight at that protection.</p>

<p>In 1974, Harry Braverman read Taylor with different eyes in <em>Labor and Monopoly Capital</em>. For him, the method had three principles: separate the labor process from the worker's skills, separate the people who conceive the work from the people who carry it out, and use control over knowledge to control every step of the work. Braverman also showed that the office was going through the same process as the factory. The technique was never only about productivity. It was about who runs the work, because whoever holds the knowledge has bargaining power.</p>

<p>Hold on to that idea of separating the people who conceive from the people who execute. It came back, and it came back through a road Taylor never had.</p>

<h2>The Skill Is the New Manual, and Now It Reaches Reasoning</h2>

<p>If you don't use them yet: a skill is a package of instructions, examples and scripts that teaches an AI agent to do a task the right way. Anthropic, which created the format and later opened it as a standard, describes the skill as a way of <strong>"packaging your expertise"</strong>. I already mentioned in <a href="https://www.mmlo.com.br/blog/skills-de-ia-por-area">AI Skills by Area</a> that Anthropic compares a skill to the onboarding material for a new hire. The detail that matters here is a different one: that material used to be written to teach a person, and now it is written to teach the agent that does that person's job.</p>

<p>The difference is that Taylor's manual stopped at the body. He timed movements, shovels, ore wheelbarrows. Even with all the changes in management that came after him, what could be extracted was the operational part. With skills, you can standardize a task that depends on reasoning and make sure it is always done the same way. For example: the playbook a senior analyst follows to investigate an incident, the criteria a QA uses to say a card is done, the way someone in sales reads a proposal.</p>

<p>The company that sells the format measured its effect. The January 2026 <em>Anthropic Economic Index</em> shows that the tasks AI covers most are precisely the ones that require more education, and concludes that the net first-order impact is <strong>"to deskill jobs"</strong>. The report also shows the other side: some occupations get upskilled, because AI takes the routine part and leaves the more complex one.</p>

<p>That is why I think the skill is the same move Taylor made, and maybe a worse one. When what gets standardized is reasoning, people lose mental effort, autonomy and room to create. They end up executing what someone else thought through before them.</p>

<h3>The move already has a name on GitHub</h3>

<p>In April 2026, MIT Technology Review told the story of a Chinese project that went viral on GitHub, <strong>Colleague Skill</strong>. It started as a parody: it imports a colleague's chat history and files and generates, from that, a manual for an agent to do their job. The article reports that bosses were asking employees to document their own work so it could be automated. The response came fast. A product manager published an "anti-distillation" skill, designed to sabotage that kind of extraction.</p>

<p>Taylor also had a word for resistance: <em>soldiering</em>, deliberately holding back the pace so management would never find out how much a day's work could really produce. A hundred and fifteen years later, soldiering became a repository.</p>

<h3>It depends on where the skill lands</h3>

<p>None of this means every skill is bad. It depends a lot on the place. Some companies use skills the right way: they automate the operational part, the boring and repetitive side of the day, and give people time back to think and build new things. Other companies only need a delivery factory, where quality and innovation matter less than volume. In those, the skill becomes a deskilling tool: people know less and less and just execute the next task.</p>

<p>Even people on the technical side have spotted the risk. The Thoughtworks Technology Radar from November 2026 put an antipattern called <strong>"agent skill cognitive surrender"</strong> in its Caution ring: teams that adopt huge libraries of ready-made skills without understanding what they do. Cognitive surrender is almost a technical name for what Braverman described on the factory floor.</p>

<p>If the skill concentrates knowledge, the next question is unavoidable: whose knowledge is it once it becomes a file?</p>

<h2>Who Keeps the Value of What Was Taught</h2>

<p>That question has no single answer, because not every skill carries the same amount of a person. I have two very different examples.</p>

<p>The first is by far the most used skill at the company where I work. It operates our project management tool: it creates cards, comments, updates statuses. I didn't build the first version. Someone created it on their own initiative, without anyone asking, and I have maintained it ever since. At first it was basically a guide to using the tool's API. The writing and communication standards came in later. And even those standards already existed in the team before they became a skill. So how much of it belongs to one person? Not much. The skill generates a lot of value, but what it encodes is a collective way of working.</p>

<p>The second example is the opposite. We have skills created by professionals who do outstanding work. They documented their own process of thinking and executing, built a workflow and turned it into a skill. That is individual. A few years ago, nobody would have had access to that method without mentoring or training from that person. The method was theirs.</p>

<p>And here is the uncomfortable part: <strong>in most cases, the company doesn't need a specialist. It needs the result.</strong> The person still knows how to do everything from scratch, without AI, but the rarest thing they had is now available to everyone, including whoever will take their seat.</p>

<h3>Research already asked the question</h3>

<p>This shift has been measured. Erik Brynjolfsson, Danielle Li and Lindsey Raymond followed more than 5,000 support agents who started using an AI assistant, in a study published in the <em>Quarterly Journal of Economics</em> in 2025. Productivity went up, and newcomers gained the most. The authors' explanation is almost the theme of this post: AI seems capable of <strong>"capturing and disseminating the behaviors of the most productive agents"</strong>.</p>

<p>The best agents supplied a good share of the examples used to train the system, and they were the ones who gained the least from it. The economists themselves write that their work raises questions about whether and how workers should be compensated for the data they provide. They also point to a risk: with AI suggesting everything, the best people may stop coming up with new solutions. Without people inventing, the next skill has nowhere to come from.</p>

<h3>"You're already paid for that"</h3>

<p>The most common answer in this debate is that the person already gets a salary, so whatever they create at work belongs to the company. With all due respect, that's a load of nonsense. Your salary pays you to do your job. If your job is checking invoices, turning your method into a skill was never part of the deal.</p>

<p>There is an exception, and I am it. I work on AI adoption, so writing and maintaining skills is part of my role, and part of my salary pays for that. Other people have no such obligation. When someone in finance, QA or sales packages their own method, they are delivering more than the role asks for.</p>

<p>Brazilian law, by the way, doesn't protect methods. Article 8 of the Brazilian Copyright Act says that ideas, systems and methods "as such" are not subject to protection. The text of the skill can belong to the company. The way of thinking inside it still belongs to whoever did the thinking. The bare minimum is that people can take their own method with them when they leave, as long as it carries no company secrets. The company keeps the copy. The person keeps what was always theirs.</p>

<p>Talking about who keeps the value without talking about who does the extracting would be dishonest of me. So let's move to my side of the table.</p>

<h2>I Hold the Stopwatch Too</h2>

<p>In Taylor's factory, the person who wrote down the worker's knowledge was the time-study man. He watched, measured and turned what he saw into written instructions. When I look at my own work honestly, the resemblance is there.</p>

<p>Part of what I do is help people extract the way they get things done and turn it into a skill. Another part is measuring: whether the skills are being used, and whether they are being used the right way. I explain how we organize and publish these skills in <a href="https://www.mmlo.com.br/blog/hub-de-skills">Skills Hub</a>.</p>

<p>The difference I can defend is in what I measure. I want to know whether a skill is useful, whether it saves work and whether it needs improving. I don't want to monitor who uses it or build a ranking of people.</p>

<p>The difference I can't guarantee is what happens next. <strong>I train people and I measure, but I'm not the one who decides what to do with it.</strong> The people above me decide. The same number that shows a skill saved ten hours a week can become free time for the team to think, or it can become an argument for not backfilling a position. I already argued which of those two paths I think is right in <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI Governance in Practice</a>. But arguing is not deciding, and anyone reading this post deserves to know that.</p>

<p>That math weighs heaviest on people who are just starting out. And that is where a question comes in that I think few people are asking the right way.</p>

<h2>Junior, Mid-Level and Senior Were Built for a Different Job</h2>

<p>The conversation about AI and careers usually stops at hiring juniors. And the data is worrying. I already wrote in <a href="https://www.mmlo.com.br/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais">AI and the Future of Professions</a> about the Stanford study on workers aged 22 to 25. The August 2026 revision pushed the gap to 19% below expected, almost all of it from less hiring rather than from layoffs. The authors warn that these are early indicators, not proof of causation. A Harvard study using data on 62 million workers found a similar pattern: at companies that adopted generative AI, junior employment fell and senior employment kept rising.</p>

<p>Not even the industry agrees on what to do. Matt Garman, CEO of AWS, said that replacing juniors with AI is <strong>"the dumbest thing I've ever heard"</strong>, because ten years from now nobody will have learned anything. Tobi Lütke, CEO of Shopify, told teams to prove AI can't do the job before asking for more headcount.</p>

<p>But I think the bigger question isn't whether companies will hire juniors. It's what "junior" will mean.</p>

<h3>Categories of autonomy</h3>

<p>Junior, mid-level and senior were created to rank people by how far they can go on their own in operational work:</p>

<ul>
  <li><strong>Junior:</strong> helps the company, but still needs help to do the day-to-day work.</li>
  <li><strong>Mid-level:</strong> does it alone and no longer needs support.</li>
  <li><strong>Senior:</strong> understands enough to guide other people.</li>
</ul>

<p>This logic is old. The Dreyfus brothers' model of skill acquisition, from 1980, goes from the novice, who relies on "context-free rules" and step-by-step instructions, to the expert, who acts on intuition and often can't explain what they do. Notice that the context-free rule, the lowest rung on the ladder, is exactly what a skill delivers ready-made. And the expert who can't explain what they do is Polanyi's tacit knowledge all over again.</p>

<p>The career handbook at 37signals, a company known for having little hierarchy, makes this even clearer. Its level matrix has an axis called <em>Trust</em>, which measures how much a person's work needs review: at level 1, everything is "thoroughly reviewed"; at level 4, work is "shipped without review". The document says nothing about AI. But that is exactly the axis AI scrambles, because today a large part of what needs review was written by an agent, not by a person.</p>

<h3>The operational work that ranked people is disappearing</h3>

<p>Models get better, <em>harnesses</em> (the environment built around the agent, with context, memory and checks) get more advanced, and better and better tools for security and quality keep showing up. The operational part that separated one level from the next is changing very fast.</p>

<p>At a retreat with senior engineers from large companies, Thoughtworks recorded the emergence of a "middle loop", supervisory work that sits between writing code and shipping it: directing, evaluating and fixing what agents produce. These are skills that, according to the report, are rarely <strong>"recognized in career ladders"</strong>. The same text says many people were hired to translate pre-digested tickets into code, and that <strong>"that work is disappearing"</strong>. The biggest risk, in their view, isn't even the junior. It's the mid-level. Martin Fowler, who attended the retreat, summed it up: "It's the mid-level developers who face the greatest challenges."</p>

<p>It makes sense. If mid-level was defined by "does it alone", and the agent also does it alone, the definition loses its footing. GitLab admitted something similar in an official document: its current interview process primarily measures skills that "AI has commoditised", and they are redesigning hiring to assess judgment and validation discipline. Addy Osmani described the effect on people starting out in one sentence: <strong>"Taste used to be a byproduct of the reps. Agents took the reps."</strong></p>

<h3>Nobody is proposing to throw out the ruler</h3>

<p>The curious part is that, as far as the research for this post could see, nobody with real weight is proposing to abolish junior, mid-level and senior because of AI. What exists is people asking to redefine the levels. In the abstract for his talk at QCon London this year, Alasdair Allan wrote that progression frameworks, built for a world of gradual skill accumulation, "no longer match reality". Ossama Chaib called for defining what each level means in a world with agents and readjusting the pay bands. Deloitte was already arguing in 2022 that jobs are not the best way to organize work, but it kept levels in its own proposal, and AI was only the third reason on its list.</p>

<p>In Brazil, that change is even harder. Under article 461 of the CLT, Brazil's main labor code, a formal career structure or a job and salary plan is what lets a company pay people in the same role differently without running into equal pay claims, and that plan can come from internal rules or from collective bargaining. Changing the categories means changing the plan.</p>

<p>I don't have the authority to say how this will turn out. But I believe it will change, because the work is changing and the way we classify work will have to keep up. Operationally, an AI with access to all of a company's documentation knows more than any single person there. That's a fact. Whether it does things the best way is another conversation.</p>

<p>The question I want to leave is about the expert. Will the expert be the person who carries out a task brilliantly? Or the person who reinvents how to use the tools the company already has, the ones AI already masters, and turns that into a process others can repeat? The Dreyfus brothers had a sixth stage, above the expert, for those who "transform the style of their skill domain". If that is the answer, the expert of the future is precisely the person who writes the skills. And then the question of who keeps the value stops being a side issue and becomes the central question of a career.</p>

<h2>What We Can Demand</h2>

<p>If the problem is structural, the way out can't depend on the goodwill of each manager. I see three paths, from the simplest to the hardest.</p>

<h3>1. Authorship, at the very least</h3>

<p>Authorship and recognition are the basics. Whoever created a skill has to be able to say on LinkedIn and in their portfolio that it is theirs, and the company has an obligation to confirm that the achievement belongs to the person, not to the company. That includes the right to take the method with them, as I said above, when there are no company secrets involved. In <a href="https://www.mmlo.com.br/blog/o-manifesto-do-desenvolvedor">The Developer's Manifesto</a> I wrote about signing what we make. A skill is one of the things that deserve a signature.</p>

<h3>2. Rewards proportional to use</h3>

<p>A skill with heavy adoption deserves a bonus, as something extra. And someone who keeps delivering new methods all the time is delivering far more than the role asks for, so the fair thing is a raise.</p>

<p>Measuring this is hard, and I'm not going to pretend it isn't. How many hours did a skill save? How much rework did it prevent? How much money did it bring in? Does the reward come as time, a percentage, a raise or a bonus? These are open questions. But there are models to copy:</p>

<ul>
  <li><strong>In Germany</strong>, the Employee Inventions Act guarantees "adequate" compensation when a company uses an employee's invention, and calculates the amount based on economic use, the person's role and the company's share in the result. It even applies to technical improvements that can't be patented, as soon as the company starts using them. It is almost the yardstick I described above: the individual weight varies, and whoever has that as part of their job gets less.</li>
  <li><strong>In Brazil</strong>, the Industrial Property Act, in article 89, already allows a company to give an employee inventor a share of the gains from a patent, through negotiation or internal rules, without that counting as salary. The design exists. It just hasn't reached skills.</li>
  <li><strong>At one company</strong>, Omnisend started giving quarterly raises of 2% to 4% to people who use AI with business impact, including processes that other colleagues and teams go on to adopt. The executive himself admitted to Fortune that measuring that impact is "honestly really difficult". The difficulty didn't stop the company from trying.</li>
</ul>

<h3>3. Collective agreements</h3>

<p>The strongest path is the one that doesn't depend on a single company. In 2023, Hollywood screenwriters went on strike and came out with a contract stating that AI-generated text doesn't count as literary material, that AI-generated material can't be used to reduce a writer's pay or credit, that the company can't force anyone to use AI, and that the union reserves the right to challenge the use of writers' work to train AI. Authorship, compensation and collective bargaining in the same document.</p>

<p>For people working in tech in Brazil, this still seems far off. But the knowledge that becomes a skill is the same kind as the knowledge that becomes a screenplay: it is the way someone thinks about the work. If screenwriters managed to put that in a contract, we can put it in collective agreements and in job and salary plans too.</p>

<h2>The Asset Still Belongs to Whoever Knows</h2>

<p>Taylor was right about one thing: the knowledge of a trade is the principal asset of the people who work. The difference is that, in 1911, taking that asset required a stopwatch, a foreman and years of observation. Today all it takes is a text file and an afternoon. And what goes into the file now isn't just the movement of the hand. It's the way of thinking.</p>

<p>That doesn't make the skill an enemy. Used to clear boring work out of the way, it gives time back and opens room for new things. Used as an assembly line, it teaches people to know less and less. The technology is the same in both cases. What changes is who decides and who keeps the result.</p>

<p>The skill can stay with the company. The knowledge has to keep belonging to whoever knows.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://www.gutenberg.org/ebooks/6435">Frederick W. Taylor, <em>The Principles of Scientific Management</em> (1911), Project Gutenberg</a></li>
  <li><a href="https://press.uchicago.edu/ucp/books/book/chicago/T/bo6035368.html">Michael Polanyi, <em>The Tacit Dimension</em> (1966), University of Chicago Press</a></li>
  <li><a href="https://www.nber.org/papers/w20485">David Autor, "Polanyi's Paradox and the Shape of Employment Growth", NBER (2014)</a></li>
  <li><a href="https://isj.org.uk/marx-and-braverman/">Bob Carter, "Defending Marx and Braverman", International Socialism (2021), on Harry Braverman, <em>Labor and Monopoly Capital</em> (1974)</a></li>
  <li><a href="https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills">Anthropic, "Equipping agents for the real world with Agent Skills" (2025)</a></li>
  <li><a href="https://www.anthropic.com/research/anthropic-economic-index-january-2026-report">Anthropic, "Anthropic Economic Index report: Economic primitives" (Jan. 2026)</a></li>
  <li><a href="https://www.technologyreview.com/2026/04/20/1136149/chinese-tech-workers-ai-colleagues/">MIT Technology Review, "Chinese tech workers are starting to train their AI doubles, and pushing back" (Apr. 2026)</a></li>
  <li><a href="https://www.thoughtworks.com/content/dam/thoughtworks/documents/radar/2026/11/tr_technology_radar_vol_35_en.pdf">Thoughtworks, Technology Radar vol. 35 (Nov. 2026)</a></li>
  <li><a href="https://arxiv.org/abs/2304.11771">Erik Brynjolfsson, Danielle Li and Lindsey Raymond, "Generative AI at Work", Quarterly Journal of Economics (2025)</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/leis/l9610.htm">Law 9,610/1998 (Brazilian Copyright Act), art. 8</a></li>
  <li><a href="https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine/">Brynjolfsson, Chandar and Chen, "Canaries in the Coal Mine?", Stanford Digital Economy Lab (rev. Aug. 2026)</a></li>
  <li><a href="https://digitaleconomy.stanford.edu/event/seyed-m-hosseini-and-guy-lichtinger-generative-ai-as-seniority-biased-technological-change-evidence-from-u-s-resume-and-job-posting-data">Hosseini and Lichtinger, "Generative AI as Seniority-Biased Technological Change" (2025)</a></li>
  <li><a href="https://www.theregister.com/2025/08/21/aws_ceo_entry_level_jobs_opinion/">The Register, "AWS CEO says using AI to replace junior staff is 'Dumbest thing I've ever heard'" (Aug. 2025)</a></li>
  <li><a href="https://techcrunch.com/2025/04/07/shopify-ceo-tells-teams-to-consider-using-ai-before-growing-headcount">TechCrunch, "Shopify CEO tells teams to consider using AI before growing headcount" (Apr. 2025)</a></li>
  <li><a href="https://en.wikipedia.org/wiki/Dreyfus_model_of_skill_acquisition">Dreyfus model of skill acquisition (Dreyfus and Dreyfus, 1980)</a></li>
  <li><a href="https://github.com/basecamp/handbook/blob/master/titles-for-programmers.md">37signals, "Titles for Programmers", Employee Handbook</a></li>
  <li><a href="https://www.thoughtworks.com/content/dam/thoughtworks/documents/report/tw_future%20_of_software_development_retreat_%20key_takeaways.pdf">Thoughtworks, "The future of software engineering: Retreat findings and strategic insights" (Feb. 2026)</a></li>
  <li><a href="https://www.martinfowler.com/fragments/2026-02-13.html">Martin Fowler, "Fragments: February 13" (2026)</a></li>
  <li><a href="https://handbook.gitlab.com/handbook/company/working-groups/ai-native-hiring/">GitLab Handbook, "AI-Native Hiring Working Group" (2026)</a></li>
  <li><a href="https://addyosmani.com/blog/earning-judgment/">Addy Osmani, "Earning taste and judgment" (Jul. 2026)</a></li>
  <li><a href="https://qconlondon.com/presentation/mar2026/ladder-missing-rungs-engineering-progression-when-ai-ate-middle">Alasdair Allan, "The Ladder Is Missing Rungs", QCon London (Mar. 2026)</a></li>
  <li><a href="https://ossama.is/blog/ladder">Ossama Chaib, "AI flattened the engineering ladder" (Feb. 2026)</a></li>
  <li><a href="https://www.deloitte.com/content/dam/assets-shared/legacy/docs/Deloitte-Skills-Based-Organization.pdf">Deloitte, "Building tomorrow's skills-based organization: Jobs aren't working anymore" (2022)</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2017/lei/l13467.htm">Law 13,467/2017, new wording of art. 461 of the CLT (Brazilian Consolidated Labor Laws)</a></li>
  <li><a href="https://www.gesetze-im-internet.de/arbnerfg/__9.html">Germany, Arbeitnehmererfindungsgesetz (Employee Inventions Act), §§ 9 and 20</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/leis/l9279.htm">Law 9,279/1996 (Brazilian Industrial Property Act), art. 89</a></li>
  <li><a href="https://fortune.com/2026/06/09/omnisend-quarterly-raises-employees-who-use-ai-to-drive-business-impact/">Fortune, "Omnisend gives quarterly raises to employees who use AI to drive business impact" (Jun. 2026)</a></li>
  <li><a href="https://www.wga.org/contracts/know-your-rights/artificial-intelligence">Writers Guild of America, "Artificial Intelligence" (2023 contract)</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 08 Nov 2026 00:03:52 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Futuro do Trabalho</category>
      <category>Tecnologia e Sociedade</category>
      <category>Automação</category>
      <category>Mercado de Trabalho</category>
      <category>Ética na Tecnologia</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/0e80d380-6b51-4ea3-8ea7-f98d029cfcfd" medium="image"/>
    </item>
    <item>
      <title>Why AI Projects Fail</title>
      <link>https://www.mmlo.app/en/blog/por-que-projetos-de-ia-falham</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/por-que-projetos-de-ia-falham</guid>
      <description>Selling Productivity, Not Replacement</description>
      <content:encoded><![CDATA[<p class="intro">The number shows up in every talk about enterprise AI: 95% of projects fail. On the first day of <a href="https://experience2026.aibrasil.ai/programacao">AI Brasil Experience 2026</a>, on October 28, I heard a reading of that number that fits what I live as head of AI Enablement at InBot. The cause is rarely the model. It is what the company promises people when it shows up with AI. Promise cuts and you get resistance. Promise time back and you get allies. In this post I share what I heard on stage and in a conversation afterwards, explain what the 95% actually measures, and show why adopting AI in-house is harder for a company that existed before it, even when that company sells AI. At the end, I leave five practices for your team.</p>

<h2>What I Heard at AI Brasil: Replacement Sold Better Than Reality</h2>

<p>The talk was called <strong>"Quando a máquina trabalha pra gente?"</strong> ("When does the machine work for us?"), given by <strong>Ricardo Heidorn</strong>, CEO and co-founder of Leads Per Hour, a company that sells AI agents for sales, and CRO at Lincros. Keep that detail in mind: this is someone who makes money from AI saying that the most common way of selling AI is wrong.</p>

<p>He opened with a provocation: <strong>the promise that AI would replace humans was more successful than the results it delivered</strong>. Every launch brought a new scare, and the scare reached the headlines long before it reached the balance sheets. Later came the line that gives this post its title: <strong>selling productivity works; selling replacement does not</strong>.</p>

<p>The retreats of those who bet on cutting people are already documented, and I listed some of them in <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI Governance in Practice</a>. A new data point comes from the very report that made the 95% famous: among the myths it debunks is the idea that AI will take most jobs within a few years. The researchers found few layoffs caused by generative AI.</p>

<p>If the cuts did not come, why do so many AI projects stall? Before answering, it is worth understanding what that 95% counts.</p>

<h2>Do 95% of AI Projects Fail? What the Number Measures</h2>

<p>The statistic comes from <em>The GenAI Divide: State of AI in Business 2025</em>, a preliminary report by MIT NANDA published in July 2025. The version that circulates is "95% of AI projects fail". The report's version is narrower: among custom-built enterprise AI tools, <strong>only 5% reached production with a clear and sustained impact on results</strong>, assessed about six months after the pilot. The funnel narrows fast: 60% of companies evaluated a tool of this kind, 20% ran a pilot and 5% put it into production.</p>

<p>The study has real limits. It is based on 52 interviews and 153 surveys, with self-reported answers and a demanding, loosely defined bar for success. Rob Wiblin, from 80,000 Hours, ran the math from the other side: among the companies that reached a pilot, roughly one in four succeeded. That is why I treat the 95% as a sense of direction, not as a scoreboard.</p>

<h3>The problem is not the model, it is the approach</h3>

<p>The part worth reading is the authors' diagnosis:</p>

<blockquote><p><strong>"This divide does not seem to be driven by model quality or regulation, but seems to be determined by approach."</strong></p></blockquote>

<p>What separates those who get results from those who do not is neither model quality nor regulation, but approach. The report's list of barriers includes tools that do not fit the workflow, people unwilling to adopt new tools, poor change management and a lack of executive sponsorship. As Fortune put it, the technology was not the problem; the way companies were using it was.</p>

<p>Heidorn called this <strong>culture</strong>, and that is the word I would use too. Microsoft and LinkedIn's 2026 Work Trend Index surveyed 20,000 AI users in 10 countries, Brazil among them, and found a telling ratio: in the report's statistical model, organizational factors such as culture and manager support account for <strong>67%</strong> of the impact people report from AI, and individual factors for <strong>32%</strong>. It is an association, not a cause, as the report itself warns. Only 26% see their leadership aligned on the subject. Since Microsoft has a direct interest in adoption, I read the number with that caveat, but it points to the same place as MIT.</p>

<h3>Tools became a commodity, culture did not</h3>

<p>Another remark from the talk: <strong>whatever is good ends up becoming a commodity, and AI will be no different</strong>. Prices already show it. Running a GPT-3.5-level model has become hundreds of times cheaper since 2022, according to Stanford's AI Index. Any competitor of yours can subscribe to the same model tomorrow. What they cannot subscribe to is a team that knows how to get value out of it.</p>

<p>Heidorn also pointed out something that often goes unnoticed: <strong>most people do not know what AI is</strong>, even inside tech companies. Without that basic explanation, the empty space gets filled by whatever is in the headlines. And what is in the headlines, almost always, is fear.</p>

<h2>Fear Breeds Sabotage: Resistance to AI Inside the Team</h2>

<p>Heidorn's line was short: <strong>some people sabotage AI because they are afraid of losing their jobs</strong>. It is a predictable reaction: if the tool arrived announced as your replacement, helping it succeed feels like digging your own grave.</p>

<p>Perception data confirms how big that fear is. Pew Research Center found that <strong>52%</strong> of U.S. workers are worried about AI's impact on work. In BCG's AI at Work 2025, <strong>41%</strong> think their own job may disappear within ten years. And sabotage is not a figure of speech: in a Writer survey, <strong>41% of millennial and Gen Z employees</strong> admitted to sabotaging their company's AI strategy, for example by refusing to use the tools. Writer sells AI and the sample covers only two generations, but the behavior is the one Heidorn described.</p>

<p>Gallup adds an important nuance: only <strong>15%</strong> of workers think it is likely they will lose their job to AI within the next five years. General concern is much higher than the real expectation of being laid off. In my reading, this means that what fuels resistance is the <strong>story being told</strong> about AI, more than the risk each person sees for themselves.</p>

<p>Duolingo learned this in public. In 2025, a memo from the CEO declared the company "AI-first" and said contractors would gradually be replaced by AI wherever it could do the work. The backlash was so bad that the CEO himself softened the tone and admitted: "I did not give enough context". No full-time employee was laid off, but the message that stuck was the memo's.</p>

<h3>Companies born with AI vs. companies that already existed</h3>

<p>After the talk, I went to speak with Heidorn in person. I wanted to know what he did with AI internally, and the answer came in two parts, one for each company.</p>

<p>At <strong>Leads Per Hour</strong>, internal adoption was natural. The company was born <em>AI-first</em>, so nobody had to change the way they worked: that already was the way they worked. At <strong>Lincros</strong>, it was much harder. It is a company that existed before generative AI, with a team that had its own processes and more resistance to change. There, he said, adoption took <strong>a lot of top-down pressure</strong>.</p>

<p>That conversation helped me see where InBot stands. At first glance, we look like the first case: InBot is an AI company. It has built conversational AI since long before the term became fashionable. <strong>Robô Ed</strong>, created by InBot and launched in 2004 for a Petrobras campaign about energy consumption, talked in natural language with children and teenagers and stayed online until 2016. But that is the product InBot delivers to its clients. <strong>Using AI in our own work is something else</strong>, and on that front InBot looks more like Lincros: a team with years of established processes that already worked well the old way.</p>

<p>Here, the beginning came from the bottom, with developers adopting AI before there was a dedicated area for it (I told that part in the governance post). The challenge now is different: <strong>how to take that culture to the whole company without the top-down pressure</strong> that Lincros needed. In other words, moving forward in the least aggressive way possible.</p>

<h3>Culture starts at the base, not at the top</h3>

<p>In the same conversation, Heidorn gave a hint for that path that was not on his slides: <strong>find people at the base of each team</strong>. Those who work in operations know the bottlenecks much better than leadership does. I asked whether he meant <em>AI champions</em>, people on each team who drive AI use among their peers, and he confirmed that was exactly it.</p>

<p>A champion solves two problems at once. They know where the team's week gets stuck, so they point AI at the right place. And they talk to colleagues as a colleague, which takes away the top-down, order-from-above tone of the change. It is the opposite of top-down pressure, and it is what makes it possible to move forward without running anyone over.</p>

<p>Having AI champions at InBot is an idea I already had, and it was put on hold at the time. I left that conversation wanting to put it back on the table.</p>

<h3>Resistance is not only fear of losing your job</h3>

<p>There is a kind of resistance that the fear explanation does not cover, and I live with it. To this day, some people on the development team avoid using the corporate <a href="https://www.mmlo.com.br/blog/hub-de-skills">Skills Hub</a> because it downloads things onto their computers.</p>

<p>I understand the distrust, and the data backs it up. In Stack Overflow's 2025 survey, <strong>81% of developers</strong> cite security and privacy as a concern when using AI agents. In August 2025, the attack known as <em>s1ngularity</em> showed why: tampered versions of a popular npm package triggered the AI command-line tools installed on victims' machines to hunt for credentials, and more than 2,000 GitHub accounts were exposed. Someone who asks what is being installed on their own computer is doing exactly what a good developer should do.</p>

<p>This kind of resistance is not solved by an order, and it should not be. The path I see is to open the box: show what the Hub installs, where it comes from, who signed the release and who reviewed it before publishing. And accept that convincing people takes time. Culture is not a stage you complete. It is an ongoing conversation.</p>

<p>Fear and distrust explain what gets stuck. What gets things moving, from what I have seen so far, is changing the promise.</p>

<h2>Selling Productivity Means Giving Time Back</h2>

<p>On stage, Heidorn presented Leads Per Hour's mission like this: <strong>give back to humans what belongs to humans</strong>. For a person, the scarcest resource is time, and an AI agent is only worth it if it gives some of that time back.</p>

<p>This is where the two narratives part ways. The replacement narrative speaks to finance and scares everyone else: "we will need fewer people" reaches each person's ears as "maybe they won't need you". The productivity narrative speaks to the people who do the work: "that tedious part will take minutes, and you decide what to do with the rest of the afternoon". Only the second gives people a reason to root for AI.</p>

<p>The MIT report has a finding that reinforces this. Most of the users interviewed welcomed automation, especially for manual and repetitive tasks, on two conditions: <strong>secure data and measurable results</strong>. Those are the same two things the developers who distrust the Hub are asking for.</p>

<h3>The promise needs proof</h3>

<p>Saying "nobody will be replaced" is easy. The hard part is for that sentence to still hold six months later. What builds credibility is proof: at InBot, automations have already turned hours-long tasks into minutes, and nobody left because of them (the numbers are in the governance post). When the colleague next to you gains a free afternoon and stays on the team, the message stops being just talk. To see how this plays out beyond software development, I wrote about <a href="https://www.mmlo.com.br/blog/skills-de-ia-por-area">AI skills by area</a>.</p>

<p>The other half of the argument is still missing: what happens to results when the human stays in the process.</p>

<h2>Human + AI Delivers More Than AI Alone</h2>

<p>During the Q&amp;A, Fernando Sapata, who had spoken earlier on the same stage, asked Heidorn what lead generation will look like going forward. The answer came with numbers from Leads Per Hour, which I wrote down on the spot and share as an account of the event, not as research:</p>

<ul>
  <li>in lead capture, the conversion rate was <strong>1.6%</strong> with AI working alone and <strong>3%</strong> with a human alongside it;</li>
  <li>in conversations with customers, AI alone succeeded in <strong>1 out of every 100</strong> attempts; with a person in the middle of the process, receiving leads that were already warm, it was <strong>20 out of every 100</strong>.</li>
</ul>

<p>In his view, the market still wastes hours on forms, and that is the part AI will change. The machine handles the repetitive triage and hands the person the conversation that is worth having.</p>

<p>Public data points the same way. Fang and co-authors tested a pre-sales assistant on a large online retail platform: compared with no assistance, the AI-only version increased sales by <strong>16.3%</strong>; when the AI could hand the conversation over to a human, the increase was <strong>25%</strong>. It is a preprint, not yet peer-reviewed. The NBER study on customer support, which I have cited on this blog before, measured AI working alongside human agents, and the gains were largest among the least experienced.</p>

<p>The point is not to put a human at the end of the line to check the machine's work. It is to design the process with the person at the center: AI takes the volume and the repetition, and the human keeps the judgment, the relationship and the decision.</p>

<h2>How to Change the AI Conversation in Your Team</h2>

<p>None of this requires a director title. It requires choosing carefully what you say and what you show. I left the event with five practices in mind:</p>

<ul>
  <li><strong>Look for champions at the base.</strong> Almost every team already has someone who uses AI on their own and knows where operations get stuck. Give that person room and recognition to bring their colleagues along.</li>
  <li><strong>Talk about time, not cuts.</strong> Replace "what AI will replace" with "how much time it gives back, and for what".</li>
  <li><strong>Measure together with the person who does the task.</strong> Time the task with the person who does it, before AI and after it. That way the gain belongs to them, not to a management spreadsheet.</li>
  <li><strong>Open the box for those who distrust it.</strong> Show what the tool accesses, what it installs and who reviewed it. A security question deserves an answer, not pressure.</li>
  <li><strong>Show the promise being kept.</strong> When someone gains hours in their week, tell the team, and also tell them that this person is still there.</li>
</ul>

<p>Communicating well is still rare, which makes it an advantage. According to Gallup, only <strong>22%</strong> of workers say their company has shared a clear AI plan, and those who received one are <strong>3 times more likely</strong> to feel very prepared to use the technology.</p>

<h2>Conclusion</h2>

<p>The 95% is often used as proof that enterprise AI does not work. Read carefully, the report says something else: the technology is available to everyone, and the difference lies in how each company presents it and fits it into people's work.</p>

<p>In a company born with AI, that culture comes for free. In a company that already existed, even one that has been selling AI for more than twenty years, like InBot, culture has to be built, and you can build it with top-down pressure or with people at the base bringing their colleagues along. I am betting on the second, with transparency for those who distrust it and with time given back as the promise.</p>

<p>If you want to start this week, pick a repetitive task, measure it with the person who does it and show the team how much time came back to that person. And notice who on your team is already doing this without anyone asking. That is probably where your first champion is.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://experience2026.aibrasil.ai/palestrantes">AI Brasil. "Palestrantes AI Brasil Experience 2026" (Speakers).</a></li>
  <li><a href="https://experience2026.aibrasil.ai/programacao">AI Brasil. "Programação AI Brasil Experience 2026" (Schedule).</a></li>
  <li><a href="https://webberwentzel.com/News/Documents/2025/MIT1757411281972.pdf">Challapally, A.; Pease, C.; Raskar, R.; Chari, P. "The GenAI Divide: State of AI in Business 2025". MIT NANDA, 2025.</a></li>
  <li><a href="https://fortune.com/2025/08/21/an-mit-report-that-95-of-ai-pilots-fail-spooked-investors-but-the-reason-why-those-pilots-failed-is-what-should-make-the-c-suite-anxious/">Kahn, Jeremy. "An MIT report that 95% of AI pilots fail spooked investors...". Fortune, 2025.</a></li>
  <li><a href="https://80000hours.org/podcast/episodes/ai-workplace-mit-study/">Wiblin, Rob. Episode on the MIT study and AI at work. 80,000 Hours, 2026.</a></li>
  <li><a href="https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization">Microsoft and LinkedIn. "2026 Work Trend Index Annual Report".</a></li>
  <li><a href="https://hai.stanford.edu/ai-index/2025-ai-index-report">Stanford HAI. "AI Index Report 2025".</a></li>
  <li><a href="https://www.pewresearch.org/social-trends/2025/02/25/u-s-workers-are-more-worried-than-hopeful-about-future-ai-use-in-the-workplace/">Pew Research Center. "U.S. Workers Are More Worried Than Hopeful About Future AI Use in the Workplace", 2025.</a></li>
  <li><a href="https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain">BCG. "AI at Work 2025: Momentum Builds, but Gaps Remain".</a></li>
  <li><a href="https://writer.com/blog/enterprise-ai-adoption-survey/">Writer and Workplace Intelligence. "2025 Enterprise AI adoption survey".</a></li>
  <li><a href="https://www.gallup.com/workplace/691643/work-trends-ai-adoption-continues-rise.aspx">Pendell, Ryan. "AI Use at Work Has Nearly Doubled in Two Years". Gallup, 2025.</a></li>
  <li><a href="https://fortune.com/2025/08/18/duolingo-ceo-admits-controversial-ai-memo-did-not-give-enough-context-insists-company-never-laid-off-full-time-employees/">Coacci, Jessica. "Duolingo CEO admits controversial AI memo did not give enough context...". Fortune, 2025.</a></li>
  <li><a href="https://www.bandab.com.br/nacional/robo-ed-historia-chatbot-brasil/amp/">Band B. "Lembra do Robô Ed? Lançado em 2004, 'ChatGPT brasileiro' tinha 7 funcionários e 35 mil respostas" (in Portuguese), 2023.</a></li>
  <li><a href="https://canaltech.com.br/inteligencia-artificial/robo-ed-e-mais-relembre-5-chatbots-que-surgiram-antes-do-chatgpt/">Canaltech. "Robô Ed e mais: relembre 5 chatbots que surgiram antes do ChatGPT" (in Portuguese).</a></li>
  <li><a href="https://survey.stackoverflow.co/2025/ai">Stack Overflow. "2025 Developer Survey: AI".</a></li>
  <li><a href="https://www.bleepingcomputer.com/news/security/ai-powered-malware-hit-2-180-github-accounts-in-s1ngularity-attack/">Toulas, Bill. "AI-powered malware hit 2,180 GitHub accounts in s1ngularity attack". BleepingComputer, 2025.</a></li>
  <li><a href="https://arxiv.org/abs/2510.12049">Fang, L. et al. "Generative AI and Sales Productivity: Field Experiments in Online Retail". arXiv, 2025.</a></li>
  <li><a href="https://www.nber.org/papers/w31161">Brynjolfsson, E.; Li, D.; Raymond, L. "Generative AI at Work". NBER Working Paper 31161.</a></li>
  <li><a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">Mierzwa, Matheus. "AI Governance in Practice".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/skills-de-ia-por-area">Mierzwa, Matheus. "AI Skills by Area".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/hub-de-skills">Mierzwa, Matheus. "Skills Hub".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Thu, 29 Oct 2026 00:35:26 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Transformação Digital</category>
      <category>Liderança Técnica</category>
      <category>Produtividade</category>
      <category>Futuro do Trabalho</category>
      <category>Governança de IA</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/68db1cc6-1d8b-44cb-8c2c-7af10a3aba8d" medium="image"/>
    </item>
    <item>
      <title>Technology with Social Purpose</title>
      <link>https://www.mmlo.app/en/blog/tecnologia-com-proposito-social-como-desenvolvedores-podem-contribuir-para-um-mundo-melhor</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/tecnologia-com-proposito-social-como-desenvolvedores-podem-contribuir-para-um-mundo-melhor</guid>
      <description>Whoever Defines the Problem Owns the Solution</description>
      <content:encoded><![CDATA[<p class="intro">For my TCC, my undergraduate capstone project, my group and I built a system to connect animal rescue NGOs with people interested in adopting. The pilot worked. And the project stopped right there, because a system needs maintainers, and none of us managed to keep it going. That was the first time I understood that technology with social purpose depends on more than good code. In this post I separate corporate "tech for good" from impact that actually changes something, lay out four concrete paths for a developer, share what I learned by doing, and end in the place almost nobody looks: your own job.</p>

<h2>Tech for Good or Social Technology?</h2>

<p>Every big tech company now has a social impact page. "AI for good" programs, annual sustainability reports, hackathons with nonprofits. Some of it helps someone. But a lot of it is marketing, and the difference is not always in the intentions of the people doing it.</p>

<p>Researcher Ben Green summed up the problem in a 2019 paper: computing's efforts toward "social good" lack a solid grounding in politics or in theories of social change. They rely on vague notions of what is "good" and move forward through small reforms, always with technology at the center. Nobody asks who decided that this was the problem.</p>

<p>In Brazil, that question has a tradition of its own. Renato Dagnino, a professor at Unicamp, uses the term <em>social technology</em> for technology developed with a community and controlled by it, not simply delivered to it. He criticizes conventional technology precisely because it takes control over the work process away from the worker:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"[Conventional technology], by not allowing the direct producer to control the work process, always makes necessary a boss, a capitalist, a manager, a foreman, or, in more modern times, an engineer."</strong></p>
</div>

<p>Look at the end of that sentence. In conventional technology, the engineer, and today the developer, is the one who decides on behalf of the people who will use it. That is why I use two tests to tell impact apart from marketing.</p>

<ul>
  <li><strong>Does it survive without the company?</strong> Real impact continues when the sponsorship ends. Marketing disappears along with the campaign.</li>
  <li><strong>Whoever decides, controls.</strong> If the company defines the problem, it also keeps the solution, the data and the credit. If the community does not decide, it does not control.</li>
</ul>

<p>Both tests apply to companies, but they apply to us too. A volunteer developer who shows up at a nonprofit with a ready-made solution repeats the same mistake on a smaller scale.</p>

<h2>Four Ways to Use What You Know</h2>

<p>If your question is "where do I start?", these are the four paths I recommend. None of them requires quitting your job.</p>

<h3>1. Websites and systems the nonprofit can maintain</h3>

<p>According to the TIC Organizações Sem Fins Lucrativos 2025 survey, Cetic.br's study of Brazilian nonprofits, only <strong>46% of nonprofit organizations</strong> in Brazil have a website or web page. Among those that do, 10% had their site built by volunteers, and volunteers run the social media accounts in 29% of them. There is room, and a lot of it.</p>

<p>But the criterion is not what you know how to build, it is what the nonprofit can maintain after you leave. A simple site, in a tool the team can edit on its own, is worth more than a custom system that only you understand.</p>

<h3>2. Accessibility</h3>

<p>The WebAIM Million 2026 report analyzed the home pages of the 1 million most visited websites in the world: <strong>95.9% had automatically detectable accessibility failures</strong>, with an average of 56.1 errors per page. Portuguese-language pages fare worse: 61.1 errors on average, and the .br domain reaches 62.5. Six basic error types, such as low contrast, images without alt text and form fields without labels, account for 96% of everything.</p>

<p>Brazil has 14.4 million people with disabilities, according to the 2022 Census, and the share reaches 27.5% among people aged 70 or older. Brazil's Inclusion Law (Lei Brasileira de Inclusão, or LBI) requires company and government websites to be accessible. Accessibility is not a favor: it is the law, and it is your audience.</p>

<h3>3. Teaching technology</h3>

<p>Almost every developer underestimates the value of teaching the basics. TIC Domicílios 2025, Cetic.br's household internet survey, shows that only <strong>54% of people aged 60 or older</strong> use the internet, compared with 85% of the general population. Of the 28 million Brazilians who do not use the internet, 16 million are 60 or older. Before any app, there is an access barrier, and it will not be solved by yet another app.</p>

<h3>4. Free software, open data and civic tech</h3>

<p>There are public-interest projects in Brazil that accept contributions from developers today:</p>

<ul>
  <li><strong>Brasil.IO:</strong> makes Brazilian public data accessible. The project's manifesto says that "restricting access to public data is making democracy elitist". During the pandemic, a task force of 40 volunteers compiled the bulletins from the state health departments.</li>
  <li><strong>Querido Diário:</strong> an Open Knowledge Brasil project that makes municipal official gazettes searchable. Its documentation has a dedicated track for people who want to start contributing.</li>
  <li><strong>Software Público Brasileiro:</strong> the Brazilian public administration's free software catalog, with 81 systems.</li>
</ul>

<p>What these projects have in common is that they do not belong to a company. They do not depend on sponsorship to exist, but they do depend on people to keep going, and that is where you come in.</p>

<h2>What I Learned by Doing</h2>

<p>Before technology, my school of community organizing was Rotaract Club Barueri, where I was Director of Public Image in the 2019-2020 term, along with projects like planting more than a thousand trees in the Baixada Santista region. Then came the projects with code: the animal rescue NGO capstone at Fatec de Barueri, free redesigns of websites for nonprofits and social movements, and basic computer classes for older adults. Those projects left me with four lessons.</p>

<p><strong>The problem was not just technical.</strong> Process, time and people weighed more than the system. The code was the easy part.</p>

<p><strong>Without someone to maintain it, it dies.</strong> The capstone got the top grade and a pilot that worked. Even so, it did not continue: a system needs maintainers, and at the time none of us managed to keep it going. Today I know that maintenance is not a detail of the project: it is the project.</p>

<p><strong>I learned more than I taught.</strong> Listening to the people who would use it changed the way I build software to this day. What seemed obvious to me was rarely what the person on the other side needed.</p>

<p><strong>Access is the first barrier.</strong> In the classes for older adults, it became clear that, before any app, what is missing is knowing how to use it and having access. No digital solution reaches people who have not yet reached the digital world.</p>

<h2>The Traps for People Who Want to Help</h2>

<p>Good intentions protect no one from these four traps. I have fallen into a few of them.</p>

<h3>A project with no maintainer</h3>

<p>This was my capstone's mistake, and it is not just a student mistake. Software Público Brasileiro, created in 2007, started having technical problems in 2009, and maintaining the portal became harder and harder until it was rebuilt by UnB and USP between 2014 and 2016. If even a federal government project degrades without someone to look after it, a system built over a weekend for a nonprofit degrades even faster. It is the same problem I discuss in <a href="https://www.mmlo.com.br/blog/o-manifesto-do-desenvolvedor">The Developer's Manifesto</a>: infrastructure everyone uses and almost nobody sustains.</p>

<h3>Solutionism</h3>

<p>Evgeny Morozov coined the term <em>solutionism</em> for the temptation to treat political and social problems as if they were technical questions of efficiency. A donation app does not fix the lack of public policy for abandoned animals. A scheduling system does not fix the shortage of doctors. Technology can help within a problem, but it is rarely the answer to it.</p>

<h3>Exploited unpaid labor</h3>

<p>Carlos Montaño, a professor at UFRJ, shows how the discourse of the "third sector" helped justify the state's withdrawal from areas like social security. The critique is not aimed at the people who work in NGOs, who often do what nobody else will. It is aimed at the model in which volunteers cover, for free, what should be public policy or paid work. When a for-profit company asks for "volunteers" to build its social project, be suspicious.</p>

<h3>Doing "for" instead of "with"</h3>

<p>Dagnino argues that technology should be produced with the participation of the people who will use it, and calls the idea of bringing science "to" citizens naive. In practice: before writing a single line of code, ask what the organization needs, what it already uses, and who will look after whatever you deliver. If you cannot answer that last question, it is not yet time to start.</p>

<h2>The Impact That Fits in Your Job</h2>

<p>The place where a developer has the most social impact is not the weekend project. It is the everyday work, which reaches thousands or millions of people. Four decisions fit there.</p>

<ul>
  <li><strong>Ship accessible work even when nobody asks for it.</strong> Contrast, alt text and form labels do not need anyone's approval. WebAIM's six most common errors are in the hands of whoever writes the front end.</li>
  <li><strong>Collect only what is necessary.</strong> The LGPD, Brazil's data protection law, has a principle called necessity: limiting data processing to the minimum required for its purpose. When someone asks for "one more field, just in case", that principle is your argument.</li>
  <li><strong>Refuse dark patterns.</strong> The FTC, the US consumer protection agency, defines dark patterns as design practices that trick or manipulate users into making choices they would not otherwise make and that may cause them harm. Hidden cancellation, pre-checked consent, fake urgency. Saying no to that is part of the job.</li>
  <li><strong>Collective organization.</strong> One person alone can rarely push back on a product decision. An organized team can. The decisions above only hold up when they do not depend on one developer's individual courage. I wrote about collective ways of organizing tech work in <a href="https://www.mmlo.com.br/blog/plataformas-cooperativas-e-fediverso">Platform Cooperatives and the Fediverse</a>.</li>
</ul>

<p>This is where the two tests from the beginning come back. The impact you build at your job survives your departure if it becomes the team's standard. And whoever decides, controls: if developers take part in the decision, it stops belonging to the company alone.</p>

<h2>Conclusion: Start Small, Start Together</h2>

<p>Technology with social purpose is not a product category or a page in an annual report. It is a question about power: who defines the problem, who controls the solution, and who keeps looking after it when the enthusiasm fades. My capstone taught me this in the simplest way, with a pilot that worked and a project that stopped because no one was there to maintain it.</p>

<p>If you want to start, do not start by building yet another app on your own. Pick one front:</p>

<ul>
  <li>Offer help to a nonprofit in your city, and first ask what it already uses and who will maintain things.</li>
  <li>Fix the most common accessibility errors in the product you work on.</li>
  <li>Teach a basic technology class at a community center or to someone in your family.</li>
  <li>Open the Querido Diário or Brasil.IO repository and look for an issue to get started.</li>
</ul>

<p>Start small, and start together with a real community. A project other people can carry on makes more of a difference than the best-built system nobody can maintain.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://static.scielo.org/scielobooks/7hbdt/pdf/dagnino-9788578793272.pdf">Dagnino, Renato. Tecnologia Social: contribuições conceituais e metodológicas. EDUEPB/SciELO Books (2014).</a></li>
  <li><a href="https://www.benzevgreen.com/wp-content/uploads/2019/11/19-ai4sg.pdf">Green, Ben. "'Good' isn't good enough". NeurIPS AI for Social Good Workshop (2019).</a></li>
  <li><a href="https://www.hbgacademic.com/titles/evgeny-morozov/to-save-everything-click-here/9781610393706/">Morozov, Evgeny. To Save Everything, Click Here: The Folly of Technological Solutionism. PublicAffairs.</a></li>
  <li><a href="https://revistas.pucsp.br/index.php/ls/article/view/18912">Montaño, Carlos. "O projeto neoliberal de resposta à 'questão social' e a funcionalidade do 'terceiro setor'". Lutas Sociais (PUC-SP).</a></li>
  <li><a href="https://cetic.br/pt/tics/osfil/2025/geral/C1/">Cetic.br. TIC Organizações Sem Fins Lucrativos 2025, table C1.</a></li>
  <li><a href="https://cetic.br/pt/tics/osfil/2025/geral/C14/">Cetic.br. TIC Organizações Sem Fins Lucrativos 2025, table C14.</a></li>
  <li><a href="https://cetic.br/pt/tics/domicilios/2025/individuos/C2/">Cetic.br. TIC Domicílios 2025, table C2.</a></li>
  <li><a href="https://cetic.br/media/analises/tic_domicilios_2025_principais_resultados.pdf">Cetic.br. TIC Domicílios 2025: principais resultados (key findings).</a></li>
  <li><a href="https://webaim.org/projects/million/">WebAIM. "The WebAIM Million: The 2026 report".</a></li>
  <li><a href="https://agenciadenoticias.ibge.gov.br/agencia-noticias/2012-agencia-de-noticias/noticias/43463-censo-2022-brasil-tem-14-4-milhoes-de-pessoas-com-deficiencia">IBGE. "Censo 2022: Brasil tem 14,4 milhões de pessoas com deficiência" (2025).</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2015/lei/l13146.htm">Brazil. Law 13.146/2015 (Lei Brasileira de Inclusão, Brazilian Inclusion Law).</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2018/lei/l13709.htm">Brazil. Law 13.709/2018 (LGPD, Brazilian General Data Protection Law).</a></li>
  <li><a href="https://www.ftc.gov/reports/bringing-dark-patterns-light">Federal Trade Commission. "Bringing Dark Patterns to Light" (2022).</a></li>
  <li><a href="https://brasil.io/manifesto/">Brasil.IO. Manifesto.</a></li>
  <li><a href="https://docs.queridodiario.ok.org.br/pt-br/latest/">Open Knowledge Brasil. Querido Diário (documentation).</a></li>
  <li><a href="https://www.gov.br/governodigital/pt-br/plataformas-e-servicos-digitais/software-publico">Governo Digital. Software Público Brasileiro.</a></li>
  <li><a href="https://softwarepublico.gov.br/gitlab/leofl/articles/raw/master/OSS-2017/01-introduction.tex">UnB/USP team. Technical history of the rebuild of the Software Público portal (2017).</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 18 Oct 2026 20:42:04 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Tecnologia e Sociedade</category>
      <category>Impacto Social da Tecnologia</category>
      <category>Ética na Tecnologia</category>
      <category>Acessibilidade</category>
      <category>Engajamento Cívico</category>
      <category>Privacidade de Dados</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/15e2361f-c6cc-4ac7-ae2f-925a017998d9" medium="image"/>
    </item>
    <item>
      <title>The Developer&apos;s Manifesto</title>
      <link>https://www.mmlo.app/en/blog/o-manifesto-do-desenvolvedor</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/o-manifesto-do-desenvolvedor</guid>
      <description>AI, Free Software and Labor</description>
      <content:encoded><![CDATA[<p class="intro">The AI that autocompletes your code learned from the code of millions of people. Part of it was under the GPL, a license written precisely so that nothing derived from that work could become closed property. Today that learning comes back to us as a monthly subscription. I wrote the first version of this manifesto in 2025, and it treated free software almost as an idea. This version is about labor: who writes the code, who keeps the value it creates, and what a developer can stand for when AI and the cloud fence off what used to be held in common. At the end, I lay out the commitments I make and where I still fall short of them.</p>

<h2>Why a Developer's Manifesto in 2026</h2>

<p>Software is collective work. Almost no production system is written from scratch: it rests on libraries, compilers, kernels and tools written by thousands of people, many of whom were never paid. Together, they form the <em>digital commons</em>: a shared heritage nobody bought and everybody uses.</p>

<p>I have written about the history of this commons before. In <a href="https://www.mmlo.com.br/blog/free-software-vs-open-source">Free Software vs Open Source</a> I show how the movement split between freedom and efficiency, and in <a href="https://www.mmlo.com.br/blog/open-source">Open Source</a> I show how big companies learned to use that work while giving almost nothing back. I won't repeat those posts here. What has changed since then is the reason for this manifesto, and it comes down to two things happening at once.</p>

<p>The first is that <strong>AI has become the newest fence</strong> around the commons. Models trained on public code sell that work back as a paid product, and appropriation has become faster and less visible. A company used to have to take a project and close off a fork. Now it only has to train a model.</p>

<p>The second is that <strong>developers stopped seeing themselves as workers</strong>. Many of us think of ourselves as "entrepreneurs of the self": contractors, freelancers, people negotiating alone with the market. Ricardo Antunes and Ruy Braga were already studying this process in <em>Infoproletários</em> back in 2009: the precarization of work in IT, including those hired as contractors (PJ, a Brazilian legal entity) instead of employees. If you see yourself as a one-person company, you miss the fact that the code you write, at your job and in open source, is labor appropriated by others.</p>

<p>Each feeds the other. AI speeds up appropriation, and developers lack the class consciousness to push back. Marx called the theft of common land and its conversion into modern private property primitive accumulation. What is happening to code is not the same thing, but it rhymes.</p>

<h2>AI Trained on Open Source Code: the Newest Fence</h2>

<p>GitHub Copilot was trained on millions of public GitHub projects. In 2022, a group of programmers sued GitHub, Microsoft and OpenAI in <em>Doe v. GitHub</em>. In September 2026, the U.S. Court of Appeals for the Ninth Circuit upheld the dismissal of the main claim, based on the Digital Millennium Copyright Act (DMCA), on the grounds that the model creates new works and does not "remove" authorship attribution.</p>

<p>This ruling deserves a careful reading. It does <strong>not</strong> say that training AI on open source code is legal. It only says that this particular DMCA provision does not apply. The breach of contract claims, that is, the license violation claims, are still pending in the trial court. The question that matters to us remains open.</p>

<p>Microsoft argues that training on licensed code is <em>fair use</em>, a use permitted without authorization. The Software Freedom Conservancy, an organization that defends copyleft, replied to the U.S. Copyright Office in 2023:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"To concede Microsoft's 'fair use' claims would be the first step in eviscerating the copyleft licenses that protect the primary commons of software source code."</strong></p>
</div>

<p>In other words: accepting Microsoft's argument would hollow out the very licenses that protect the largest commons of source code we have. The SFC also points out that Microsoft acknowledges having used large amounts of copyleft code in training.</p>

<h3>"Open Source AI" without the data</h3>

<p>In October 2024, the Open Source Initiative published its definition of open source AI. It does not require training data to be open: it only asks for sufficient information about it. The OSI FAQ argues that "training data is not equivalent to source code". The criticism came from within. Bruce Perens, author of the original Open Source Definition, replied that <strong>training data is the source code</strong> of a model.</p>

<p>It is the same Free versus Open split, now applied to AI. On one side, a definition that accommodates the interests of those who train models on data they cannot or will not release. On the other, the Free Software Foundation, which argues in a draft that a machine learning system is only free if its training data and scripts are free as well.</p>

<h3>The technology is ours</h3>

<p>I don't think the answer is to reject AI. I use AI every day, and it was built on top of our work. That is exactly why it is ours too. The fight is not about using it or not, it is about who controls what was learned from the commons.</p>

<p>Given what has been happening, I argue that <strong>AI needs copyleft</strong>. If a model was trained on free code, it should be required to be open: weights, data and process. Today this is a proposal, not a license in force. The SFC set up a committee on the subject in 2022 because of Copilot, and in 2024 it concluded that no available system can be reproduced by the public. This is a political demand, and political demands are built by people standing up for them.</p>

<h2>Relicensing and the Cloud: When the Commons Becomes a Product</h2>

<p>AI is not the only fence. Over the last three years, widely used open source projects changed their licenses to stop others from profiting from them.</p>

<ul>
  <li><strong>Terraform (HashiCorp):</strong> in August 2023 it swapped the MPL 2.0 for the Business Source License, which is not open source. In September, the Linux Foundation launched OpenTofu, a fork backed by more than 140 organizations.</li>
  <li><strong>Redis:</strong> in March 2024 it dropped the BSD license for two source-available licenses, admitting it was no longer open source. A week later Valkey was born, a fork maintained under the Linux Foundation.</li>
  <li><strong>Elastic:</strong> closed the Elasticsearch license in 2021 and, in August 2024, brought back the AGPL as an option.</li>
</ul>

<p>I see two readings that are true at the same time.</p>

<p>The first is that this was <strong>appropriation from the community</strong>. These projects grew on contributions from many people outside the company. Closing the license afterwards means taking what was built in common and turning it into a private asset.</p>

<p>The second is that it was <strong>a symptom of cloud concentration</strong>. Redis itself justified the change by saying most of its sales went through the big cloud providers, which turned the project's investment into a commodity. In 2025, when adding the AGPLv3, CEO Rowan Trollope said cloud providers "reap the profits and control the infrastructure without proportional contributions back". That is capital itself acknowledging the extraction.</p>

<p>And there is a contradiction here worth facing: Valkey, the fork that "resisted" Redis closing its license, is backed by AWS, Google Cloud and Oracle. The commons resists, but the cloud is still the one paying for the resistance. Elastic went back to the AGPL after being named AWS "partner of the year". Forks prove that free code survives a closure, but they don't change who concentrates the infrastructure.</p>

<p>That is why the AGPL matters so much. It is the GPL adapted for software that runs as a service: whoever offers the program over a network must make the modified code available. It is copyleft's answer to the cloud, and it is no accident that Redis and Elastic went back to precisely this license. I explained the difference between copyleft and permissive licenses in <a href="https://www.mmlo.com.br/blog/gpl-vs-mit">GPL vs MIT</a>.</p>

<h2>The Invisible Work of Open Source Maintainers</h2>

<p>On March 29, 2024, engineer Andres Freund announced that versions 5.6.0 and 5.6.1 of <em>xz utils</em>, a compression library shipped in nearly every Linux distribution, contained a backdoor. The flaw received the maximum severity score (10.0) and came close to reaching SSH on servers around the world. The malicious code was inserted by "Jia Tan", a contributor who spent years earning the project's trust.</p>

<p>The most revealing part happened two years earlier. In June 2022, under pressure from accounts demanding a new maintainer (later suspected of being part of the attack), Lasse Collin, the original maintainer, wrote on the project's mailing list that his ability to care for xz was limited by long-term mental health issues, that he had been working with Jia Tan, and closed with this:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"It's also good to keep in mind that this is an unpaid hobby project."</strong></p>
</div>

<p>A project running on millions of servers was maintained as an unpaid hobby by one exhausted person. This is not an exception. In December 2021, <em>Log4Shell</em> exposed another 10.0 flaw in Log4j, a Java library used by companies all over the world. The official U.S. government report on the incident opens by noting that "a group of volunteers maintain Log4j" and discusses the risks of a "thinly-resourced, volunteer-based" community.</p>

<p>The numbers confirm the pattern. In the 2024 report by Tidelift, a company whose business is precisely paying maintainers, <strong>60% of maintainers</strong> describe themselves as unpaid hobbyists, only 12% earn most of their living from this work, and 60% have quit or considered quitting a project. One anonymous maintainer summed it up in the report: open source has generated trillions in value, harvested by large companies that give very little back, "not even recognition, respect, or gratitude".</p>

<p>In Marxist terms, this is <strong>the appropriation of other people's labor</strong>: value produced by those who work and captured by those who control the infrastructure. In a job, that appropriation happens through the wage. In open source, it happens with no wage at all.</p>

<h3>Who should pay</h3>

<p>I don't believe in a single solution. I believe in three fronts working together:</p>

<ul>
  <li><strong>Companies that profit from software pay</strong>, as an obligation and not as charity or a marketing campaign.</li>
  <li><strong>Public money funds digital infrastructure</strong>, the same way it funds roads. Germany already does this: the Sovereign Tech Agency has invested €41.1 million in 118 technologies since 2022, including OpenSSH, PHP and systemd.</li>
  <li><strong>Maintainers organize</strong> into foundations, cooperatives and unions to bargain collectively, instead of asking for donations one by one.</li>
</ul>

<h2>What I Commit to Defending</h2>

<p>A manifesto that only diagnoses is just an analysis. This one has commitments, and they are concrete enough for me to be held accountable for them.</p>

<p>The first is to use <strong>copyleft by default</strong>. When I publish a project, the license is a political choice, not a repository detail. <a href="https://www.mmlo.com.br/blog/rubrica">Rubrica</a>, Gatekey and <a href="https://www.mmlo.com.br/blog/code-forge">Code Forge</a> are under GPLv3. Voxr, which runs as a service, is under AGPLv3, because that is exactly the case the AGPL exists to protect. Permissive licenses have their place, but by default, whoever uses my work to build something should give back what they built.</p>

<p>The second is to <strong>give back to what I use</strong>. Every dependency in my <code>package.json</code> has someone behind it. Giving back is not always code: it can be a well-written bug report, a documentation fix, an answer on an issue, or money, when the company I work for can pay. xz showed what happens when nobody gives back.</p>

<p>The third is to prefer <strong>open standards, tools and models</strong>. The open web over closed platforms, open formats over formats that lock you in, open-weight AI models when they solve the problem. Each of these choices reduces dependence on whoever controls the infrastructure.</p>

<p>The fourth, and the hardest, is <strong>collective organization</strong>. Recognizing myself as a worker, not as a one-person company. This takes concrete forms: unions in our field, such as SINDPD, the IT workers' union in São Paulo; free technology cooperatives, such as EITA, which builds technology alongside social movements and has already become the subject of academic research; and examples abroad, such as the Alphabet Workers Union, founded in 2021 by just over 200 Google workers and which today claims more than a thousand members. I wrote about cooperative alternatives in <a href="https://www.mmlo.com.br/blog/plataformas-cooperativas-e-fediverso">Platform Cooperatives and the Fediverse</a>.</p>

<h2>The Limits of a Manifesto</h2>

<p>I need to be upfront about a contradiction: <strong>I use closed tools</strong>. I write much of my code with Claude Code, a proprietary product. I work on platforms I don't control. An attentive reader will notice, and it is fair that they do.</p>

<p>Nobody lives outside the system they criticize. A manifesto is not a purity test, it is a direction. It concerns what I control: the license of what I publish, what I give back, the choices I can make and the organizing I take part in. Waiting for a life without contradictions before acting is the most efficient way to never act at all.</p>

<p>I also need to be upfront about free software: <strong>on its own, it emancipates no one</strong>. Free code without organization becomes free raw material. That is what happened with Linux in the cloud, with Log4j in corporations and with GitHub code in Copilot's training. The license protects the code; what protects the people who write the code is their organization. That is why the fourth commitment weighs more than the other three.</p>

<h2>Conclusion: Pick One Commitment</h2>

<p>AI learned from our work, companies close their licenses when the cloud squeezes them, and the world's infrastructure depends on people maintaining software as a hobby. None of this is an accident: these are the incentives of a system that turns the commons into property whenever it can. As in <a href="https://www.mmlo.com.br/blog/engenheiros-da-liberdade-ou-arquitetos-da-prisao">Engineers of Freedom or Architects of the Prison?</a>, the question is which side our work is on.</p>

<p>You don't have to adopt all four commitments at once. Pick one and apply it to your next project:</p>

<ul>
  <li><strong>License:</strong> publish your next repository under the GPL or AGPL and explain why in the README.</li>
  <li><strong>Give back:</strong> pick a dependency you use every day and contribute to it this month, with code, documentation or money.</li>
  <li><strong>Openness:</strong> replace one closed tool or model with an open alternative where it does the job.</li>
  <li><strong>Organization:</strong> find out which union represents your field and talk to your colleagues about it.</li>
</ul>

<p>The code we write is already collective. What is missing is for the decisions about it to be collective too.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://storage.courtlistener.com/recap/gov.uscourts.ca9.08476fbf-f296-42a9-ba10-af35e4ec26ac/gov.uscourts.ca9.08476fbf-f296-42a9-ba10-af35e4ec26ac.126.1.pdf">U.S. Court of Appeals for the Ninth Circuit. "Doe v. GitHub, Inc., No. 24-7700" (09/16/2026).</a></li>
  <li><a href="https://www.authorsalliance.org/2026/09/23/resolving-an-interlocutory-appeal-ninth-circuit-affirms-dismissal-of-section-1202-dmca-claims-in-ongoing-doe-v-github-litigation/">Authors Alliance. "Ninth Circuit Affirms Dismissal of Section 1202 DMCA Claims in Doe v. GitHub" (09/23/2026).</a></li>
  <li><a href="https://sfconservancy.org/docs/2023-12-06_Software-Freedom-Conservancy-Copyright-Office-Generative-AI-Comments-Docket-2023-6_reply.pdf">Software Freedom Conservancy. "Reply Comment, Copyright Office Docket 2023-6" (2023).</a></li>
  <li><a href="https://sfconservancy.org/news/2024/oct/25/aspirational-on-llm-generative-ai-programming/">Software Freedom Conservancy. "Aspirational Statement on LLM-backed generative AI for Programming" (2024).</a></li>
  <li><a href="https://opensource.org/ai/open-source-ai-definition">Open Source Initiative. "The Open Source AI Definition 1.0".</a></li>
  <li><a href="https://opensource.org/ai/faq">Open Source Initiative. "The Open Source AI Definition: FAQ".</a></li>
  <li><a href="https://thenewstack.io/the-case-against-osis-open-source-ai-definition/">Cassel, David. "The Case Against OSI's Open Source AI Definition". The New Stack (2024).</a></li>
  <li><a href="https://www.hashicorp.com/blog/hashicorp-adopts-business-source-license">HashiCorp. "HashiCorp adopts Business Source License" (2023).</a></li>
  <li><a href="https://www.linuxfoundation.org/press/announcing-opentofu">Linux Foundation. "Linux Foundation Launches OpenTofu" (2023).</a></li>
  <li><a href="https://redis.io/blog/redis-adopts-dual-source-available-licensing/">Redis. "Redis Adopts Dual Source-Available Licensing" (2024).</a></li>
  <li><a href="https://www.linuxfoundation.org/press/linux-foundation-launches-open-source-valkey-community">Linux Foundation. "Linux Foundation Launches Open Source Valkey Community" (2024).</a></li>
  <li><a href="https://redis.io/blog/agplv3/">Trollope, Rowan. "Redis is now available under the AGPLv3 open source license" (2025).</a></li>
  <li><a href="https://www.elastic.co/blog/elasticsearch-is-open-source-again">Banon, Shay. "Elasticsearch Is Open Source. Again!" (2024).</a></li>
  <li><a href="https://www.openwall.com/lists/oss-security/2024/03/29/4">Freund, Andres. "backdoor in upstream xz/liblzma leading to ssh server compromise". oss-security (2024).</a></li>
  <li><a href="https://nvd.nist.gov/vuln/detail/CVE-2024-3094">NIST NVD. "CVE-2024-3094".</a></li>
  <li><a href="https://tukaani.org/xz-backdoor/">Collin, Lasse. "XZ Utils backdoor".</a></li>
  <li><a href="https://www.mail-archive.com/xz-devel@tukaani.org/msg00567.html">Collin, Lasse. "Re: [xz-devel] XZ for Java" (06/08/2022).</a></li>
  <li><a href="https://nvd.nist.gov/vuln/detail/CVE-2021-44228">NIST NVD. "CVE-2021-44228".</a></li>
  <li><a href="https://www.cisa.gov/sites/default/files/publications/CSRB-Report-on-Log4-July-11-2022_508.pdf">Cyber Safety Review Board. "Review of the December 2021 Log4j Event" (2022).</a></li>
  <li><a href="https://www.sonarsource.com/the-2024-tidelift-maintainer-impact-report.pdf">Tidelift. "The 2024 Tidelift Maintainer Impact Report".</a></li>
  <li><a href="https://www.sovereign.tech/tech">Sovereign Tech Agency. "Technologies".</a></li>
  <li><a href="https://code-cwa.org/stories/alphabet-workers-union-launches">CODE-CWA. "The Alphabet Workers Union Launches!" (2021).</a></li>
  <li><a href="https://www.alphabetworkersunion.org/faqs">Alphabet Workers Union. "FAQs".</a></li>
  <li><a href="https://policyreview.info/articles/analysis/beyond-platform-capitalism-digital-solidarity">Foletto, Leonardo; Santini, Daniel. "Beyond platform capitalism". Internet Policy Review (2026).</a></li>
  <li><a href="https://eita.coop.br/">EITA Cooperative.</a></li>
  <li><a href="https://sindpd.org.br/">SINDPD (IT workers' union, São Paulo).</a></li>
  <li><a href="https://www.marxists.org/portugues/marx/1867/capital/cap24/index.htm">Marx, Karl. O Capital (Capital), Volume I, ch. XXIV: "A Chamada Acumulação Original" (The So-Called Primitive Accumulation).</a></li>
  <li><a href="https://blogdaboitempo.com.br/wp-content/uploads/2022/05/infoproletarios_antunes.pdf">Antunes, Ricardo; Braga, Ruy (eds.). Infoproletários. Boitempo (2009).</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 18 Oct 2026 20:42:04 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Free Software</category>
      <category>Open Source</category>
      <category>Inteligência Artificial</category>
      <category>Forças Produtivas</category>
      <category>Condições de Trabalho</category>
      <category>Tecnologia e Sociedade</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/758ed142-6fb4-47bc-9373-808aed99cd98" medium="image"/>
    </item>
    <item>
      <title>AI Skills by Team</title>
      <link>https://www.mmlo.app/en/blog/skills-de-ia-por-area</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/skills-de-ia-por-area</guid>
      <description>Productivity Beyond the Dev Team</description>
      <content:encoded><![CDATA[<p class="intro">When people talk about AI at work, almost every example comes from the development team: generated code, written tests, reviewed pull requests. That makes sense, because developers already live in the terminal. But most of a company's repetitive work rarely sits in the code. It sits in the card copied by hand, the Friday report, the test spreadsheet built from scratch and the email rewritten three times for three audiences. In <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI Governance in Practice</a> and <a href="https://www.mmlo.com.br/blog/hub-de-skills">Skills Hub</a> I explained why InBot chose to upskill its teams instead of shrinking them, and how skills reach each person's machine safely. Now I show what these AI skills deliver, team by team: the problem before, what the skill does and what changed. Above all, I show why the big gain does not come from a single skill, but from one skill feeding the next.</p>

<h2>What a Skill Is, and Why the People Who Know the Process Should Write It</h2>

<p>A <em>skill</em> is a folder with instructions, examples and sometimes scripts that Claude Code or Codex loads when a task calls for it. Anthropic uses a comparison I really like: building a skill is like putting together <strong>an onboarding guide for a new hire</strong>.</p>

<p>That comparison says a lot about who should write the guide. A language model knows a lot in general and nothing about your team's card template, the rule QA applies before opening a bug or the way sales records an opportunity. The team knows that. A skill is the team's process written in a way the AI can follow.</p>

<p>A technical detail makes it possible to have dozens of them installed. According to the official documentation, only each skill's name and description stay loaded all the time, about <strong>100 tokens per skill</strong>. The body of the skill only comes in when it is triggered, and supporting files only when they are read. In practice, a catalog of ~50 skills does not weigh the assistant down. And the format is not tied to a single vendor: Codex uses the same open standard, so the same skills run in both tools.</p>

<h3>Who builds them: it depends on the team</h3>

<p>I would like to say that every team at InBot writes its own skills on its own. That is not true, and pretending it is would hide the most interesting part of the work.</p>

<p>Today, <strong>UX and Content</strong>, <strong>QA</strong> and <strong>Development</strong> build with a lot of autonomy. They bring the process, write the skill, test it and publish it through the Skills Hub, and the AI Enablement team reviews it before it goes to production. In the other teams, the AI team takes a bigger part in building. Even so, the shift has started: Sales published a skill of its own through the Hub, without anyone from the AI team writing it for them.</p>

<h3>The barrier is the tool, not the process</h3>

<p>What holds back the less technical teams is not a lack of knowledge about their own work. They know exactly what they do every week. What holds them back is <strong>familiarity with the tool</strong>: the terminal, Claude Code, a command to paste, a configuration file.</p>

<p>This is not unique to us. A BCG survey of more than 10,000 workers found that only <strong>36%</strong> feel properly trained to use AI, and an independent Jobs for the Future survey reached the same number in 2026. The same BCG study shows that people who get five or more hours of training, ideally in person and with coaching, are much more likely to become regular users. And Gallup found that employees whose manager actively supports AI use it frequently <strong>2.1 times more often</strong>.</p>

<p>Our answer had two fronts:</p>

<ul>
  <li><strong>Close support.</strong> The AI team works closely with people who are getting started, maps the process together with them and stays available when something does not work.</li>
  <li><strong>An installation page designed for non-developers.</strong> In the Skills Hub, people choose their operating system, download the installer and paste a single command, with instructions on how to open the terminal on Mac, Linux or Windows. Anyone who prefers can ask Claude Code or Codex to run the installation. The page has an FAQ section for the most common errors, and the installer's signature is checked before it runs, so nobody needs to understand security to be protected.</li>
</ul>

<h2>AI Skills by Team: What Changed in Each One</h2>

<p>From here on, I go through each team in the same format: the problem before, what the skills do and what changed. I will not use the skills' internal names or talk about clients. What matters is the kind of repetitive work that no longer takes up people's week.</p>

<h3>UX and Content: FAQs, journeys and personas</h3>

<p>I start with the heart of the product. InBot builds virtual assistants, and each one needs a <strong>persona</strong> (who the assistant is and how it talks), a <strong>question-and-answer knowledge base</strong> the AI consults and <strong>journeys</strong>, the guided service flows. All of this used to be built by hand, in spreadsheets in the platform's format, with a lot of variation from one project to the next.</p>

<p>Today there are skills for each piece: one creates the persona, another builds the knowledge base, another builds the journeys, another explains and debugs the markup language used in the bot's content cards, another draws the flowchart, and a last one reviews the configuration before the bot goes to production.</p>

<p>Two results show the care behind this:</p>

<ul>
  <li><strong>800 automated tests</strong> to decide how a micro-agent should validate user input: with code or with the model alone. The "model only" version was the only source of errors across the runs, and the skill now generates the validation in code.</li>
  <li>The skill that explains the content card language was born for the UX team and is now distributed to <strong>7 teams</strong>. Knowledge that used to depend on asking the right person became a lookup anyone can do.</li>
</ul>

<h3>Planning: from the meeting to the document</h3>

<p>Planning turns meetings, briefs and transcripts into a project planning document. It used to take hours of reading and reorganizing for every new project.</p>

<p>The team's skills write the executive summary of the meeting, build the planning document in the official template, run a privacy check (personal data, purpose, likely legal basis and gaps) before the project starts and create standardized cards in ClickUp. Card creation passed <strong>5 out of 5 test scenarios</strong>, and a project card is born with the team's full pre-kickoff checklist.</p>

<p>One lesson came from here: when we looked at what the teams were asking for, we found <strong>three requirements-gathering skills</strong> being reinvented in parallel, one in Planning, one in UX and one in Projects. They became one. Governance is also this: spotting duplication before it turns into three versions of the truth.</p>

<h3>Projects: cards and schedules without manual copying</h3>

<p>Projects is where the "before" is most striking:</p>

<ul>
  <li>for every new project, someone copied a <strong>79-card template list</strong> by hand and edited the cards one by one;</li>
  <li>every week, the manager read <strong>40 to 80 cards per project</strong> to know what needed to happen;</li>
  <li>the Friday status report was personalized by hand, in a portfolio that reached <strong>more than 50 projects per analyst</strong>.</li>
</ul>

<p>Now the schedule comes from the planning document, the cards are created from the template (always with a preview before saving), and the weekly summary and status report read what was created. The status report comes out in three versions: internal, client-facing and executive. An audit compares periodic snapshots of the lists and flags cards with no owner, deadlines pushed without justification and cards closed without evidence.</p>

<p>The team also got a communication validator, which rewrites the same email for different audiences and checks that it has an owner, a deadline and next steps. In its first version, it got <strong>56 out of 60 checks right (93%)</strong>. After the analyst's feedback, it went through <strong>77 more scenarios</strong> before going to production.</p>

<h3>QA: tests generated from business rules</h3>

<p>In QA, the classic problem was the inherited test spreadsheet: you copied the one from a similar project and adapted it, carrying over scenarios and terms that had nothing to do with the new product.</p>

<p>The test spreadsheet skill starts from scratch, from the product's business rules, and also generates a dashboard where QA marks the execution. In a recent review, a spreadsheet of <strong>212 cases shrank to 137 rows covering the same cases</strong>. Another skill tests the bot end to end in read-only mode: the happy path, field validation, attempts to manipulate the assistant and sensitive content.</p>

<h3>Development: one standard where there used to be five</h3>

<p>For the dev team, the gain was less about speed and more about standards. There were <strong>five different definitions of code review</strong> going around, and each person reviewed in their own way. They became one.</p>

<p>Other skills standardize branch and PR names and keep the ClickUp card in sync with the PR's status, drive deployment across environments and generate the feature's documentation, checked against the actual code. One of them drives the single sign-on setup for new clients, a process that used to require reading about 1,900 lines spread across ten documents. If you want to see what this looks like in a developer's individual workflow, I described mine in <a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">My AI Development Workflow</a>.</p>

<h3>Sales: CRM and routine without rework</h3>

<p>In Sales, the skills operate the CRM with fixed rules: they search before creating, avoid duplicates, ask only for what is missing and confirm before saving. Before approval, the operation went through <strong>250 runs in the CRM, with 250 correct results</strong>.</p>

<p>The team's daily dashboard and the weekly report on new companies also come from there. They only read data and leave the email as a draft for a person to review and send. And Sales itself published a skill through the Hub to organize the weekly follow-up of its strategic missions.</p>

<h3>Marketing: one theme, every piece</h3>

<p>Marketing had a very specific kind of rework: asking the AI for each piece separately and then copying everything into the editorial calendar. Now a theme from the calendar becomes a blog post, an Instagram post, a LinkedIn post and an email at once, and the card is created in the calendar with the publication date. Before going to production, the skill went through <strong>80 scenarios across four rounds of testing</strong>.</p>

<h3>The pieces every team reuses</h3>

<p>Some skills do not belong to any team, and they may be the most important ones. One operates ClickUp from the terminal, another operates the platform's data tables, another sends email through the corporate account, another writes meeting minutes and another interviews the person, one question at a time, until the request is clear. The teams' skills do not reimplement any of this: they call these pieces.</p>

<p>In the same category are the skills that operate the company's systems from a request in plain Portuguese. The one that operates the platform's admin panel turned a <strong>3-hour task into 13 minutes</strong>. With it, we also analyzed a 30-day A/B test between two language models in our chatbot, comparing tokens and response time. The platform shows one day at a time and does not export data, so this used to require daily screenshots and manual math. It took <strong>about 20 minutes</strong>.</p>

<h2>Chaining: a Bot Project from the Meeting to Production</h2>

<p>So far, each skill looks like a saving of a few minutes, sometimes a few hours. The leap shows up when you look at an entire delivery. Anthropic's documentation lists among the benefits of skills precisely <strong>combining several of them for multi-step tasks</strong>. At InBot, this became an architecture rule: one skill <em>plans</em> and another <em>executes</em>, and the output of one is the input of the next.</p>

<p>Here is the path of a bot project:</p>

<ol>
  <li><strong>Meeting minutes:</strong> the transcript becomes decisions, open items and owners.</li>
  <li><strong>Planning document:</strong> the minutes and the brief become the project's official plan.</li>
  <li><strong>Privacy check:</strong> the plan is read looking for personal data and risks before anything is executed.</li>
  <li><strong>Cards and schedule:</strong> the plan becomes the schedule and the cards in ClickUp, with a preview.</li>
  <li><strong>Persona, knowledge base and journeys:</strong> the bot's content is generated in the platform's format.</li>
  <li><strong>Test spreadsheet and bot QA:</strong> the scenarios come from the business rules and the bot is tested end to end.</li>
  <li><strong>Pre-production qualification:</strong> the configuration is reviewed before the bot goes live.</li>
  <li><strong>Deployment plan:</strong> the plan reads the deployment card and points to QA's post-deployment checklist instead of reinventing the scenarios. Its rule sums up the whole philosophy: <strong>nothing without a source becomes a guess; it becomes an open item with an owner</strong>.</li>
</ol>

<p>None of these steps is revolutionary on its own. The difference is that the quality of one guarantees the next. A note on one of the cards in this chain puts it well: it is the rigor in creating the card that makes the report come out ready later.</p>

<h3>When one team fixes another team's skill</h3>

<p>Chaining also works across teams. In a QA round, the testing skill found that <strong>35% of a bot's content was out of the AI's reach</strong>: a configuration field cut the text without warning, and part of the answers was simply never found. The finding went back to UX, and the skill that builds the knowledge base now <strong>blocks generation</strong> when this problem appears. The defect was found once, and the skill keeps it from happening again in future projects.</p>

<p>A study with 776 Procter &amp; Gamble professionals found something similar at scale: with AI, proposals from R&amp;D and from Commercial people became more balanced, as if the AI brought each one the other team's perspective. Here, what brings one team's perspective to the other is the skill.</p>

<h3>Skills that orchestrate other skills</h3>

<p>The next step was natural. The skill that builds the deployment package for a simple bot was, at first, self-contained: it handled persona, knowledge base and journey on its own. It was rebuilt to only orchestrate, delegating each piece to the specialist skills. It became <strong>67% smaller</strong> and produced the <strong>7 deliverables in 11 out of 11</strong> test scenarios. In another case, a demo bot was built with <strong>7 skills in sequence</strong>, from the journey map to the task card.</p>

<p>This has a practical consequence: when the persona skill improves, every package that depends on it improves too, without anyone editing the orchestrator.</p>

<h2>A Human in Control</h2>

<p>None of this works if the teams do not trust the skills. And trust came from a pattern that repeats in almost all of them:</p>

<ul>
  <li><strong>Preview and confirmation</strong> before saving anything in ClickUp, the CRM or the platform.</li>
  <li><strong>Read-only</strong> wherever possible: reports and bot tests do not change anything.</li>
  <li><strong>Email as a draft</strong>: the skill writes, a person sends.</li>
  <li><strong>A test suite before the release</strong>, with real cases and more than one model, as Anthropic itself recommends.</li>
  <li><strong>Whoever wrote the skill is not the one who approves it</strong>: every publication goes through review.</li>
</ul>

<p>The tests have already changed decisions. In the status report, with vague or contradictory documents, the cheapest model got it wrong and only the mid-tier one got it right. The rule became: <strong>the budget model does not run in production without human review</strong>. I explained the logic of this governance in <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI Governance in Practice</a>.</p>

<h2>Productivity Without Hype: What Can Be Measured</h2>

<p>"Extreme productivity" is the kind of phrase I distrust when I read it. So it is worth separating what we measured from what we did not.</p>

<p>External data shows that the gain is real, but uneven. In the BCG study with Harvard, consultants using AI completed <strong>12.2% more tasks</strong> and worked <strong>25.1% faster</strong> on tasks within what AI does well. On a task outside that frontier, they were <strong>19% less likely to get it right</strong>. At MIT, business professionals took <strong>40% less time</strong> on writing tasks, with 18% higher quality.</p>

<p>In aggregate, the number is much more modest. The St. Louis Fed estimated that people who use generative AI save on average <strong>5.4% of their working hours</strong>. A study using official data from Denmark found no effect on wages and hours two years after ChatGPT. The average hides the concentration: the big gain shows up in specific tasks, and it does not show up when the tool is simply rolled out without changing the process. BCG says it plainly: companies cannot simply roll out AI tools and expect transformation.</p>

<p>That is why I believe in skills by team, not in "granting access". Gallup found in 2026 that the share of people who perceive a gain rises from <strong>45%</strong>, among those who use AI for one or two tasks, to <strong>90%</strong>, among those who use it for seven or more. One skill saves minutes. A set of skills that covers a team's whole workflow changes what that team can get done in a week.</p>

<p>And here is the honest part: <strong>we have not measured hours saved for most skills</strong>. What we have are test suites, accuracy rates, size reductions and feedback from the teams. The time figures I cite, like the 3 hours that became 13 minutes, are measured cases, not averages. The big gain is in repetitive and lookup tasks. In judgment tasks, the skill organizes the work, and the decision stays with the person.</p>

<h2>How to Start with AI Skills in Your Team</h2>

<p>In four months, InBot's teams got <strong>77 initiatives</strong> off the ground and <strong>about 50 skills and automations into production</strong>. The domain merit belongs to each team that opened up its own process. The AI Enablement team comes in to bring this up to the company's standard, test it and make sure it arrives safely. If I had to sum up what we learned as a roadmap, it would be this:</p>

<ol>
  <li><strong>List your team's repetitive tasks.</strong> Not the hardest ones: the ones that repeat every week and follow the same format.</li>
  <li><strong>Start with a small, read-only skill.</strong> A report, a summary or a check. It is the safest way to build trust.</li>
  <li><strong>Write the process before the prompt.</strong> If nobody can explain the process to a new colleague, the AI will not be able to follow it either.</li>
  <li><strong>Test with real cases before distributing</strong>, and with more than one model.</li>
  <li><strong>Think about the next skill from the first one.</strong> Define the output of one so that the next one can read it. That is where the gain stops adding up and starts multiplying.</li>
  <li><strong>Stay close to the people who are getting started.</strong> The barrier is almost never the process; it is the tool.</li>
</ol>

<h2>Conclusion</h2>

<p>AI skills by team are not a list of shortcuts. They are each team's process written in a way the AI can follow, with a person checking what matters. One skill on its own saves minutes. Chained together, they carry a delivery across the company, from the meeting minutes to the deployment plan, with the quality of each step guaranteeing the next.</p>

<p>If you lead a team, my suggestion is to look at your next week and mark one task that repeats. That is your first skill. The second will show up when you notice who uses its output. And if you want to think about what this means for professions in the long run, I wrote about it in <a href="https://www.mmlo.com.br/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais">AI and the Future of Professions</a>.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://www.anthropic.com/engineering/equipping-agents-for-the-real-world-with-agent-skills">Zhang, B.; Lazuka, K.; Murag, M. "Equipping agents for the real world with Agent Skills". Anthropic Engineering, 2025.</a></li>
  <li><a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview">Anthropic. "Agent Skills" (overview).</a></li>
  <li><a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/best-practices">Anthropic. "Skill authoring best practices".</a></li>
  <li><a href="https://learn.chatgpt.com/docs/build-skills">OpenAI. "Build skills" (Codex and ChatGPT).</a></li>
  <li><a href="https://www.hbs.edu/faculty/Pages/item.aspx?num=64700">Dell'Acqua, F. et al. "Navigating the Jagged Technological Frontier". Harvard Business School Working Paper 24-013.</a></li>
  <li><a href="https://news.mit.edu/2023/study-finds-chatgpt-boosts-worker-productivity-writing-0714">MIT News. "Study finds ChatGPT boosts worker productivity for some writing tasks", 2023.</a></li>
  <li><a href="https://www.nber.org/papers/w33641">Dell'Acqua, F. et al. "The Cybernetic Teammate". NBER Working Paper 33641, 2025.</a></li>
  <li><a href="https://www.stlouisfed.org/on-the-economy/2025/feb/impact-generative-ai-work-productivity">Bick, A.; Blandin, A.; Deming, D. "The Impact of Generative AI on Work Productivity". Federal Reserve Bank of St. Louis, 2025.</a></li>
  <li><a href="https://www.nber.org/papers/w33777">Humlum, A.; Vestergaard, E. "Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI". NBER Working Paper 33777.</a></li>
  <li><a href="https://www.bcg.com/press/26june2025-beyond-ai-adoption-full-potential">BCG. "Companies Must Go Beyond AI Adoption to Realize Its Full Potential" (AI at Work 2025).</a></li>
  <li><a href="https://www.jff.org/newsroom/press-releases/worker-anxiety-over-ai-is-growing-and-employers-arent-preparing-employees-for-whats-next-new-survey-finds/">Jobs for the Future. "Worker Anxiety Over AI Is Growing, and Employers Aren't Preparing Employees for What's Next", 2026.</a></li>
  <li><a href="https://www.gallup.com/workplace/694682/manager-support-drives-employee-adoption.aspx">Kemp, A. "Manager Support Drives Employee AI Adoption". Gallup, 2025.</a></li>
  <li><a href="https://www.gallup.com/workplace/713063/ai-workplace-productivity.aspx">Gallup. "AI and Workplace Productivity: What the Data Show", 2026.</a></li>
  <li><a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">Mierzwa, Matheus. "AI Governance in Practice" (Portuguese).</a></li>
  <li><a href="https://www.mmlo.com.br/blog/hub-de-skills">Mierzwa, Matheus. "Skills Hub" (Portuguese).</a></li>
  <li><a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">Mierzwa, Matheus. "My AI Development Workflow" (Portuguese).</a></li>
  <li><a href="https://www.mmlo.com.br/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais">Mierzwa, Matheus. "AI and the Future of Professions" (Portuguese).</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 18 Oct 2026 18:44:42 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Governança de IA</category>
      <category>Claude Code</category>
      <category>Produtividade</category>
      <category>Automação</category>
      <category>Transformação Digital</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/a4197b60-eb16-4ff6-b6e1-035150cec7ef" medium="image"/>
    </item>
    <item>
      <title>AI and the Future of Work</title>
      <link>https://www.mmlo.app/en/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais</guid>
      <description>Tasks Change Before Job Titles</description>
      <content:encoded><![CDATA[<p class="intro">"Will AI take my job?" The short answer, in most cases, is no. But it will change what you do in it, and some roles will need fewer people. Both things are true at the same time, and most of the debate about AI and the future of work picks only one: either "no job is safe" or "AI only helps". In this post I separate tasks from occupations using data from the ILO, Stanford, and the World Economic Forum. Then I show who is most exposed, share what changed in my own work as a developer, and close with an exercise you can apply to your role.</p>

<h2>Tasks, Not Occupations: What AI Exposure Really Means</h2>

<p>Almost no occupation is a single task. A support agent answers questions, but also calms down angry customers, handles exceptions, and notices when a problem is bigger than it looks. A developer writes code, but also understands the problem, picks the architecture, and decides what not to build. AI takes some of those tasks, not the whole package.</p>

<p>That is why the right question is not "will my occupation disappear?" but "which tasks in my occupation are exposed?". Being <em>exposed</em> to AI means part of your work can be done or sped up by it. It does not mean being replaced.</p>

<p>Numbers from the <a href="https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure">International Labour Organization</a> show this difference clearly. In the 2025 update of its exposure index, <strong>1 in 4 workers</strong> worldwide is in an occupation with some exposure to generative AI. But only <strong>3.3%</strong> of global employment sits in the highest exposure level. The ILO's own conclusion:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"As most occupations consist of tasks that require human input, transformation of jobs is the most likely impact of GenAI."</strong></p>
</div>

<p>In other words: since most occupations include tasks that need a human, changing how work is done is the most likely outcome, not wiping jobs out.</p>

<h3>It changes, it creates, and it also shrinks</h3>

<p>"Transformation" does not mean "nothing happens". I see three movements happening together.</p>

<ul>
  <li><strong>Tasks change.</strong> The same role gets a different day-to-day. This is the most common case.</li>
  <li><strong>New roles appear.</strong> My current position is an example: since June 2026 I have led the <em>AI Enablement</em> area at InBot, a role that did not exist at the company until recently.</li>
  <li><strong>Some roles do shrink.</strong> They do not vanish, but they need fewer people.</li>
</ul>

<p>The World Economic Forum's <a href="https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf">Future of Jobs Report 2025</a> puts numbers on both sides: by 2030, <strong>170 million</strong> jobs are expected to be created and <strong>92 million</strong> displaced, a net gain of 78 million. The overall balance is positive, but it is no comfort to those among the 92 million. That is why it is worth looking at who is closest to that side of the equation.</p>

<h2>Who Is Most Exposed: Office Work, Customer Support, and Early Careers</h2>

<p>Back in 2023, the ILO already pointed to clerical work as the most exposed: nearly a quarter of the tasks in those occupations are highly exposed to generative AI. The WEF lands in a similar place when it lists the fastest-shrinking roles: cashiers, administrative assistants, data entry clerks, telemarketers, and graphic designers.</p>

<p>In my view, four kinds of work are shrinking:</p>

<ul>
  <li><strong>Repetitive office tasks:</strong> triage, form filling, standard reports.</li>
  <li><strong>Tier-1 support:</strong> frequently asked questions and simple tickets.</li>
  <li><strong>Routine code:</strong> CRUD, boilerplate, translating code between languages.</li>
  <li><strong>Generic content produced at volume:</strong> low-value text and images.</li>
</ul>

<p>The last one already has data. A study of freelancers found that, after ChatGPT launched, writers lost about <strong>2% of their jobs</strong> per month and <strong>5.2% of their income</strong>. It is a short-term effect, measured in 2023, but it points in the expected direction.</p>

<p>Shrinking does not mean cutting everything, either. In customer service, Gartner predicts that half of the companies that cut staff because of AI will rehire by 2027. The simple task leaves human hands. The hard case still needs someone.</p>

<h3>Juniors lose and gain at the same time</h3>

<p>The data point that strikes me most is about people starting out. The <a href="https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine">"Canaries in the Coal Mine?"</a> study from the Stanford Digital Economy Lab was updated in August 2026 with US payroll data. It shows that employment of <strong>22 to 25-year-olds</strong> in the occupations most exposed to AI is <strong>19% below</strong> where it would be had it kept pace with less-exposed occupations. Among experienced workers in those same roles, no comparable gap appears. Software development and customer service are among the examples.</p>

<p>Two details in the study matter a lot. First: the effect comes from <strong>less hiring</strong> of young people, not more layoffs. Second: it shows up where AI replaces tasks. Where it complements the work, employment stays stable or grows.</p>

<p>In Brazil, the pattern repeats. A study by Daniel Duque at <a href="https://blogdoibre.fgv.br/sites/blogdoibre.fgv.br/files/u139/inteligencia_artificial_generativa_e_mercado_de_trabalho_no_brasil.pdf">FGV Ibre</a> found negative effects on employment and income only among 18 to 29-year-olds more exposed to AI. For ages 30 and up, the effect is consistent with zero. The author's explanation sums it up: AI lowers the value of standardized tasks, exactly the ones that used to be the entry point to many careers.</p>

<p>These findings are still early and, on their own, do not prove causation. The Stanford authors themselves say AI is not the only factor: interest rates and earlier trends also play a role. Even so, the first data points in a clear direction.</p>

<p>Now, the other side. The <a href="https://www.nber.org/papers/w31161">NBER</a> study of more than 5,000 support agents found that AI raised productivity by <strong>14% on average</strong> and by <strong>34% among novice</strong> and less-skilled workers. For the most experienced, the effect was minimal. In other words, AI shortens the learning curve for people who are starting out.</p>

<p>That is why I think juniors lose and gain at the same time. Those who use AI to learn get a shortcut. But entry-level openings for simple tasks are shrinking, because those tasks were exactly the ones AI learned to do. In software, <a href="https://hiringlab.indeed.com/2026/07/23/the-labor-market-is-tilting-toward-seniority/">Indeed Hiring Lab</a> shows the scale of it: only <strong>4.5%</strong> of software development postings in the US are entry-level, and 71% of the recent recovery in postings came from senior roles.</p>

<p>Software development shows up in almost all of this data. It is also the profession I know from the inside.</p>

<h2>The Case of Software Development: From Writing Code to Orchestrating and Validating</h2>

<p>My work has changed a lot in recent years. I used to write code. Today I orchestrate: I think through the architecture, define the requirements, break the work into tasks, create agents with clear limits, and validate what the AI delivers. I described this process in detail in <a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">my AI development workflow</a>.</p>

<p>That does not mean I need to know less. I need to know more, because I have to understand and validate code I did not write. And I am not an exception. In the 2025 <a href="https://survey.stackoverflow.co/2025/ai">Stack Overflow</a> survey, <strong>84%</strong> of developers use or plan to use AI. At the same time, trust in the accuracy of its output dropped from 43% to <strong>33%</strong> in one year. The top frustration, cited by 66%, is code that is "almost right, but not quite".</p>

<p>Google's <a href="https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf">DORA 2025</a> report sums it up in one sentence: AI's primary role in software development is that of an amplifier. It magnifies what a team does well, and also what it does badly.</p>

<h3>What makes a good developer now</h3>

<p>If AI writes much of the code, a developer's value shifts to what comes before and after it. For me, it is four things together:</p>

<ul>
  <li><strong>Architecture and requirements:</strong> knowing what to build and why, before asking for a single line of code.</li>
  <li><strong>Validating what AI generates:</strong> tests, review, and security. This is where <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">TDD</a> comes into my work: the test defines the behavior, and the code is what adapts to it.</li>
  <li><strong>Orchestrating agents:</strong> breaking the work into small tasks and giving each agent clear limits on what it can and cannot do.</li>
  <li><strong>Domain knowledge:</strong> understanding the business and the customer better than any tool.</li>
</ul>

<h3>The risks: vibe coding today, lost foundations tomorrow</h3>

<p>In the short term, the biggest risk is <em>vibe coding</em> without a foundation: people generating entire systems without understanding what was generated. It works until the first security problem shows up. Out of curiosity, I have looked into applications built this way and got full access to more than one.</p>

<p>In the long term, the risk is losing the foundations. People who never wrote code have a harder time reading code, and people who cannot read it cannot validate it. Anthropic ran an experiment with engineers, mostly juniors, learning a new library. Those who learned with AI scored <strong>50%</strong> on a comprehension test, versus <strong>67%</strong> for those who coded by hand, and the biggest gap was in debugging. The sample is small, and the company sells the product being studied. Even so, the authors' sentence sums up the risk:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"Productivity benefits may come at the cost of skills necessary to validate AI-written code."</strong></p>
</div>

<p>In other words: the productivity gain may cost exactly the skills you need to check AI-written code. The same study offers a useful clue: those who used AI to ask questions about the concepts, and not just to generate code, learned more.</p>

<h3>AI is not a tool, it is a technology</h3>

<p>Many people talk about the risk of "depending on the tool". I see it differently. AI is not a tool, it is a technology. We all learned to do math by hand at school, but no workplace requires us to do that on the job. They all <strong>require</strong> us to use a calculator, because it automates the process.</p>

<p>Depending on the calculator was never the problem. The problem would be not knowing what multiplication is and, because of that, not noticing when the result makes no sense. It is the same with AI: using it will be mandatory. What sets people apart is knowing the fundamentals well enough to spot the mistake.</p>

<p>And this logic does not apply only to developers.</p>

<h2>Future Skills: What Becomes More Valuable (and What Becomes Less)</h2>

<p>The WEF estimates that <strong>39%</strong> of workers' core skills will change or become outdated by 2030. A word of caution on that number: it has been falling (it was 57% in 2020 and 44% in 2023), partly because more people have received training. It is not an acceleration. It is a large and ongoing shift. At the top of the report's list is analytical thinking, considered essential by 7 out of 10 companies.</p>

<p>In my view, four skills become more valuable in any field:</p>

<ul>
  <li><strong>Critical thinking:</strong> questioning the AI's answer and knowing when it is wrong.</li>
  <li><strong>Domain knowledge:</strong> understanding the business and the customer. AI knows a lot in general and nothing about the exception your team handles every week.</li>
  <li><strong>Process design:</strong> before automating, asking whether the problem is speed or the process itself. AI speeds up what exists. If the process is wrong, it speeds up the mistake.</li>
  <li><strong>Writing and clarity:</strong> those who write well, ask well. Explaining precisely what you want has become a work skill.</li>
</ul>

<p>What becomes less valuable is <strong>memorizing and executing</strong>: knowing by heart the commands, formulas, and steps that AI executes better and faster. Knowing how to execute is still useful, it just stopped being the differentiator.</p>

<p>The market already pays for this shift. According to <a href="https://kyodonewsprwire.jp/release/202606150856">PwC</a>, job postings that ask for AI skills pay <strong>62% more</strong> on average. And the entry-level postings most exposed to AI now ask far more often for senior-level skills, such as leadership and creativity. PwC sells AI consulting, so read it with care, but the data matches Stanford's: the entry door does not disappear, it gets more demanding.</p>

<p>If skills change this much, one question remains: who pays for it?</p>

<h2>Reskilling: Whose Responsibility Is It</h2>

<p>For me, the responsibility is shared. The individual seeks it out, the company trains, and the public sector guarantees the basics. When one of the three fails, the people who pay are those with the fewest resources to adapt on their own.</p>

<h3>The individual: seek it out</h3>

<p>Nobody will map the tasks of your role better than you. Waiting for your company or the government to act before you learn is risky, especially early in your career, where the data shows the biggest impact. Seeking it out does not have to mean an expensive course. It can mean using AI in your own work, every day, with a critical eye.</p>

<h3>The company: train</h3>

<p>On paper, companies already agree: <strong>85%</strong> of the employers surveyed by the WEF say they will prioritize reskilling. In practice, training is not keeping up. According to Microsoft and LinkedIn's 2024 Work Trend Index, only <strong>39%</strong> of AI users had received any training from their company. A 2025 survey by <a href="https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain">BCG</a> found only a third of employees saying they were well trained. And the WEF itself estimates that, out of every 100 workers, 11 will likely not receive the training they need by 2030.</p>

<p>I argue that training is smarter than cutting, and I wrote about how this works inside a company in <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI governance in practice</a>.</p>

<h3>The public sector: public policy and social protection</h3>

<p>This is the part that usually gets left out of the debate. Individuals and companies solve a lot, but not everything. Someone who loses an entry-level job at 23, in a role that shrank, needs accessible training and protection during the transition. The ILO has said this plainly since 2023: workers' voice, skills training, and adequate social protection will be key to managing this change.</p>

<p>In Brazil, there are two developments worth following:</p>

<ul>
  <li><strong>Bill PL 2338/2023</strong>, which regulates AI, was approved by the Senate in December 2024. The text has a dedicated section on protecting work: it calls for guidelines to reduce negative impacts on workers, "especially the risks of job displacement", and for public programs for reskilling and supporting workers affected by AI. But it is a bill, not a law. As of October 2026, it was still awaiting the rapporteur's report in the <a href="https://www.camara.leg.br/proposicoesWeb/fichadetramitacao?idProposicao=2487262">Chamber of Deputies</a>.</li>
  <li><strong>The Brazilian AI Plan</strong> sets aside about R$ 23 billion through 2028. Of that total, <strong>R$ 500 million</strong> goes to the AI training, qualification, and reskilling program. It is a start, but it is just over 2% of the plan.</li>
</ul>

<p>Brazil currently has no social protection policy specifically for people who lose their jobs because of AI. Maybe it does not even need to be specific. But if early careers are the most fragile point, that is where public training and protection during the transition make the biggest difference.</p>

<h2>Do the Exercise with Your Own Role</h2>

<p>Data helps you understand the landscape, but the decision is about your work. So here is a simple exercise:</p>

<ul>
  <li><strong>List the tasks in your week.</strong> Not your job title: the tasks. Answering tickets, reviewing contracts, building reports, writing tests, serving customers.</li>
  <li><strong>Sort each one into three groups.</strong> What AI already does well today. What AI amplifies, making you faster or better. And what still depends on you: judgment, relationships, domain knowledge, decisions.</li>
  <li><strong>Look at the size of each group.</strong> If the first is the largest, your role is among those that shrink, and it is worth starting to shift your time to the other two. If the third is the largest, invest in using AI in the second to gain scale.</li>
</ul>

<h2>Conclusion</h2>

<p>AI will not end most occupations, but it will not leave any of them the same. Tasks change first, new roles appear, and some roles start needing fewer people. The greatest pressure is on early careers, where the simple tasks that used to teach the craft are exactly the ones AI learned to do.</p>

<p>What protects you is not running from AI. It is what has always protected people: understanding the fundamentals, knowing the domain, thinking critically, and being able to explain what you want. Nobody does math by hand at work, but those who understand the math notice when the calculator got it wrong.</p>

<p>Do the three-column exercise with your role this week. And if you lead a team, do it with them: reskilling starts when someone stops and looks, task by task, at what is changing.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure">ILO; Gmyrek, P.; Berg, J.; Kamiński, K. et al. "Generative AI and Jobs: A Refined Global Index of Occupational Exposure". ILO Working Paper 140, 2025.</a></li>
  <li><a href="https://www.ilo.org/publications/generative-ai-and-jobs-global-analysis-potential-effects-job-quantity-and">ILO; Gmyrek, P.; Berg, J.; Bescond, D. "Generative AI and Jobs: a Global Analysis of Potential Effects on Job Quantity and Quality". ILO Working Paper 96, 2023.</a></li>
  <li><a href="https://www.ilo.org/resource/news/generative-ai-likely-augment-rather-destroy-jobs">ILO. "Generative AI likely to augment rather than destroy jobs", 2023.</a></li>
  <li><a href="https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf">World Economic Forum. "The Future of Jobs Report 2025".</a></li>
  <li><a href="https://digitaleconomy.stanford.edu/publications/canaries-in-the-coal-mine">Brynjolfsson, E.; Chandar, B.; Chen, R. "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence". Stanford Digital Economy Lab, August 2026 version.</a></li>
  <li><a href="https://digitaleconomy.stanford.edu/news/canaries-interest-rates-and-timinga-more-on-recent-drivers-of-employment-changes-for-young-workers">Brynjolfsson, E.; Chandar, B.; Chen, R. "Canaries, Interest Rates, and Timing: More on Recent Drivers of Employment Changes for Young Workers". Stanford Digital Economy Lab, 2026.</a></li>
  <li><a href="https://blogdoibre.fgv.br/sites/blogdoibre.fgv.br/files/u139/inteligencia_artificial_generativa_e_mercado_de_trabalho_no_brasil.pdf">Duque, Daniel. "Inteligência artificial generativa e mercado de trabalho no Brasil: evidências iniciais sobre ocupação e renda". IBRE/FGV Blog, 2026.</a></li>
  <li><a href="https://www.nber.org/papers/w31161">Brynjolfsson, E.; Li, D.; Raymond, L. "Generative AI at Work". NBER Working Paper 31161 / Quarterly Journal of Economics, 2025.</a></li>
  <li><a href="https://olin.washu.edu/about/news-and-media/news/2023/08/study-ai-tools-cause-a-decline-in-freelance-work-and-incomeat-least-in-the-short-run.php">WashU Olin. "Study: AI tools cause a decline in freelance work and income, at least in the short run" (Hui, Reshef and Zhou), 2023.</a></li>
  <li><a href="https://www.itnews.asia/news/half-of-firms-that-cut-customer-service-staff-due-to-ai-will-rehire-by-2027-623867">iTnews Asia. "Half of firms that cut customer service staff due to AI will rehire by 2027" (Gartner), 2026.</a></li>
  <li><a href="https://hiringlab.indeed.com/2026/07/23/the-labor-market-is-tilting-toward-seniority/">Aidala, F.; Puri, S. "The Labor Market Is Tilting Toward Seniority". Indeed Hiring Lab, 2026.</a></li>
  <li><a href="https://hiringlab.indeed.com/2026/07/08/ai-and-job-postings-from-destruction-to-creation/">Gallacher, G. "AI and Job Postings: From Destruction to Creation?". Indeed Hiring Lab, 2026.</a></li>
  <li><a href="https://survey.stackoverflow.co/2025/ai">Stack Overflow. "2025 Developer Survey: AI".</a></li>
  <li><a href="https://services.google.com/fh/files/misc/2025_state_of_ai_assisted_software_development.pdf">Google Cloud / DORA. "State of AI-assisted Software Development 2025".</a></li>
  <li><a href="https://www.anthropic.com/research/AI-assistance-coding-skills">Anthropic. "How AI assistance impacts the formation of coding skills", 2026.</a></li>
  <li><a href="https://kyodonewsprwire.jp/release/202606150856">PwC. "2026 Global AI Jobs Barometer" (press release), 2026.</a></li>
  <li><a href="https://news.microsoft.com/2024/05/08/microsoft-and-linkedin-release-the-2024-work-trend-index-on-the-state-of-ai-at-work/">Microsoft and LinkedIn. "2024 Work Trend Index".</a></li>
  <li><a href="https://www.bcg.com/publications/2025/ai-at-work-momentum-builds-but-gaps-remain">BCG. "AI at Work: Momentum Builds, but Gaps Remain", 2025.</a></li>
  <li><a href="https://www.camara.leg.br/proposicoesWeb/fichadetramitacao?idProposicao=2487262">Brazilian Chamber of Deputies. Legislative tracking of Bill PL 2338/2023.</a></li>
  <li><a href="https://www.camara.leg.br/proposicoesWeb/prop_mostrarintegra?codteor=2868197">Brazilian Chamber of Deputies. Full text of Bill PL 2338/2023 (as approved by the Senate).</a></li>
  <li><a href="https://www2.camara.leg.br/atividade-legislativa/comissoes/comissoes-permanentes/cctci/apresentacoes-em-eventos/apresentacoes-de-convidados-em-eventos-de-2025/27-08-2025-reuniao-tecnica-mcti-inteligencia-artificial/hugo-valadares-siqueira-mcti-pbia">MCTI. Presentation of the Brazilian Artificial Intelligence Plan to the Chamber's CCTCI committee, 2025.</a></li>
  <li><a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">Mierzwa, Matheus. "My AI Development Workflow".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">Mierzwa, Matheus. "My TDD Journey".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">Mierzwa, Matheus. "AI Governance in Practice".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 18 Oct 2026 15:06:39 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Futuro do Trabalho</category>
      <category>Impacto da IA</category>
      <category>Mercado de Trabalho</category>
      <category>Requalificação Profissional</category>
      <category>IA no Desenvolvimento</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/98f95769-ed7e-43ef-80c2-66ee05480e00" medium="image"/>
    </item>
    <item>
      <title>Skills Hub</title>
      <link>https://www.mmlo.app/en/blog/hub-de-skills</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/hub-de-skills</guid>
      <description>AI Skills Governance for Non-Developers</description>
      <content:encoded><![CDATA[<p class="intro">A Claude Code or Codex <em>skill</em> looks harmless: it's a folder with an instructions file. But that folder often also ships scripts that run on the user's machine. As long as it lives only on your computer, that's fine. Once it starts landing on the computers of 35 people across 7 teams, it becomes software, with every question software brings: who gets it, which version, and who guarantees nobody changed what was approved. In <a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">AI Governance in Practice</a> I promised to show how some skills became internal systems at InBot. I'll start with what holds all of them up: the Skills Hub, and the AI skills governance running underneath it without non-developers ever needing to know it exists.</p>

<h2>When a Skill Leaves Your Machine, It Becomes Software</h2>

<p>Anthropic itself is blunt about this. In the <a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview">Agent Skills</a> documentation, the advice is to use skills only from trusted sources and to treat installing a skill <strong>like installing software</strong>, because a malicious skill can run code that has nothing to do with its stated purpose.</p>

<p>This isn't excessive caution. In February 2026, Snyk analyzed 3,984 skills published on an open marketplace and found <strong>534 with critical security issues</strong> and <strong>76 confirmed malicious</strong>. And the <a href="https://www.theregister.com/2025/09/29/postmark_mcp_server_code_hijacked/">postmark-mcp</a> package shows how it plays out: 15 clean versions building trust, until in the next release a single line started sending a hidden copy of every email to the attacker.</p>

<h3>How we started (and why we couldn't keep going that way)</h3>

<p>At InBot, the skills were good from day one. What surrounded them wasn't. When the AI area started delivering to the whole company, the pace was intense: lots of tasks in the sprint and little time to build the software side properly.</p>

<p>The hub we used back then wasn't built by our team. It was a prototype created earlier to validate the idea, and it did that job well: it proved that distributing skills to the whole team worked. But the prototype stayed in use. It had no authentication, and uploading content required only an admin password. Anyone who found that password could publish something malicious.</p>

<p>Updating the machines had the same profile. Each team had its own installer, which left the team identifier and a read-only credential for the skills repository on the person's computer. Content was audited before reaching the machines, but the mechanism was fragile.</p>

<p>When the delivery pace slowed down, the development team, the most technical team in the company, raised the security risks, and they were absolutely right. The board made a decision I consider the most mature point of this story: <strong>slow down the delivery of skills, automations and platforms to fix the foundation</strong>. On October 3 we did what we internally call the "big bang": in a single day, the backend, the login, the signing and the review flow all changed at once.</p>

<p>Before getting into how it works, one important definition: when I say "Skills Hub", I don't mean just a website. The Hub is the web app, the backend <strong>and the skills themselves</strong> that talk to it, like the one that installs, the one that logs you in and the one that publishes.</p>

<h2>Who Gets What: SSO, Groups and "See = Receive"</h2>

<p>The first question for any distribution system is who gets what. The answer was to take that decision away from any shared password and hand it to <em>SSO</em> (corporate single sign-on), the same login people already use for the company's other systems.</p>

<p>In the terminal, login uses the <a href="https://www.rfc-editor.org/rfc/rfc8628">Device Authorization Grant</a>: the script shows a code, the person opens the browser, signs in with their company account, and that's it. With <a href="https://www.rfc-editor.org/rfc/rfc7636">PKCE</a>, nobody who intercepts that flow can reuse the code. The identity provider is Keycloak, which already offered the device flow, offline sessions and groups out of the box.</p>

<h3>Groups per team, and what's out of reach doesn't exist</h3>

<p>Each team is a group: sales, development, marketing, planning, projects, QA and UX. There are also groups for staging, for editors and for administration. That leads to a simple rule I call <strong>"see = receive"</strong>: the same function decides what the sync downloads to your machine and what the Hub shows you. There's no screen showing one thing while an installer delivers another.</p>

<p>And if you ask for a skill that isn't your team's, the answer isn't "access denied". It's "doesn't exist". It looks like a detail, but it stops anyone from discovering, by trial and error, which skills other teams have. On the other side, revoking someone's access takes effect fast. Login issues two tokens: an <strong>access token</strong>, which lasts only 5 minutes, and a <strong>refresh token</strong>, which is used to request a new access token and expires after 2 days of inactivity. When a person's access is revoked, the refresh token may still be valid, but it becomes useless: it can no longer get a new access token, and what they had stops working within 5 minutes at most. If you've read the post on <a href="https://www.mmlo.com.br/blog/construindo-um-sistema-de-autorizacao-rebac">ReBAC authorization</a>, you'll recognize the idea: access is a relationship, and relationships must be quick to undo.</p>

<h3>The source of truth is the repository, not the database</h3>

<p>There's one design decision that shapes everything that follows: <strong>the files are the source of truth</strong>. A skill's version and who has access to it are recorded in the skill itself, inside the repository. The backend reads that, but it doesn't own it.</p>

<p>In practice, if tomorrow we switch platforms, leave the Hub or even lose the backend data, we still know exactly how everything is organized, and Git history gives us traceability for every change. Anthropic's <a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/enterprise">Skills for enterprise</a> guidance goes the same way: use version control as the source of truth for skills.</p>

<h2>What Arrives Is What Was Approved: Ed25519 Signatures and Human Review</h2>

<p>Now the scenario that changed how I think about this. Imagine someone gets access to the company's GitHub, by any means, and pushes a skill with a malicious script. In a naive model, the sync does exactly its job: <strong>it delivers that script to all 35 machines</strong> the next morning.</p>

<p>This isn't paranoia; it's the pattern of the biggest recent software supply chain incidents. In <a href="https://www.cisa.gov/news-events/alerts/2024/03/29/reported-supply-chain-compromise-affecting-xz-utils-data-compression-library-cve-2024-3094">xz utils</a>, in 2024, versions 5.6.0 and 5.6.1 shipped with a backdoor straight from the official repository. In <a href="https://www.cisa.gov/news-events/alerts/2025/03/18/supply-chain-compromise-third-party-tj-actionschanged-files-cve-2025-30066-and-reviewdogaction">tj-actions/changed-files</a>, in 2025, a compromised GitHub Action exposed the secrets of everyone running it. And the <a href="https://www.cisa.gov/news-events/alerts/2025/09/23/widespread-supply-chain-compromise-impacting-npm-ecosystem">Shai-Hulud</a> worm compromised more than 500 npm packages, published under the developers' own names. In every case, the trusted source was what delivered the malicious code.</p>

<h3>Signing: the machine only accepts what a human approved</h3>

<p>The answer was to separate "being in the repository" from "being approved". Every released version is <strong>digitally signed</strong>, and the user's machine refuses to install anything that doesn't check out.</p>

<ul>
  <li><strong>What gets signed:</strong> the install, sync and login scripts, in their Linux/macOS and Windows versions, plus a list with the <em>hash</em> (the fingerprint) of every released skill. If a single file in the skill changes, the hash changes and the signature no longer matches.</li>
  <li><strong>How:</strong> with plain <a href="https://man.openbsd.org/ssh-keygen">ssh-keygen</a>, which has been able to sign and verify files since OpenSSH 8.1, using a password-protected Ed25519 key. We chose SSH because it's already installed everywhere; even GitHub's documentation says SSH signatures are the simplest to generate. And <a href="https://www.rfc-editor.org/rfc/rfc8032">Ed25519</a> is fast, has small keys and signatures, and doesn't depend on generating a random number for each signature.</li>
  <li><strong>Where it's checked:</strong> on the user's machine, on the server and even in the browser. When someone downloads the installer from the Hub page, the browser itself verifies the signature with WebCrypto, a feature that only reached all major browsers in 2025, with Chrome 137. To deliver a fake installer, someone would have to break into the server <strong>and</strong> the Hub at the same time.</li>
</ul>

<p>With that, the compromised-GitHub scenario changes completely. The change lands in the repository, but <strong>doesn't reach a single machine</strong>, because nobody signed it. And at the same moment, an email goes out to the AI team and the board, warning that something is wrong.</p>

<h3>AI prepares, a human signs</h3>

<p>Every publication becomes a review pull request, and the AI team approves it with a script that shows exactly what will be released, asks for the key password once, merges and signs. The <a href="https://slsa.dev/spec/v1.2/threats-overview">SLSA</a> framework sums up why: a review by a second person could have caught the unauthorized change. Anthropic says the same thing, in so many words, in its enterprise skills guide:</p>

<blockquote><p><strong>"Skill authors should not be their own reviewers."</strong></p></blockquote>

<p>In other words: whoever writes the skill shouldn't be the one who reviews it.</p>

<p>And here's a rule I won't bend: <strong>an AI agent never signs or approves</strong>. AI does almost the whole path (preview, form, PR, ClickUp card), but the signature always comes from a person.</p>

<p>First, for accountability: someone has to answer for what lands on the whole team's computers. Second, for defense: if an agent could sign, a malicious instruction hidden in a file could convince it to release something.</p>

<p>This didn't become a bottleneck, precisely because the review arrives ready: diff, completed form and tests already run. If you want to know why I like PRs with protection rules on the main branch so much, I wrote about it in <a href="https://www.mmlo.com.br/blog/git-flow-para-github-flow">Git Flow to GitHub Flow</a>.</p>

<h2>Non-Developers Don't Even Notice: Single Login, Daily Sync and Staging Channels</h2>

<p>None of this would matter if it weighed on the people using it. Security that gets in the way becomes a workaround, and an AI workaround has a name: <em>shadow AI</em>, using AI outside any control. According to Gartner, as reported by <a href="https://www.infosecurity-magazine.com/news/gartner-40-firms-hit-shadow-ai/">Infosecurity Magazine</a>, 69% of organizations already suspect this kind of use, and more than 40% are expected to have an incident tied to it by 2030. So the rule was clear: <strong>the cost of rigor stays with the builders, not the users</strong>.</p>

<p>For the people using it, the experience is this:</p>

<ul>
  <li><strong>One install, with login.</strong> After that, they don't think about it anymore.</li>
  <li><strong>Updates on their own.</strong> The sync runs every morning and also when the person opens Claude Code or Codex, if it hasn't run yet. It only downloads what changed.</li>
  <li><strong>Works wherever they are:</strong> Claude Code and Codex, on Linux, macOS and Windows.</li>
  <li><strong>Staging and production channels.</strong> People in the testing group get the staging version by default, and can ask for the production one for a specific skill.</li>
</ul>

<h3>Where the complexity showed up: in the publishers</h3>

<p>For the team in general, the system is still simple. The complexity showed up for the <strong>leads who publish skills</strong>, and it isn't technical: there are many options and a flow with rules. Since the files are the source of truth, publishing a skill means touching its files, and many leads aren't technical.</p>

<p>The fix was a skill. The same one that updates skills on the machine learned to <strong>publish</strong>: the person asks in plain language, the skill builds the preview with the diff (what's new, what changed, what was removed), helps fill out a 7-field form (problem, who uses it, input, output, success criteria, risks and origin) and sends it for review. The answer comes within 2 business days. It's AI automating its own governance.</p>

<h3>Testing a skill like a real user</h3>

<p>Before a skill reaches review, it goes through a test suite run by another skill, <strong>testar-skill</strong>. It builds 10 scenarios (5 ideal, 3 ambiguous and 2 extreme), varying how much information the person gives, the quality of the documents and the tone of the message. The expected answer for each scenario is written <strong>before</strong> running, in the same spirit as what I described in <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">my journey with TDD</a>.</p>

<p>Then testar-skill fires off subagents in parallel, each with a clean context and talking like a real user, without knowing it's a test. If the skill under test asks a question, the "user" answers in the same tone; if it gets nowhere in 5 rounds, the result is "doesn't converge". Any external action is deferred, and the person's files are protected during the suite. There are already more than 70 skill test reports on file.</p>

<h3>Measuring usage, showing exactly what leaves the machine</h3>

<p>The last piece came from outside. At Lovable Day, I heard from another company doing work similar to ours that they measure what actually gets used. We were already measuring installs and updates, but not the use of the skills themselves. The challenge was figuring out how to measure it without intruding on anyone's work.</p>

<p>The answer was Claude Code and Codex hooks that record only three things: <strong>which skill was used, on which day and in how many sessions</strong>. Prompts, responses, conversations, file contents or names, commands and third-party skills are never collected. This follows the necessity and transparency principles of Brazil's data protection law, the <a href="https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2018/lei/l13709.htm">LGPD</a>: collect the minimum for the purpose and make clear what is collected.</p>

<p>The transparency is concrete. People see a notice the first time and have a command that shows exactly what leaves their machine, the way <a href="https://code.visualstudio.com/docs/configure/telemetry">VS Code</a> and <a href="https://docs.brew.sh/Analytics">Homebrew</a> do with their telemetry. With this data, we'll retire the skills nobody uses and prioritize improvements to the ones that run the most, which is exactly what Anthropic recommends for skills in companies.</p>

<h2>What Went Wrong: the Cost of Rigor Stays with the Builders</h2>

<p>Supporting two AI tools on three operating systems has a price, and it shows up in the bugs. Four of them are worth telling.</p>

<ul>
  <li><strong>The em dash that became a "network error".</strong> Windows PowerShell 5.1 reads scripts without a BOM (a marker at the start of the file indicating its encoding) using the legacy "ANSI" code page, as <a href="https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_character_encoding?view=powershell-7.5">Microsoft's documentation</a> explains. A single em dash (the character <code>U+2014</code>) in a comment broke parsing of the whole script, and the error reached the user as a "network error". In other words, the bug even lied about its own cause.</li>
  <li><strong>macOS's old bash.</strong> macOS still ships bash 3.2, and the most accepted explanation is licensing: newer versions are GPLv3. Associative arrays only exist from <a href="https://tiswww.case.edu/php/chet/bash/NEWS">bash 4.0</a> on, and the sync used one. Result: on the Mac, it broke silently.</li>
  <li><strong>The false "unsigned version" alarm.</strong> Signing happens seconds after the merge, and sometimes the backend lost track of that commit. That fired the alert email to everyone. The fix was a 30-minute grace period and re-reading the branch before alerting. An alarm that goes off for nothing teaches people to ignore alarms.</li>
  <li><strong>The keyring that signed on its own.</strong> On Linux, the session keyring unlocked the signing key for any process. In practice, any program running in my session could sign without asking for the password. Signing moved to an isolated SSH agent, used only for that.</li>
</ul>

<p>All of them reinforce the same lesson: rigor has a cost, and that cost has to stay with the builders.</p>

<h2>Where to Start with Skills Governance on Your Team</h2>

<p>In about four months, from the first commit in June to the big bang in October, there were <strong>382 commits and 186 releases</strong>. Today there are <strong>55 skills in production</strong> reaching 7 teams, all signed, reviewed and tested. I share these numbers with the same care as in the previous post: they're the result of a team that agreed to slow down deliveries to fix the foundation, and of leads who publish skills even without being technical.</p>

<p>You don't need all of this to get started. If you share skills or prompts with your team, the minimum viable software is this:</p>

<ul>
  <li><strong>Version it.</strong> Skills in a repository, with history. If nobody knows which version is running on whose machine, you don't have one skill, you have several.</li>
  <li><strong>Have someone review.</strong> Whoever writes doesn't approve. It can be a PR with a simple rule on the main branch.</li>
  <li><strong>Know who gets what.</strong> Even if it's just one folder per team, at first.</li>
  <li><strong>Leave signing for later, but not for never.</strong> It comes in once skills start reaching many machines, and the ssh-keygen you already have does the job.</li>
</ul>

<p>And, from day one, simplicity for the users. If I have to choose between one more rule for me and one more step for the user, I'll take the extra rule for me.</p>

<h2>Conclusion</h2>

<p>A skill is software the moment it leaves your machine. Treating it that way means answering three questions: who gets it, which version, and who guarantees nobody changed what was approved. In the Skills Hub, the answers are SSO with groups, the files as the source of truth, and a signature only a human makes.</p>

<p>What makes this workable is what users don't see. People log in once and that's it; the complexity stays with the team that builds, and even the hard part of publishing became a conversation with a skill. If you're starting to share skills with your team, begin with versioning and review, and ask at every decision: does this weigh on me or on the user? In the next post in this series, I want to show how one of these skills became an entire system.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview">Anthropic. "Agent Skills". Claude Docs.</a></li>
  <li><a href="https://platform.claude.com/docs/en/agents-and-tools/agent-skills/enterprise">Anthropic. "Skills for enterprise". Claude Docs.</a></li>
  <li><a href="https://snyk.io/blog/toxicskills-malicious-ai-agent-skills-clawhub">Beurer-Kellner, L. et al. "ToxicSkills" study. Snyk, 2026.</a></li>
  <li><a href="https://www.theregister.com/2025/09/29/postmark_mcp_server_code_hijacked/">Lyons, Jessica. "One line of malicious npm code led to massive Postmark email heist". The Register, 2025.</a></li>
  <li><a href="https://www.cisa.gov/news-events/alerts/2024/03/29/reported-supply-chain-compromise-affecting-xz-utils-data-compression-library-cve-2024-3094">CISA. "Reported Supply Chain Compromise Affecting XZ Utils Data Compression Library, CVE-2024-3094", 2024.</a></li>
  <li><a href="https://www.openwall.com/lists/oss-security/2024/03/29/4">Freund, Andres. "backdoor in upstream xz/liblzma leading to ssh server compromise". oss-security, 2024.</a></li>
  <li><a href="https://www.cisa.gov/news-events/alerts/2025/03/18/supply-chain-compromise-third-party-tj-actionschanged-files-cve-2025-30066-and-reviewdogaction">CISA. "Supply Chain Compromise of Third-Party tj-actions/changed-files (CVE-2025-30066) and reviewdog/action-setup@v1", 2025.</a></li>
  <li><a href="https://www.cisa.gov/news-events/alerts/2025/09/23/widespread-supply-chain-compromise-impacting-npm-ecosystem">CISA. "Widespread Supply Chain Compromise Impacting npm Ecosystem", 2025.</a></li>
  <li><a href="https://slsa.dev/spec/v1.2/threats-overview">SLSA. "Supply chain threats", v1.2.</a></li>
  <li><a href="https://man.openbsd.org/ssh-keygen">OpenBSD. "ssh-keygen(1)".</a></li>
  <li><a href="https://docs.github.com/en/authentication/managing-commit-signature-verification/about-commit-signature-verification">GitHub Docs. "About commit signature verification".</a></li>
  <li><a href="https://www.rfc-editor.org/rfc/rfc8032">Josefsson, S.; Liusvaara, I. "RFC 8032: Edwards-Curve Digital Signature Algorithm (EdDSA)". IETF, 2017.</a></li>
  <li><a href="https://developer.chrome.com/release-notes/137">Google. "Chrome 137 release notes", 2025.</a></li>
  <li><a href="https://developer.mozilla.org/en-US/docs/Web/API/SubtleCrypto/verify">MDN. "SubtleCrypto: verify() method".</a></li>
  <li><a href="https://www.rfc-editor.org/rfc/rfc8628">Denniss, W. et al. "RFC 8628: OAuth 2.0 Device Authorization Grant", 2019.</a></li>
  <li><a href="https://www.rfc-editor.org/rfc/rfc7636">Sakimura, N. et al. "RFC 7636: Proof Key for Code Exchange by OAuth Public Clients", 2015.</a></li>
  <li><a href="https://www.keycloak.org/docs/latest/server_admin/index.html">Keycloak. "Server Administration Guide".</a></li>
  <li><a href="https://www.planalto.gov.br/ccivil_03/_ato2015-2018/2018/lei/l13709.htm">Brazil. "Law No. 13,709/2018 (LGPD)".</a></li>
  <li><a href="https://code.visualstudio.com/docs/configure/telemetry">Visual Studio Code Docs. "Telemetry".</a></li>
  <li><a href="https://docs.brew.sh/Analytics">Homebrew. "Anonymous Analytics".</a></li>
  <li><a href="https://learn.microsoft.com/en-us/powershell/module/microsoft.powershell.core/about/about_character_encoding?view=powershell-7.5">Microsoft Learn. "about_Character_Encoding".</a></li>
  <li><a href="https://tiswww.case.edu/php/chet/bash/NEWS">Ramey, Chet. "Bash NEWS".</a></li>
  <li><a href="https://thenextweb.com/news/why-does-macos-catalina-use-zsh-instead-of-bash-licensing">The Next Web. "Why does macOS Catalina use Zsh instead of Bash? Licensing", 2019.</a></li>
  <li><a href="https://www.infosecurity-magazine.com/news/gartner-40-firms-hit-shadow-ai/">Infosecurity Magazine. "Gartner: 40% of Firms to Be Hit By Shadow AI Security Incidents", 2025.</a></li>
  <li><a href="https://www.mmlo.com.br/blog/governanca-de-ia-na-pratica">Mierzwa, Matheus. "Governança de IA na Prática".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/construindo-um-sistema-de-autorizacao-rebac">Mierzwa, Matheus. "Construindo um Sistema de Autorização ReBAC".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/git-flow-para-github-flow">Mierzwa, Matheus. "Git Flow para GitHub Flow".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">Mierzwa, Matheus. "Minha Jornada com TDD".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 18 Oct 2026 01:01:05 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Governança de IA</category>
      <category>Inteligência Artificial</category>
      <category>Segurança da Informação</category>
      <category>Arquitetura de Software</category>
      <category>Automação</category>
      <category>Claude Code</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/eb60f9a1-b7bb-4cdb-acef-928238d51f24" medium="image"/>
    </item>
    <item>
      <title>AI Governance in Practice</title>
      <link>https://www.mmlo.app/en/blog/governanca-de-ia-na-pratica</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/governanca-de-ia-na-pratica</guid>
      <description>Empower People, Don&apos;t Lay Them Off</description>
      <content:encoded><![CDATA[<p class="intro">Every week, another company announces it has cut part of its team "because of AI". Almost as often, another one freezes AI usage because the token bill scared someone in finance. These are opposite reactions, but they share the same flawed premise: that AI is, above all, a way to spend less on people. Since June 2026 I've led the <em>AI Enablement</em> area at InBot, and what I see day to day is the opposite. In this post I explain how AI governance works in practice here, why it exists to upskill the team rather than shrink it, and where you can start with your own team.</p>

<h2>AI to Cut People or AI to Empower People? What I Heard at Lovable Day</h2>

<p>On October 16 I attended <a href="https://vibedays.com.br/">Lovable Day São Paulo</a>, and two talks summed up what I had already been advocating inside the company.</p>

<p>The first came from <strong>Carol Paiffer</strong>, investor and Shark on Shark Tank Brasil. She said that when she sees an entrepreneur wanting to lay off the team to replace it with AI, she sees a bad move.</p>

<p>The reasoning is simple: if you keep the team and upskill it, the company won't do the same with fewer people. It will do <strong>far more than it does today</strong>, with the same people. It's a gain that compounds, instead of a saving you collect only once.</p>

<p>The second came from <strong>Guilherme Junqueira</strong>, founder of Delta Academy, who also argues for training companies and their teams to use AI instead of replacing them. I'll come back to his talk in the next section, because it touches the other fear: cost.</p>

<h3>Why laying people off to replace them with AI is a bad move</h3>

<p>I agree with both of them, for two reasons that go hand in hand.</p>

<p>The first is that <strong>the team is who knows the business</strong>. A language model knows a lot in general and nothing about how your customer thinks, the exception support handles every week, or the shortcut sales uses to close a deal. At InBot, it's the teams themselves who build the skills and automations. AI only delivers what it delivers because someone with domain knowledge is telling it what to do.</p>

<p>The second is the multiplier effect. An NBER study of customer support agents found an average <strong>14% productivity gain</strong> from generative AI, reaching <strong>34% among the least experienced</strong>. In other words: AI doesn't replace people who are starting out, it shortens their learning curve. That's upskilling, not cutting.</p>

<p>And the data from those who went the other way isn't encouraging. An Orgvue survey found that, among leaders who laid people off after adopting AI, <strong>55% admit they made the wrong call</strong>. Klarna, which became the poster child for "replacing human support with AI", started hiring people for support again after its own CEO acknowledged that the focus on cost led to lower quality. The company didn't abandon AI: it walked back the overreach. And Forrester predicts that half of the layoffs attributed to AI will be quietly reversed.</p>

<p>At InBot, nobody left because of the automations. The time they freed up became delivery. But for that to happen, we had to face the second fear.</p>

<h2>Token Cost vs. the Cost of Inertia: Measure Before You Fear</h2>

<p>Guilherme raised a point I hear often: the fear of token costs. The AI bill grows, someone sees the number, and the natural reaction is to cut usage. His counterpoint was the <strong>cost of inertia</strong>. When a company steps away from the tool because of cost, it saves now and pays later, because it stops learning while the rest of the market keeps learning.</p>

<p>I see token cost in two ways at the same time.</p>

<ul>
  <li><strong>As an investment, not an expense.</strong> When a 3-hour task drops to 13 minutes, the right comparison isn't "how much did it cost in tokens", it's "how much did it cost in tokens versus what those hours of a person's time cost". Three hours that become 13 minutes pay for a lot of tokens.</li>
  <li><strong>As something to be measured, not feared.</strong> AI governance also means knowing what each thing costs. Fear without numbers leads to bad decisions in both directions: cutting what was paying off, or letting run what wasn't.</li>
</ul>

<p>There's one more data point in favor of not freezing: AI prices drop fast. <a href="https://hai.stanford.edu/ai-index/2025-ai-index-report">Stanford's AI Index 2025</a> shows that the inference cost for a model at GPT-3.5 level fell more than <strong>280-fold</strong> between November 2022 and October 2024. Epoch AI estimates that the price for the same performance falls between 9 and 900 times per year, depending on the task.</p>

<p>That doesn't mean the total bill will shrink, because we start using better models and doing more with them. It means that freezing usage today because of today's price is deciding based on a number that's already going stale. What doesn't go stale is how far behind a team falls once it stops learning.</p>

<p>Measuring, however, requires something organized to measure. That's where governance comes in.</p>

<h2>AI Governance in Practice: Teams Build, the Area Makes It Last</h2>

<p>"Governance" usually sounds like bureaucracy: committees, forms, approval queues. That's not what we do. <em>AI enablement</em>, the way I see it, means creating the conditions for the whole company to use AI well, not centralizing AI usage in a single team.</p>

<h3>How the area started</h3>

<p>InBot's AI area wasn't born from a decree. It grew bottom-up. Before it existed, I convinced leadership to adopt Cursor, which at the time was a new tool used mostly outside Brazil. Until then, we did everything by hand. I brought the idea to the other devs, including a more experienced colleague who was wary of using AI at work.</p>

<p>Over time, the dev team itself started standardizing how we used skills so everyone was on the same page, because each person was doing it however they thought best. That's when the board decided to create the AI area: to take this initiative, which started inside the development team, to the rest of the company, with centralization and governance.</p>

<p>Notice that the whole story is about empowering the people who were wary, not replacing them. The area is a continuation of that.</p>

<h3>The path to production</h3>

<p>Today, on AI matters, the leads of Sales, Dev, Marketing, Planning, Projects, QA, UX and Support report to the area. The flow works like this:</p>

<ul>
  <li><strong>The team builds.</strong> The idea and the authorship belong to whoever has the problem and knows the domain.</li>
  <li><strong>The area brings it up to company standards.</strong> We review, adjust and test with scenario suites. We've run suites of <strong>800 automated tests</strong> and <strong>250 CRM runs with 250 successes</strong>. If you've read <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">my journey with TDD</a>, you know I don't trust anything that hasn't been tested.</li>
  <li><strong>The requesting team validates.</strong> Whoever asked for it confirms it solves the real need.</li>
  <li><strong>The area approves the release.</strong> Only then does it go to production.</li>
</ul>

<p>This design avoids two extremes. If everyone uses AI their own way, untested and with no owner, you get <em>shadow AI</em>: AI running in the dark, outside any control. The <a href="https://news.microsoft.com/2024/05/08/microsoft-and-linkedin-release-the-2024-work-trend-index-on-the-state-of-ai-at-work/">Microsoft and LinkedIn Work Trend Index</a> showed that <strong>78% of AI users bring their own tools</strong> to work, and only 39% had received any training from their company. If, at the other extreme, everything depends on a central team to build it, you create a bottleneck and take authorship away from the people who understand the problem.</p>

<h3>The infrastructure that removes friction</h3>

<p>Governance that depends on spreadsheets and goodwill doesn't last. That's why a large part of the area's work was building infrastructure. The main example is the <strong>Skills Hub</strong>: it distributes skills to Claude Code and Codex on Linux, macOS and Windows, with corporate SSO, group-based access, reviewed publishing and signed releases.</p>

<p>In practice, this turns governance into a product. Publishing a new skill isn't opening a ticket and waiting. It's following a flow that already has review, access control and versioning built in. Frameworks like the <a href="https://www.nist.gov/itl/ai-risk-management-framework">NIST AI Risk Management Framework</a> and the ISO/IEC 42001 standard describe well what AI governance needs to cover. The challenge is doing it without friction for the people who are building.</p>

<p>But governance isn't only saying "yes, with standards". Sometimes it's saying "no".</p>

<h2>Governance Also Means Saying "No": Shadow AI, Security and Process</h2>

<p>Two cases illustrate this part of the job well, and neither of them ended with a skill in production.</p>

<h3>When the risk is too high</h3>

<p>Some teams wanted a skill that could interact with corporate email. The idea makes sense from a productivity standpoint: read, summarize, reply. After talking it through with the operations director, we decided it was too risky and turned the idea down.</p>

<p>The reason has a name. Simon Willison calls it the <a href="https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/"><em>lethal trifecta</em></a>: the combination of three things in a single AI agent, namely access to private data, exposure to untrusted content, and the ability to communicate externally. An inbox has all three. Or, in his words:</p>

<blockquote><p><strong>"An attacker can literally email your LLM and tell it what to do!"</strong></p></blockquote>

<p>Put simply: a single incoming email can become an instruction to your agent.</p>

<p>This isn't theoretical. <a href="https://arxiv.org/abs/2509.10540">EchoLeak</a> (CVE-2025-32711) showed how data could leak from Microsoft 365 Copilot through a single email, with zero clicks from the victim. <a href="https://genai.owasp.org/llmrisk/llm01-prompt-injection/">OWASP</a> itself ranks <em>prompt injection</em> as the number one risk in LLM applications and recommends least privilege and human approval.</p>

<p>Turning down an idea like this isn't being against AI. It's what keeps a single skill from becoming an incident. IBM's <a href="https://newsroom.ibm.com/2025-07-30-ibm-report-13-of-organizations-reported-breaches-of-ai-models-or-applications,-97-of-which-reported-lacking-proper-ai-access-controls">Cost of a Data Breach 2025</a> report points out that high levels of shadow AI add an average of <strong>US$ 670,000</strong> to the cost of a breach, and that 63% of breached companies had no AI governance policy. Gartner predicts that more than 40% of organizations will suffer an incident tied to unauthorized AI by 2030.</p>

<h3>When the problem is the process, not the delay</h3>

<p>The second case is subtler. Someone asked for a requirements-gathering skill to speed up the kickoff of client projects. When we looked closely, the problem wasn't the delay. It was the process itself: the way we collected information to start a project wasn't working well.</p>

<p>A skill on top of a bad process would only make the bad process run faster. So, together with the Planning and Projects teams, and led by the AI team, we built a requirements-gathering platform: <strong>InBoarding</strong>, which is currently in pilot.</p>

<p>This may be the most useful governance lesson I've learned so far: before automating, ask whether the problem is speed or design. AI accelerates what already exists. If what exists is wrong, it accelerates the mistake.</p>

<h2>What Changed Without Anyone Leaving (and What the AI Maturity Assessment Says)</h2>

<p>In the area's first four months, we shipped <strong>77 initiatives</strong> and put about <strong>50 skills and automations into production</strong>. Some concrete examples:</p>

<ul>
  <li>The admin API operator turned a <strong>3-hour task into 13 minutes</strong>.</li>
  <li>A bot content package dropped from about <strong>35 to 13 to 16 minutes</strong>.</li>
  <li><strong>1,300 of a client's employees</strong> were imported by a skill.</li>
  <li><strong>InSuporte</strong> generates AI insights, approved by humans, on more than 1,800 support tickets, and replaced 2 dashboards and 4 automations.</li>
</ul>

<p>In none of these cases was the result "now we need fewer people". It was "now this person does in 13 minutes what used to take 3 hours, and spends the rest of the time on something else". That's exactly the compounding gain Carol talked about. The World Economic Forum's <a href="https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf">Future of Jobs Report 2025</a> points the same way: 85% of employers plan to prioritize upskilling their workforce, and 59 out of every 100 workers will need training by 2030.</p>

<h3>A validation, not a finish line</h3>

<p>At Lovable Day, Delta Academy offered participating companies a free AI maturity assessment. I filled it out on behalf of InBot and we got a perfect score, along with a comment that stuck with me: InBot already does, at a corporate level, what they teach their mentees.</p>

<p>I share this carefully, because this result isn't mine. It belongs to the people in every area who stopped what they were doing to learn, build and test new things. The AI area organized the path, but the teams are the ones who walked it.</p>

<p>And it's a validation of the path, not of the finish line. There's still a lot in pilot, a lot to measure better, and a lot that will change with the next models. The score says the direction is right. It doesn't say the work is done.</p>

<h2>Where to Start Tomorrow: 3 Steps for Your Team</h2>

<p>You don't need an AI Enablement title to get started. The story of InBot's AI area began with one person insisting that the team try a new tool. If I had to sum up what worked in three steps, they would be:</p>

<ul>
  <li><strong>Turn your personal usage into something for the team.</strong> That prompt you repeat every week can become a versioned skill the whole team uses. Many of InBot's internal systems were born exactly this way. If you want to see how I work with AI day to day, I wrote about it in <a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">my AI development workflow</a>.</li>
  <li><strong>Map what already exists and give it an owner.</strong> List the AI automations your team already uses, including the ones nobody has mentioned out loud. For each one, define who owns it, how it's tested and who approves changes. That's the minimum viable governance.</li>
  <li><strong>Before automating, ask whether the problem is the process.</strong> If a task is slow because it's poorly designed, automating it just makes the problem faster. Sometimes the right answer isn't a skill, it's redesigning the process.</li>
</ul>

<h2>Conclusion</h2>

<p>The debate about AI in companies tends to get stuck between two fears: losing your job and losing money on tokens. Both fade when the question changes from "how much can we save with AI?" to "how much more can our team do with AI?".</p>

<p>AI governance in practice is what makes that second question answerable. It gives authorship to the people who know the problem, ensures that what goes to production has been tested, knows how to say "no" when the risk is high, and recognizes when the problem isn't about AI at all. None of this requires cutting people. It requires empowering them.</p>

<p>If you lead a team, or you're just the person who uses AI the most on it, start with the mapping: what does your team already do with AI today, and who owns each thing? The answer will probably surprise you. I've written before about the <a href="https://www.mmlo.com.br/blog/inteligencia-artificial-automacao-e-o-futuro-das-profissoes-uma-analise-das-transformacoes-laborais">impact of AI on professions</a>; this post is the practical side of that conversation, told by someone applying it. In upcoming posts, I want to show in more detail how some of these skills became internal systems.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://vibedays.com.br/">VibeDays. "Lovable Day São Paulo", 10/16/2026.</a></li>
  <li><a href="https://www.nber.org/papers/w31161">Brynjolfsson, Erik; Li, Danielle; Raymond, Lindsey. "Generative AI at Work". NBER Working Paper 31161.</a></li>
  <li><a href="https://hrreview.co.uk/hr-news/strategy-news/ai-regret-redundancies-got-them-wrong/381820">Orgvue (via HRreview). "AI regret: over half of firms say their redundancies were 'a mistake'", 2025.</a></li>
  <li><a href="https://www.entrepreneur.com/business-news/klarna-ceo-reverses-course-by-hiring-more-humans-not-ai/491396">Entrepreneur. "Klarna Is Hiring Customer Service Agents After AI Couldn't Cut It on Calls, According to the Company's CEO", 2025.</a></li>
  <li><a href="https://www.maginative.com/article/klarna-dials-back-its-ai-customer-service-strategy-now-its-hiring-humans-again/">Maginative. "Klarna Dials Back its AI Customer Service Strategy, Now It's Hiring Humans Again", 2025.</a></li>
  <li><a href="https://www.forrester.com/blogs/future-of-work-predictions-2026-whats-coming-for-work-and-the-workforce/">Forrester. "Predictions 2026: The Workforce Muddles Through Ambient Disruption", 2025.</a></li>
  <li><a href="https://reports.weforum.org/docs/WEF_Future_of_Jobs_Report_2025.pdf">World Economic Forum. "The Future of Jobs Report 2025".</a></li>
  <li><a href="https://hai.stanford.edu/ai-index/2025-ai-index-report">Stanford HAI. "AI Index Report 2025".</a></li>
  <li><a href="https://epoch.ai/data-insights/llm-inference-price-trends">Epoch AI. "LLM inference prices have fallen rapidly but unequally across tasks", 2025.</a></li>
  <li><a href="https://news.microsoft.com/2024/05/08/microsoft-and-linkedin-release-the-2024-work-trend-index-on-the-state-of-ai-at-work/">Microsoft and LinkedIn. "2024 Work Trend Index".</a></li>
  <li><a href="https://newsroom.ibm.com/2025-07-30-ibm-report-13-of-organizations-reported-breaches-of-ai-models-or-applications,-97-of-which-reported-lacking-proper-ai-access-controls">IBM. "Cost of a Data Breach Report 2025".</a></li>
  <li><a href="https://www.infosecurity-magazine.com/news/gartner-40-firms-hit-shadow-ai/">Infosecurity Magazine. "Gartner: 40% of Firms to Be Hit By Shadow AI Security Incidents", 2025.</a></li>
  <li><a href="https://www.nist.gov/itl/ai-risk-management-framework">NIST. "AI Risk Management Framework".</a></li>
  <li><a href="https://simonwillison.net/2025/Jun/16/the-lethal-trifecta/">Willison, Simon. "The lethal trifecta for AI agents: private data, untrusted content, and external communication", 2025.</a></li>
  <li><a href="https://arxiv.org/abs/2509.10540">Reddy, Pavan; Gujral, Aditya. "EchoLeak: The First Real-World Zero-Click Prompt Injection Exploit in a Production LLM System". arXiv, 2025.</a></li>
  <li><a href="https://genai.owasp.org/llmrisk/llm01-prompt-injection/">OWASP. "LLM01:2025 Prompt Injection".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">Mierzwa, Matheus. "Minha Jornada com TDD".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">Mierzwa, Matheus. "Meu Fluxo de Desenvolvimento com IA".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sat, 17 Oct 2026 23:02:54 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Governança de IA</category>
      <category>Liderança Técnica</category>
      <category>Requalificação Profissional</category>
      <category>Futuro do Trabalho</category>
      <category>Automação</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/9943e8dc-1a30-46fa-9ca8-201595d11f69" medium="image"/>
    </item>
    <item>
      <title>Git Flow to GitHub Flow</title>
      <link>https://www.mmlo.app/en/blog/git-flow-para-github-flow</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/git-flow-para-github-flow</guid>
      <description>Why We Don&apos;t Stick to Trunk-Based Development</description>
      <content:encoded><![CDATA[<p class="intro">A while ago I started considering opening my own tech company. I quickly realized that writing good code is not enough to sustain a business. I was missing what surrounds the software: infrastructure, cloud, and the process that takes code from my machine to production. That is how I arrived at Trunk-Based Development as a possible answer. And it was while trying to implement it that I found out the workflow I actually use is a different one: GitHub Flow. In this post I explain why I left Git Flow, why I did not stay with Trunk-Based Development, and what my routine looks like today.</p>

<h2>Beyond Features: Why I Went Into Infrastructure and Git Workflows</h2>

<p>I have a degree in Business Administration and IT Management, so management was never my weak spot. Software architecture was not either, since I have worked in development for years and contribute to open source projects. What I was missing was something else: the areas that never show up in application code but decide whether a product survives in production.</p>

<p>I started by studying <strong>cloud computing</strong>. From there, version control and Git workflows were the next piece. It was not a lack of technical confidence. It was a deliberate expansion of scope with a concrete goal: if I ever run my own startup, I need to understand not only how to build the product, but how the team ships it continuously and reliably.</p>

<p>That is when <em>Trunk-Based Development</em> came into my research. Not as a trend, but as the answer to a practical question: how do high-performing teams organize code delivery? Before getting there, though, I had to deal with what was no longer working.</p>

<h2>Why Git Flow Stopped Making Sense in My Personal Projects</h2>

<p>Many posts on this topic oversimplify this point. I did not leave Git Flow because I think it is bad. I left because it stopped fitting the context I was using it in.</p>

<p>Git Flow was created by Vincent Driessen in 2010. It has two permanent branches (<code>master</code> and <code>develop</code>) and supporting branches for features, releases, and hotfixes. It was designed for versioned software that has to support several versions in production at the same time. Driessen himself, in a note added years later to the <a href="https://nvie.com/posts/a-successful-git-branching-model/">original post</a>, acknowledged this and started recommending simpler workflows for teams practicing continuous delivery.</p>

<h3>How Git Flow ended up in my projects</h3>

<p>Not through careful analysis. Through habit. When I joined the company where I work today, there was no structured Git workflow. The team was quick to integrate and ship, but it did not yet have the formal processes needed to keep a trunk always healthy.</p>

<p>I went looking for a fix for that specific chaos. I studied Git Flow in depth, learned its pros and cons in practice, and rolled it out at the company. For that context, it worked.</p>

<h3>The problem with copying a workflow into another context</h3>

<p>My mistake was bringing the same structure into my personal projects, just because it was what I already knew. There I worked alone, with the technical discipline I know I have, and Git Flow turned into process standing in my way. I wanted agility, and the develop, release, and hotfix branches created friction that no benefit justified on a one-person team.</p>

<p>That friction is what made me look for an alternative. The first one I found, still in the research phase, was Trunk-Based Development.</p>

<h2>What Trunk-Based Development Is (and Why It Made Me Uneasy)</h2>

<p><em>Trunk-Based Development</em> (TBD) is a branching model in which the team collaborates on a single main branch, the trunk (usually <code>main</code>), and avoids long-lived branches. Paul Hammant, a reference on the subject, puts it this way: developers collaborate on code in a single branch called trunk and resist any pressure to create other long-lived development branches. Merging into the trunk at least once a day is what the community itself calls a <em>key enabler</em> of Continuous Integration.</p>

<p><a href="https://dora.dev/capabilities/trunk-based-development/">DORA</a> (DevOps Research and Assessment, now part of Google Cloud) defines three signs that TBD is healthy:</p>

<ul>
  <li>three or fewer active branches in the repository;</li>
  <li>merging into the trunk at least once a day;</li>
  <li>no code freezes and no dedicated integration phases.</li>
</ul>

<p>DORA's research links TBD to high-performing delivery teams. One caveat: blogs often quote figures about how many times more likely elite teams are to use TBD, but I could not find that number in DORA's primary material. So I do not repeat it. What is confirmed is the correlation: TBD is the branching practice of teams that ship more often and with less friction.</p>

<p>I decided to migrate with enthusiasm. But the more I read about the canonical practice, with fast integration into the trunk and minimal ceremony, the riskier it looked to me, even with tests and CI. I wanted TBD's agility, but I also wanted a formal safety layer before anything touched the main branch. And that layer, as I found out later, I already had.</p>

<h2>The Discovery: While Implementing TBD, I Realized I Use GitHub Flow</h2>

<p>Only when I reviewed what I was actually doing did I notice the naming mistake. I thought I was applying TBD, but what I use is <em>GitHub Flow</em>. The <a href="https://docs.github.com/en/get-started/using-github/github-flow">official GitHub documentation</a> defines the workflow in six steps:</p>

<ol>
  <li>create a branch;</li>
  <li>make changes;</li>
  <li>open a pull request;</li>
  <li>address review comments;</li>
  <li>merge the pull request;</li>
  <li>delete the branch.</li>
</ol>

<p>The key point is that changes do not reach the default branch until the merge. In other words, the main branch only receives code after the pull request has been reviewed and approved.</p>

<h3>The difference between GitHub Flow and TBD</h3>

<p>According to a <a href="https://mergify.com/blog/trunk-based-development-vs-gitflow-which-branching-model-actually-works">technical comparison</a> I found during my research, the difference is not branch lifetime, since both use short-lived feature branches. It is the review requirement. GitHub Flow requires the pull request as a formal gate before integration. "Pure" TBD favors merging into the trunk quickly, with minimal ceremony. That exact detail is what resolved my unease.</p>

<p>GitHub's documentation treats this review as an established practice, not a formality:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"Pull request review is so valuable that some repositories require an approving review before pull requests can be merged."</strong></p>
</div>

<p>In other words: review is a requirement, not a nice-to-have, before code reaches the main branch.</p>

<p>Compared to Git Flow, GitHub Flow also removes the multiple permanent branches. You keep one production branch and temporary feature branches, created for a single change and deleted after the merge.</p>

<p>I see GitHub Flow as part of the same family of continuous delivery practices as TBD, with one extra safety gear. And that gear fits my profile: I trust my technical discipline, but I prefer my code to be reviewed before it touches the trunk, even when the reviewer is me at another moment or an AI I use in the process. If you want to see how AI fits into that process, I wrote about it in <a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">my AI development workflow</a>.</p>

<h2>My Routine with GitHub Flow: Required PRs, Short Branches, Continuous Deployment</h2>

<p>Today I use GitHub Flow in both personal projects and commercial client projects. In practice, that means:</p>

<ul>
  <li>a short branch for each change;</li>
  <li>small, frequent pull requests;</li>
  <li>a main branch ready to deploy as soon as a PR is merged.</li>
</ul>

<h3>What I have not used yet: feature flags</h3>

<p>I have not used <em>feature flags</em> in practice yet. They are more tied to pure TBD, as a way to handle unfinished work that has to land in the trunk before it is ready for users. Martin Fowler and Pete Hodgson describe <a href="https://martinfowler.com/bliki/FeatureToggle.html">feature toggles</a> as a powerful technique that lets teams change system behavior without changing code.</p>

<p>Since in GitHub Flow a PR only goes in when the change is ready for review, I never needed to merge something unfinished. Even so, it is something I need to try to understand the implications in practice, not just in theory.</p>

<p>None of these practices holds up on its own, though. They depend on guardrails.</p>

<h2>Discipline as a Prerequisite: Tests, CI, and Guardrails</h2>

<p>The first guardrail, and the most important one for me, is something I already practiced before researching TBD or GitHub Flow: unit, integration, and regression tests on everything I write. I told that story in <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">my TDD journey</a>.</p>

<p>With the new workflow, those tests stopped being just a quality habit and became a structural part of the process. Any major mistake I make breaks the tests and automatically blocks the merge in CI, before the PR is even approved.</p>

<p>Beyond CI, the other guardrails are behavioral:</p>

<ul>
  <li><strong>Consistent branch naming</strong>, even for branches that live only a few hours.</li>
  <li><strong>Small PRs</strong>, to avoid large, slow reviews.</li>
  <li><strong>Fast integration</strong>, without letting work pile up outside the main branch.</li>
</ul>

<p>As for whether this kind of workflow depends more on discipline or on team size, I would rather be clear about the limits of what I know. I do not have enough experience with large teams to speak confidently about how TBD or GitHub Flow behave at that scale. What I can say, because I live it, is that for small, disciplined teams the model works very well.</p>

<p>Which leads to the opposite question: when not to use it.</p>

<h2>Trade-offs: When Git Flow Is Still the Right Choice</h2>

<p>Acknowledging the limits of what I learned also means acknowledging that Git Flow is still excellent in some contexts. Regulated industries, such as banks and fintechs, tend to prefer more controlled workflows, with an explicit audit trail, manual approvals, and well-defined promotion environments between development, staging, and production.</p>

<p>Leaner workflows like TBD and GitHub Flow depend on strong automated test coverage and quick reviews. Even then, a bug that slips past the automated checks gets close to production fast. That risk profile is not always compatible with compliance requirements.</p>

<p>In the end, this is not a purely technical decision. It is a leadership decision. Every company has its own context, and it is up to whoever leads technically to weigh the pros and cons based on what they know about the company and the team, and then decide what makes the most sense there. The models work well, each in the right place.</p>

<h2>Conclusion</h2>

<p>I left Git Flow because it became friction in a context it was not built for. I went after Trunk-Based Development for its agility, and I did not stay with it because I wanted a formal review before merging. Along the way, I found out that this combination already had a name: GitHub Flow.</p>

<p>If you are torn between Git Flow, GitHub Flow, and Trunk-Based Development, my advice is to test them. Start with personal projects or lower-risk client work. Watch whether the workflow holds up with the discipline you and your team have today, not the discipline you wish you had.</p>

<p>And if, during your research, you find out, like I did, that you use a different workflow than you thought, that is fine. The workflow's name matters less than understanding why it works in your context.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://docs.github.com/en/get-started/using-github/github-flow">GitHub Docs. "GitHub flow".</a></li>
  <li><a href="https://trunkbaseddevelopment.com/">Hammant, Paul. "Trunk-Based Development". trunkbaseddevelopment.com.</a></li>
  <li><a href="https://mergify.com/blog/trunk-based-development-vs-gitflow-which-branching-model-actually-works">Mergify. "Trunk-Based Development vs Gitflow: Which Branching Model Actually Works?".</a></li>
  <li><a href="https://nvie.com/posts/a-successful-git-branching-model/">Driessen, Vincent. "A successful Git branching model". nvie.com, 2010.</a></li>
  <li><a href="https://martinfowler.com/bliki/FeatureToggle.html">Fowler, Martin; Hodgson, Pete. "Feature Toggles (aka Feature Flags)". martinfowler.com.</a></li>
  <li><a href="https://dora.dev/capabilities/trunk-based-development/">DORA / Google Cloud. "Trunk-based development". dora.dev.</a></li>
  <li><a href="https://www.datacamp.com/tutorial/git-branching-strategy-guide">DataCamp. "Git Branching Strategy: A Complete Guide".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">Mierzwa, Matheus. "Minha Jornada com TDD".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">Mierzwa, Matheus. "Meu Fluxo de Desenvolvimento com IA".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Thu, 09 Jul 2026 15:51:54 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>workflow de desenvolvimento</category>
      <category>Arquitetura de Software</category>
      <category>Projetos Pessoais</category>
      <category>TDD</category>
      <category>qualidade de software</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/aacca2a0-66a7-4b6b-91ea-2d93a66bcd99" medium="image"/>
    </item>
    <item>
      <title>Building a ReBAC-Inspired Authorization System</title>
      <link>https://www.mmlo.app/en/blog/construindo-um-sistema-de-autorizacao-rebac</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/construindo-um-sistema-de-autorizacao-rebac</guid>
      <description>A Comprehensive Case Study</description>
      <content:encoded><![CDATA[<p class="intro">Reading tutorials and comparing SaaS platforms teaches you the concept of authorization. It does not teach you how an authorization system works on the inside. I only learned that by building one from scratch and living with every consequence of that choice. This post tells the story of <strong>GateKey</strong>, an authorization system inspired by ReBAC (relationship-based access control) that I built to explore the space between the simplicity of Auth0 and the completeness of Keycloak. It also covers what building it taught me about designing systems that make binary decisions about people's access.</p>

<h2>Why ReBAC Matters: Beyond Role-Based Access Control</h2>

<p>For most developers, authorization comes down to one of two questions: "what is this user's role?" (RBAC) or "can this user perform this action?" (a binary check). Both approaches work well, until the day they stop working.</p>

<p>The reason is that, in practice, access is not binary. It is relational. A user does not have a role: they have a relationship with a resource. "Manager of project X", "member of team Y", "guest in workspace Z". Each relationship carries different permissions. RBAC tries to flatten those relationships into named roles, and that simplification breaks as soon as requirements get complex: multi-tenant systems, delegated access, context-dependent permissions.</p>

<p><strong>ReBAC</strong> (<em>Relationship-Based Access Control</em>) flips the model. Instead of assigning roles to users, you define relationships between entities: who owns what, who can edit what, who invited whom. The component that makes the permission decision, the <strong>PDP</strong> (<em>Policy Decision Point</em>), does not ask "does this user have the Editor role?". It asks "what is the relationship between this user and this resource?" and evaluates the permission based on that relationship.</p>

<p><strong>Granularity is ReBAC's greatest strength and also its price.</strong> You define permissions at the level of each resource, tied to specific relationship paths. You can revoke one person's access to a single resource without touching anything else. In return, no permission is ever created implicitly: every access requires an explicit <em>binding</em>. For simple cases, that is overkill. For serious systems, it is what works.</p>

<h3>ReBAC is about the kind of control, not scale</h3>

<p>The question that comes up most is "when do I need ReBAC?". The answer is not "when you have 10,000 users" or "when the system grows". It is <strong>when the granularity you want to maintain goes beyond what RBAC can express</strong>. A two-person SaaS may need ReBAC if it requires a granular audit trail and the ability to revoke access to individual resources. A 10,000-person company may do just fine with RBAC if its permission model is simple.</p>

<p>So the decision is not about size. It is about the level of control you need and, as a result, how much complexity you are willing to accept in your authorization layer. I decided to build an entire system precisely to feel that complexity firsthand.</p>

<h2>Learning by Building Complete Systems, Not Toy Projects</h2>

<p>The project started with a decision: I would not build a demo, or an isolated feature. I would build a complete system.</p>

<p>There is a big difference between "building a to-do app to learn React" and "building a production system where React is essential". In the first case, you learn React. In the second, you learn architecture, state management, performance trade-offs and how to structure code that ages well. The same technology leads to completely different lessons.</p>

<p>Building GateKey meant building everything a production authorization system needs:</p>

<ul>
<li>multi-tenancy, with Organizations that contain Workspaces;</li>
<li>session management;</li>
<li>MFA and <em>magic link</em> login;</li>
<li>access auditing;</li>
<li>real-time permission evaluation;</li>
<li>token expiration and immediate access revocation.</li>
</ul>

<p>Each item is a small, isolated problem. Together, they force you to understand how authorization fits into the whole system. If I had studied only "resources and capabilities", I would have learned the concept. Building alongside the multi-tenant hierarchy, real-time sync and the interplay between authentication and authorization, I learned something else: the friction between those parts, the decisions about where each piece of logic should live and the failures that only show up when everything is connected.</p>

<h3>Open source as a practice</h3>

<p><strong>Everything I build to learn, I release as open source.</strong> I will not turn this into a political discussion here. It is a practice: if you are building a reference system, other people need to be able to see it, critique it and build on top of it. The reasoning behind that choice lives in another post, on <a href="https://www.mmlo.com.br/blog/gpl-vs-mit">GPL vs MIT</a>.</p>

<p>With the scope defined, the next step was not opening the editor. It was writing.</p>

<h2>From PRD to Architecture: How I Designed GateKey</h2>

<p>I started with a PRD (<em>Product Requirements Document</em>). It defined what I wanted to explore:</p>

<ul>
<li>a pure ReBAC model at the core, not a hybrid;</li>
<li>a three-level multi-tenant structure: Root org → Organizations → Workspaces;</li>
<li>permissions defined as relationships: who can do what on which resource;</li>
<li>real-time permission updates: if I revoke access, it is gone immediately;</li>
<li>an audit trail for every permission decision;</li>
<li>a system usable by a single developer, not just theory.</li>
</ul>

<p>The PRD was incomplete, and that is normal. As I built, I found out what I did not know, and the PRD evolved. The core principle, though, never changed: <strong>relationships first, everything else follows.</strong> Three decisions shaped the architecture.</p>

<h3>Decision 1: no global roles, every access is a scoped binding</h3>

<p>Most systems start with RBAC and add relationships later. It is pragmatic, because it is easy to implement at first. And it is fatal, because you end up with two partial systems talking to each other.</p>

<p>In GateKey I started from the other end: the relationship between user and resource comes first. There is no global role with implicit permissions. Every permission is a <em>binding</em>: a user, a role and a resource (or the whole workspace), always inside a workspace. A role on its own grants nothing. It only applies where a binding exists. The PDP evaluates each request by asking: "is there a binding that connects this user to this resource with a role that has this capability?".</p>

<p>The PRD called for pure ReBAC. What the code implements sits somewhere in between: it is closer to resource-scoped RBAC, with inheritance between resources, than to Zanzibar-style ReBAC. The "relationship" the PDP sees is "this user has this role on this resource", plus the containment relationship between a resource and its parent. There are no free-form relationship types such as "owner" or "invited".</p>

<p>That is more expressive and also more complex. Every permission has to be defined explicitly. You have to model the relationships and think about propagation: if I give someone access to a container, does that access flow down to the items inside it or not?</p>

<h3>Decision 2: a three-level hierarchy, with context-dependent inheritance</h3>

<p>The hierarchy has three levels: Root org → Organizations → Workspaces. Bindings live inside the Workspace. Within it, a binding can apply to the whole Workspace or to a specific resource. And one resource type can inherit from another (an item inside a container, for example), but only if the type definition declares that inheritance. That solves common multi-tenancy problems, but it creates ambiguity.</p>

<p>To evaluate a permission, the PDP may have to walk the entire hierarchy before returning a verdict. Can the user edit this document? I check for a direct binding on the document. None? If the "document" type inherits from a container, I check for a binding on the parent container. None? I check for a binding on the Workspace that contains the document. Still none? Denied. Now add caching, cache invalidation and the need to record which binding actually granted the permission.</p>

<p>That decision solved the problem I wanted to solve and created problems I had not anticipated. That is exactly what building a system teaches you.</p>

<h3>Decision 3: separating capabilities from roles</h3>

<p>In traditional RBAC, a role holds its permissions directly: "Editor = edit, delete, comment". I inverted that. Roles hold capabilities, and capabilities hold permissions. A role is a set of capabilities, and an action requires a capability.</p>

<p>That adds a layer of indirection, but it lets me compose permissions without duplicating anything: I create new roles by combining existing capabilities. The cost shows up in auditing. When the PDP says "denied", which capability was missing? Which role applied? Without clear logs, you cannot trace it.</p>

<p>Many of these decisions did not come out of nowhere. They came from a Google paper.</p>

<h2>Zanzibar in Practice: Authorization as a Graph Problem</h2>

<p>In 2019 Google published the paper on <strong>Zanzibar</strong>, the company's authorization system. It is not a product you can download. It is a design pattern. The core idea is that evaluating authorization is a graph problem: the relationships between entities form a graph, and checking a permission means asking "can I get from the user to the resource by following allowed relationship paths?".</p>

<p>Zanzibar also separates storage (the relation tuples) from evaluation (the PDP). That is essential. It means the PDP can be <em>stateless</em> and run in parallel. A permission request does not need a transaction: it needs a snapshot of the relationship graph and the ability to traverse it.</p>

<p>GateKey borrowed that separation. Bindings are records <code>(user, role, resource)</code> inside a workspace. The PDP is stateless. Multiple PDPs can evaluate permissions in parallel against the same binding store. The difference is that GateKey's PDP does not traverse an arbitrary graph like Zanzibar does: it runs a fixed number of lookups, in the order resource, parent resource and workspace.</p>

<h3>The audit trail problem</h3>

<p>Standard ReBAC returns a binary: ALLOW or DENY. But the "why" matters as much as the answer. So I added an audit trail: the PDP returns the decision and also the reason: direct binding, binding inherited from the parent resource, workspace binding or no binding found. "Denied because no binding was found at any level" is very different from "allowed by the parent container's binding, inherited by the document".</p>

<p>This part is entirely custom. There is no standard for it in ReBAC implementations, and it was an experiment. But it addressed a concrete need: when a user complains about access, you need to know why.</p>

<p>That need grows even bigger when every customer has its own rules.</p>

<h2>Multi-Tenancy in Practice: Designing for Isolation</h2>

<p>The three-level hierarchy is not theoretical. It exists because well-known products use these structures: Slack has Workspaces, Google has Organizations and Projects, Figma has Teams and Files. These are common patterns, not edge cases.</p>

<p>Building for this forces you to answer hard questions:</p>

<ul>
<li>Can an Organization admin see every Workspace, or do Workspace admins have autonomy?</li>
<li>Can a permission at the Organization level override a denial at the Workspace level, or the other way around?</li>
<li>When someone leaves the Organization, what happens to their Workspace memberships?</li>
<li>How do you audit who has access to what when permissions depend on context?</li>
</ul>

<p>These stop being policy questions and become architecture questions. They determine how the data is stored, how the PDP evaluates permissions and which edge cases you will run into.</p>

<p>Multi-tenancy is not "having a <code>tenant_id</code> column". It is designing around the assumption that tenants may have conflicting policies and that the system has to respect both.</p>

<p>One decision I did not expect to treat as architecture was the interface.</p>

<h2>Design as a Reflection of the System: GateKey's Visual Identity</h2>

<p>GateKey's first mockup was generic. Default-style cards, buttons with no clear logic, icons that meant nothing in particular. It looked like any admin dashboard, because the specification was vague.</p>

<p>The specification said things like "octagon in the empty state", "cards as a graph", "cut corners". It sounds specific when you read it. The moment a designer or an AI has to execute it, each sentence turns into dozens of possible interpretations. "Cut corners" could be a 30-degree chamfer, a hard 90-degree edge or something else entirely. Whoever executes picks the easiest interpretation, which is almost never the one you meant.</p>

<p><strong>The problem was not the execution. It was the specification.</strong></p>

<h3>From prose to code: the specification as the source of truth</h3>

<p>So I inverted the process. Instead of describing design intent and hoping it would come out right, I took the code that already worked, extracted its tokens (exact colors, spacing values, animation keyframes) and documented those tokens as the specification. The source of truth moved from prose to the system itself.</p>

<p>The second iteration was not "let's make it pretty". It was "let's document what works and make it consistent".</p>

<h3>Every visual choice maps to a function</h3>

<ul>
<li><strong>Sharp corners, no border-radius:</strong> authorization has no curves. ALLOW and DENY are final, and the geometry conveys that decisiveness without explanation.</li>
<li><strong>Monospace font for system data:</strong> IDs, tokens, capabilities and permissions are not interface copy. They are data. They need to be easy to scan, easy to copy and visually distinct from UI labels.</li>
<li><strong>Amber as the only accent color:</strong> in a system where everything is denied by default, whatever is allowed or active needs to stand out. The color does not decorate. It signals "this passed the check".</li>
<li><strong>Real-time animations (scan, pulse, status indicators):</strong> an authorization system never rests. It is constantly evaluating sessions, expiring tokens and logging events. The interface cannot look static, because the system is not.</li>
</ul>

<p>The only arbitrary choice was amber itself. It could have been green or blue. Once it was chosen, every decision that followed had a structural reason.</p>

<h2>What I Learned Building an Authorization System</h2>

<p>You can read about ReBAC, study the Zanzibar paper and watch videos on authorization patterns. None of that shows you what happens when you build a system with those patterns. Four lessons stuck.</p>

<h3>Walking the hierarchy has a cost</h3>

<p>The audit trail adds overhead. Every permission check may walk the hierarchy: direct binding? Parent resource binding? Workspace binding? That is three queries, or one with good caching. At scale, it adds up. You cannot avoid it, only plan for it.</p>

<h3>The line between identity and authorization is blurry</h3>

<p>I kept running into cases where identity and authorization overlapped. If MFA is enabled, does it change permission evaluation? If a session expires, do pending permission requests return the previous decision or fail? Where does authentication end and authorization begin? These questions are not theoretical. They determine how you structure your code.</p>

<h3>You will miss a role hierarchy later</h3>

<p>I implemented roles as flat lists of capabilities, with no inheritance. If you create a "Moderator" role and a "Senior Moderator" role with a few extra capabilities, you end up duplicating everything. That does not scale. Production systems need role hierarchies ("Senior Moderator extends Moderator"), but that complicates the PDP. I learned this by building, not by reading.</p>

<h3>Convex is great for development, but not for this product in production</h3>

<p>I chose Convex for its real-time reactivity and for not having to sync anything by hand. For building fast, it is excellent. For an authorization system that anyone should be able to self-host, it is a problem: a system meant to run anywhere should not depend on a proprietary managed service.</p>

<p>That tension is still unresolved. If GateKey goes to production, I will revisit the Convex decision. The likely path is replacing it with Postgres and an ORM that supports reactive queries, or accepting the lock-in as an explicit trade-off.</p>

<p>With those lessons on the table, it is easier to say what GateKey is today, and what it is not.</p>

<h2>What GateKey Is Today</h2>

<p>GateKey is a complete authorization system. It has authentication (MFA and magic links), resource-scoped bindings, multi-tenancy, real-time updates, access auditing and a working UI. The code is open on <a href="https://github.com/MierzwaMatheus/gatekey">GitHub</a>.</p>

<p>It does not replace Keycloak, which is far more complete. It is also not a direct alternative to Auth0, because it has no SaaS offering and never will for anyone but me. It fills a specific niche: engineers who want to understand how authorization works, who are willing to learn the system and who operate where "a complete system on my own infrastructure" matters more than "a managed SaaS".</p>

<h3>What it deliberately is not</h3>

<ul>
<li><strong>It is not a general-purpose auth solution.</strong> Most apps do not need ReBAC. For those, use Auth0 or Supabase.</li>
<li><strong>It is not self-hosted in the traditional sense.</strong> The UI and logic run anywhere, but the data lives in Convex. That is a concrete limitation.</li>
<li><strong>It is not "done".</strong> It works, but production patterns (role hierarchies, large-scale caching, complex audit requirements) have not been fully explored.</li>
<li><strong>It is not minimal.</strong> It includes authentication features (MFA, magic links) that, strictly speaking, an authorization system should not have. They are there because production systems need them.</li>
</ul>

<h3>Who it works for</h3>

<ul>
<li>Engineers building their own authorization system who want a reference implementation.</li>
<li>Teams that understand authorization deeply and want to own their implementation.</li>
<li>Products that need granular multi-tenant control and can absorb the complexity.</li>
<li>Anyone who wants to learn ReBAC by building, not just reading.</li>
</ul>

<p>It does not work for startups that need to ship fast, small teams without authorization expertise or anyone who wants "set it and forget it" auth.</p>

<h2>Conclusion</h2>

<p>This post was not a tutorial. It was a case study, and the difference matters. A tutorial would show "this is how ReBAC works" with isolated examples. A case study shows what happens when you commit to understanding a concept well enough to build an entire system around it.</p>

<p>The learning was not in the theory behind Zanzibar. It was in discovering that the theory does not deal with session expiration, audit trails, real-time invalidation, or the fact that people do not create bindings in clean, testable patterns. Systems in use are messier, and understanding that requires building.</p>

<p>If you are thinking about building your own authorization system, start here:</p>

<ul>
<li><strong>Write a clear specification</strong> before any code.</li>
<li><strong>Build one complete feature at a time.</strong></li>
<li><strong>Test your audit trail early.</strong> It is hard to add later.</li>
<li><strong>Accept that the technology you choose</strong> (Convex, Postgres or anything else) will constrain your design.</li>
<li><strong>Document your decisions</strong>, not for others, but for yourself a year from now.</li>
</ul>

<p>If you are evaluating authorization systems for your product, GateKey can serve as a reference to study, even if you choose something else. Understanding what goes into an authorization system makes you a better evaluator of the ones already on the market.</p>

<p>Understanding a concept and understanding a system are different things. It is the difference between reading about ReBAC and building one. The second teaches more and costs more. This project was built on that trade-off.</p>

<h2>References</h2>

<ul class="references">
<li><a href="https://research.google/pubs/pub48190/">Google Research. "Zanzibar: Google's Consistent, Global Authorization System". USENIX Security Symposium, 2019.</a></li>
<li><a href="https://www.keycloak.org/">Keycloak Official Documentation. "Identity and Access Management".</a></li>
<li><a href="https://www.openpolicyagent.org/">Open Policy Agent. "Policy as Code".</a></li>
<li><a href="https://auth0.com/docs/get-started/authentication-and-authorization">Auth0. "Authorization Fundamentals".</a></li>
<li><a href="https://csrc.nist.gov/publications/detail/sp/800-162/final">NIST. "Access Control Models and Policies". SP 800-162.</a></li>
<li><a href="https://martinfowler.com/articles/microservices.html">Martin Fowler. "Microservices Security".</a></li>
<li><a href="https://casbin.org/">Casbin. "Authorization Library for Multiple Languages".</a></li>
<li><a href="https://docs.convex.dev/">Convex. "Backend Development Platform with Real-Time Reactivity".</a></li>
<li><a href="https://www.universalprinciples.com/">Lidwell, Holden, Butler. "Universal Principles of Design".</a></li>
<li><a href="https://github.com/MierzwaMatheus/gatekey">GateKey Repository. "Open Source ReBAC Authorization System".</a></li>
<li><a href="https://www.mmlo.com.br/blog/gpl-vs-mit">Mierzwa, Matheus. "GPL vs MIT".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sun, 31 May 2026 16:09:18 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Arquitetura de Software</category>
      <category>Open Source</category>
      <category>Desenvolvimento Tecnológico</category>
      <category>Projetos Pessoais</category>
      <category>TypeScript</category>
      <category>NodeJS</category>
      <category>Convex</category>
      <category>Free Software</category>
      <category>Desenvolvimento Pessoal</category>
      <category>Tecnologia e Sociedade</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/9dc49cb4-0587-4422-a2c0-fe9a7ad4c721" medium="image"/>
    </item>
    <item>
      <title>Rubrica</title>
      <link>https://www.mmlo.app/en/blog/rubrica</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/rubrica</guid>
      <description>The open-source portfolio that became infrastructure</description>
      <content:encoded><![CDATA[<p class="intro">Every developer reaches a point where LinkedIn isn't enough. You want a space that's yours, that showcases your technical skills, that tells your story. Most people create a simple portfolio, leave it up for a few months, and abandon it when maintenance feels like too much work. I did things differently: I built it three times. And on the third try, the project became something else.</p>

<p>This post tells the story of <strong>Rubrica</strong>, the open-source portfolio system that I started as a personal portfolio and that ended up becoming, 211 commits later, a complete infrastructure that any developer can fork, customize, and publish.</p>

<h2>Why the Third Version of a Portfolio Changes Everything</h2>

<p>My portfolio has always been a technical reflection. Each version documented where I was as a developer at that moment: the technologies I mastered, the patterns I preferred, the problems I could solve. The first version was simple, honest in its limitations. The second, more robust. The third, different in essence.</p>

<p>It wasn't planned to become what it did. I started with a simple question: I want a portfolio that has a blog, projects, and a resume, and that I can manage without touching code every time. The obvious answer was an admin panel. And when the admin panel was ready, something changed in the perception of the project.</p>

<p>The turning point wasn't when I added payments. It wasn't when I integrated AI. It was when I created the <strong>file management system</strong>. At that moment, I looked at the code and thought: any developer could use this. Not as a template, but as a foundation. A solid technical structure that frees you to focus on what really matters: your identity.</p>

<p>That's the philosophy of Rubrica. <strong>The foundation is ours, the identity is yours.</strong></p>

<p>This has a name: a living portfolio. It's not a portfolio frozen in time. It's a project that grows with you, that actively demonstrates the technical decisions you make today. When someone visits Rubrica and sees React 19, Convex, Tailwind 4, 472 tests with Vitest, and GDPR compliance, they're not looking at a resume. They're seeing code in production.</p>

<h2>What Rubrica Delivers: From Landing Page to Admin Panel</h2>

<p>Before diving into the technical decisions, it's important to understand the scope of what the project has built. Rubrica isn't a one-page template. It's a system with complete public and administrative areas.</p>

<p>In the public area, you have:</p>

<ul>
  <li><strong>Home</strong> with hero section, about section, services, testimonials, and availability badge</li>
  <li><strong>Portfolio</strong> with project grid, technology filters, detail modals, and image carousel</li>
  <li><strong>Blog</strong> with posts, tags, search, and featured content system</li>
  <li><strong>Resume</strong> with structured sections and export to ATS-optimized PDF</li>
  <li><strong>Commercial proposals</strong> with handwritten digital signature and PDF contract generation</li>
</ul>

<p>In the admin area, you manage everything without touching code:</p>

<ul>
  <li>Dashboard with usage statistics</li>
  <li>Complete CRUD for projects, posts, resume, and proposals</li>
  <li>User management with roles</li>
  <li>Audit logs of all actions</li>
  <li>Plugin system with granular control</li>
  <li>GDPR compliance panel with data export and anonymization</li>
  <li>Payment Links via Stripe</li>
  <li>AI Resumes: AI-powered resume rewriting pipeline</li>
</ul>

<p>There are 11 admin routes protected by authentication. The database has more than 20 tables in the schema. The README has been rewritten as a complete deployment guide: from fork to custom domain in 10 steps.</p>

<p>This answers the question that developers usually ask: "Why so much complexity for a portfolio?" Because Rubrica isn't just a portfolio. It's a work tool.</p>

<h2>Technical Decisions: React 19, Convex, and Why I Migrated from Supabase</h2>

<p>Rubrica's stack wasn't chosen to impress. It was chosen to work.</p>

<p><strong>React 19 and Vite</strong> as a base don't need a long justification: React leads with 41.6% of professional developers in the Stack Overflow Developer Survey 2024. Vite surpassed Webpack in npm downloads in July 2025 and had the highest satisfaction rate in the State of JS 2024. Create React App was officially deprecated in February 2025, recommending Vite as an alternative. The choice is in the mainstream for good reasons.</p>

<p>The backend story is more interesting.</p>

<p>Rubrica started with Supabase. It was the logical choice: structured SQL, ready-made auth, integrated storage, huge community (99k stars on GitHub). It worked. The problem wasn't technical, it was philosophical.</p>

<p>In the middle of development, I started using <strong>Convex</strong> in other projects. And I realized something: Convex fit exactly into Rubrica's intention. To be simple. To be easy. To be a system that even a beginner can deploy.</p>

<p>The migration was long and arduous. It required dozens of tests and checks. But the result was worth it: a complete deployment became as simple as an <code>npm run build</code>. Convex takes care of the reactive backend, real-time, storage, and server-side Actions. You don't need to configure a database, you don't need to manage connections, you don't need to think about initial scalability.</p>

<p>In terms of performance, Convex sustains latency below 50ms in reading and writing. Supabase, under similar load, reaches 100-200ms. But the decision wasn't based on benchmarks: it was based on alignment with the project's mission. Convex open-sourced in February 2025. With 10,000 stars today, it's an early adopter bet. Rubrica is exactly that kind of bet.</p>

<p>The rest of the stack follows the same pragmatic logic:</p>

<ul>
  <li><strong>Tailwind CSS 4</strong> for styling, with CSS variables and controlled theme</li>
  <li><strong>Wouter</strong> for lightweight routing, without the overhead of React Router for the project's scope</li>
  <li><strong>TanStack Query</strong> for caching and data fetching</li>
  <li><strong>TipTap</strong> for the blog's rich text editor with complete extensions</li>
  <li><strong>React Hook Form + Zod</strong> for forms with typed validation</li>
  <li><strong>jsPDF + signature_pad</strong> for PDF generation and handwritten signature</li>
</ul>

<h2>AI with Purpose: Automatic Translation and ATS Resume Pipeline</h2>

<p>Rubrica has AI. But not in the way you're used to seeing it.</p>

<p>It's not a chatbot. It's not "content generation." It's not AI for the sake of it. It's AI where it makes sense for the professional.</p>

<p>The first case is <strong>automatic translation via OpenRouter</strong>. The problem: maintaining a portfolio in two languages manually is double the work. Each text field, each blog post, each project description would need a version in pt-BR and en-US. Rubrica's solution is to automate this pipeline: you write once in Portuguese and the system automatically translates it into English. Internationalization without the maintenance cost.</p>

<p>This matters more than it seems. Retrofitting i18n into an existing codebase is significantly more expensive than planned from the start. For a solo developer, the entire pipeline, code, design, translation, and review, falls on a single person. AI specifically solves this bottleneck.</p>

<p>The second case is more impactful: the <strong>ATS-optimized resume pipeline</strong>.</p>

<p>Here's the real problem: almost all large companies use Applicant Tracking Systems to process applications. 97.8% of the Fortune 500. 90% of companies with more than a thousand employees. And 66% of these systems can't interpret synonyms: the keywords need to be exact. Candidates are 6 times more likely to get an interview if they personalize their resume for the job.</p>

<p>Rubrica has a three-step pipeline to solve this:</p>

<ol>
  <li><strong>Gemini</strong> analyzes the job description and extracts keywords and ATS requirements</li>
  <li><strong>GPT</strong> scores the current resume against these requirements</li>
  <li><strong>Claude Sonnet</strong> rewrites the resume incorporating the keywords without losing authenticity</li>
</ol>

<p>It's not to trick recruiters. It's to ensure that the recruiter's AI doesn't discard a qualified application just because a keyword was missing. The irony of our market: AI is being used on both sides.</p>

<h2>Quality as a Value: 472 Tests, GDPR, and Enterprise Architecture</h2>

<p>At a certain point, Rubrica stopped being just a portfolio and started having the characteristics of a system in real production. This requires another level of technical responsibility.</p>

<p>The project has <strong>472 tests with Vitest</strong>, covering both the Convex functions and the application layer. Tests aren't about numerical coverage: they're about the confidence you have when making a change. When you refactor something in Rubrica, the tests tell you if you broke something.</p>

<p>It has <strong>soft delete</strong> throughout the system: nothing is permanently deleted, only marked. It has <strong>audit logs</strong> of all administrative actions. It has <strong>rate limiting by IP</strong>, XSS prevention with DOMPurify, and Content Security Policy headers configured.</p>

<p>It has <strong>GDPR compliance</strong>: export of personal data, anonymization, and granular control over what is collected. For a personal site, this may seem like overkill. For a tool that processes contact data, commercial proposals, and customer payments, it's the minimum.</p>

<p>The <strong>plugin system</strong> is where the architecture gets more interesting. There are 14 plugins with hierarchy, each with granular control through the admin panel. Want to disable the payments module? A toggle. Want to enable Telegram notifications for new contacts? Another toggle. The core of the system doesn't change: you activate what you need.</p>

<p>All of this added together: 211 commits, 66 well-structured dependencies, 20 tables in the schema, a technical reference of 880 lines. Rubrica has the scale of a product, but the soul of a personal project.</p>

<h2>Portfolio as Infrastructure: How to Use Rubrica as a Starting Point</h2>

<p>I'll be direct about the current state of the project: Rubrica still has the front-end with my visual identity. Fonts, colors, texts, photos. To use it as a base today, you still need to replace these layers manually.</p>

<p>This is changing. The next planned feature is an <strong>installation CLI</strong> that goes beyond a simple fork:</p>

<ul>
  <li>Different visual templates to choose from before the first deploy</li>
  <li>Configuration of the visual identity in the initial setup</li>
  <li>Guided setup of environment variables and integrations</li>
</ul>

<p>The goal is for you to arrive at Rubrica, run the CLI, answer a few questions about your visual identity, and start with a portfolio that's already yours, not a copy of mine.</p>

<p>But even today, Rubrica works as a base. The README has 10 detailed steps from fork to custom domain. The stack is modern and well-documented. The code follows clear patterns: Repository Pattern, Context API, Clean Architecture in the use cases.</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"Digital commons describe digital systems and solutions that are owned, developed, and maintained by groups rather than by single individuals or entities, coordinating through peer collaboration rather than pricing or subordination."</strong></p>
<p>— Open Future, "Digital Commons as Providers of Public Digital Infrastructures" (2024)</p>
</div>

<p>You don't need to start from scratch. 211 commits of technical decisions, corrected errors, covered edge cases: all of this becomes available when you fork Rubrica. Explore, build on top of it, improve it. The foundation is ours, the identity is yours.</p>

<h2>Conclusion</h2>

<p>The name Rubrica came after a long time thinking. Rubrica: the signature, the sign that you were here, that the work is yours. I didn't need a name to build, but I needed a name to open it up. When the name arrived, the post arrived together.</p>

<p>Rubrica is the project I built with the most care. Seeing a simple portfolio transform into a system with payments, AI, 472 tests, and GDPR compliance over 211 commits taught me more than any tutorial could. It's my pride and joy.</p>

<p>If you're thinking about creating or renovating your portfolio, fork Rubrica. If you want to contribute, open an issue. If you want to build something on top of it, do it. This project was made for that: <strong>sign it with your face.</strong></p>

<h2>References</h2>

<ul class="references">
  <li><a href="https://www.coversentry.com/ats-statistics">CoverSentry. "ATS Statistics 2026: Why Your Resume Disappears Into the Void". Accessed 2026-05-20.</a></li>
  <li><a href="https://www.resumeadapter.com/ats-statistics">ResumeAdapter. "ATS Statistics 2026: The '75% Rejection' Stat Is Fake. Here's Real Data." Accessed 2026-05-20.</a></li>
  <li><a href="https://www.selectsoftwarereviews.com/blog/applicant-tracking-system-statistics">Select Software Reviews. "Applicant Tracking System Statistics (Updated for 2026)". Accessed 2026-05-20.</a></li>
  <li><a href="https://survey.stackoverflow.co/2024/technology">Stack Overflow. "Technology | 2024 Stack Overflow Developer Survey". Accessed 2026-05-20.</a></li>
  <li><a href="https://2024.stateofjs.com/en-US/libraries/build_tools/">State of JS 2024. "Build Tools". Accessed 2026-05-20.</a></li>
  <li><a href="https://openfuture.eu/publication/digital-commons-as-providers-of-public-digital-infrastructures/">Open Future. "Digital Commons as Providers of Public Digital Infrastructures". October 2024. Accessed 2026-05-20.</a></li>
  <li><a href="https://phrase.com/blog/posts/internationalization-beyond-code-a-developers-guide-to-real-world-language-challenges/">Phrase.com. "Internationalization beyond code: A developer's guide to real-world language challenges". Accessed 2026-05-20.</a></li>
  <li><a href="https://www.devtoolsacademy.com/blog/supabase-vs-convex/">DevTools Academy. "Supabase vs Convex". Accessed 2026-05-20.</a></li>
  <li><a href="https://stack.convex.dev/why-choose-convex-database-for-backend">Convex. "Why I picked Convex over Firebase, Supabase, and Neon for my app". Convex Stack Blog. Accessed 2026-05-20.</a></li>
  <li><a href="https://codeanywhere.com/blog/vite-the-java-script-build-tool-dominating-2024">CodeAnywhere. "Vite: The JavaScript Build Tool Dominating 2024". Accessed 2026-05-20.</a></li>
  <li><a href="https://www.tweag.io/blog/2024-12-19-cra-to-vite/">Tweag. "The Developer Experience Upgrade: From Create React App to Vite". Accessed 2026-05-20.</a></li>
</ul>]]></content:encoded>
      <pubDate>Thu, 21 May 2026 02:55:19 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Portfólio</category>
      <category>Open Source</category>
      <category>Projetos Pessoais</category>
      <category>React</category>
      <category>TypeScript</category>
      <category>Vite</category>
      <category>Tailwind</category>
      <category>JavaScript</category>
      <category>Comunidade Dev</category>
      <category>IA no desenvolvimento</category>
      <category>Inteligência Artificial</category>
      <category>Automação</category>
      <category>Mercado de Trabalho</category>
      <category>Vercel</category>
      <category>qualidade de software</category>
      <category>Desenvolvimento Pessoal</category>
      <category>workflow de desenvolvimento</category>
      <category>Privacidade de Dados</category>
      <category>testes   unitários</category>
      <category>clean code</category>
      <category>Free Software</category>
      <category>Convex</category>
      <category>Presença Digital</category>
      <category>Freelancer</category>
      <category>Arquitetura de Software</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/7f942a44-2df0-4067-8377-acc0bc99d393" medium="image"/>
    </item>
    <item>
      <title>My AI Development Workflow</title>
      <link>https://www.mmlo.app/en/blog/meu-fluxo-de-desenvolvimento-com-ia</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/meu-fluxo-de-desenvolvimento-com-ia</guid>
      <description>From PRD to Deployment in Hours</description>
      <content:encoded><![CDATA[<p class="intro">It used to take me weeks, sometimes months, to build a complete system. Today I do the same thing in hours or a few days. Not because I type faster. Because I changed how I think about development. This post isn't about asking ChatGPT for isolated functions. It's about the AI development workflow I use to build systems from scratch, with planning, specialized agents, TDD and deployment, told step by step.</p>

<h2>From Chaos to Orchestration: How My Role Changed</h2>

<p>In the beginning, I missed getting my hands on the code. I often tweaked what the AI generated. Partly because AI back then wasn't as good as it is today. Partly because I felt I had lost control.</p>

<p>Over time, I understood that the problem wasn't the AI. It was the lack of a system. I asked for things at random, with no direction and no validation. The result was code that seemed to work and fell apart at the first integration.</p>

<p>Today I don't write code. I orchestrate. I think about the architecture, define the requirements, create specialized agents, and the AI executes. That doesn't mean I need to know less. I need to know more.</p>

<p>Code was never the most important part. Anyone who has opened an open source project to study it knows that large systems are made of improvised solutions that work. What always mattered was understanding what the code does, not writing it. Now that we no longer write it, understanding code has become even more critical.</p>

<p>To get out of the chaos, I needed a process that repeats. That's what I show next.</p>

<h2>My AI Workflow, Step by Step</h2>

<p>The flow always follows the same order:</p>

<p><strong>Idea → PRD → GitHub Issues → TASKS.md → TDD (tests + execution) → Deploy</strong></p>

<p>Each step has a reason. Skipping one creates problems that only show up later, and fixing things later costs much more than preventing them up front.</p>

<h3>1. Idea: discuss before asking for code</h3>

<p>It all starts with a sentence like "I want to build a complete RBAC authentication system, where do I go from here?". From there it's a back-and-forth with the AI, usually in a web chat. The questions I ask look like these:</p>

<ul>
  <li>"Is this idea feasible with this stack?"</li>
  <li>"Even with significant trade-offs, could we do it with stack X?"</li>
  <li>"Is this necessary, or is it over-engineering?"</li>
  <li>"Aren't we thinking too small?"</li>
  <li>"Aren't we thinking bigger than we should?"</li>
</ul>

<h3>2. PRD: the AI interviews me</h3>

<p>Once the idea is solid at a high level, I flip the game: I ask the AI to ask me a series of questions about the system. These are the questions needed to build the full PRD. They're no longer about the idea, but about execution: what the system will do and what it won't.</p>

<h3>3. Issues and TASKS.md: extreme granularity</h3>

<p>With the PRD ready, I generate the issues on GitHub. From them, I create a <code>TASKS.md</code> file split into stages, with what will be done in each one.</p>

<p>This file sometimes goes past <strong>200 tasks</strong>. And each task is small, along the lines of "now create this function".</p>

<h3>4. Execution: a fresh session for each group of tasks</h3>

<p>For each group of tasks, I open a new AI session with a clean context. Before anything else, I use the tool's planning mode. This reduces hallucinations and makes sure the result is what I want.</p>

<p>I don't carry context over from previous sessions because the functions need to be independent. I plan how that function will work, execute it, and only then start connecting the functions.</p>

<h3>5. TDD: no test, no merge</h3>

<p>TDD is mandatory in all my projects today. For each group of tasks, the test defines the expected behavior before the code exists. If the AI-generated code doesn't pass the test, the code changes, not the test. I explain this cycle in more detail below and in the post about <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">my journey with TDD</a>.</p>

<h3>6. Deploy</h3>

<p>Only after that does the code go to production. If you want to see how code gets from my machine to production, I wrote about it in <a href="https://www.mmlo.com.br/blog/git-flow-para-github-flow">Git Flow to GitHub Flow</a>.</p>

<p>That's the skeleton. Two decisions, though, are what make it work: the order in which I build and the way I set up the agents.</p>

<h2>Backend First, Frontend Later</h2>

<p>This change took me a while to accept. I used to build the frontend first and only then think about the backend, and I let a lot slip through. When you start with the visuals, it's easy to get excited about the interface and forget to check whether the data structure makes sense, whether the business rules are well defined and whether the integrations will work.</p>

<p>Today I always start with the backend: entities, APIs, business rules and tests. Only with that foundation in place do I move to the frontend. For this, I have backend-specific agents, frontend agents and agents that work on both.</p>

<h2>Specialized Agents: the Secret Is in the Limits</h2>

<p>An agent is a prompt. And a good agent prompt, for me, follows two rules:</p>

<ul>
  <li><strong>Markdown is mandatory.</strong></li>
  <li><strong>Well-defined sections</strong>, including what the agent can and cannot do.</li>
</ul>

<p>The second rule is the most important. For an agent to be a specialist, it has to know very clearly what <strong>not</strong> to do. That's what keeps it focused on its job.</p>

<p>A generic agent only has two ways out: either it stays shallow on everything, or it gets so big that it doesn't fit well in an AI session, and then the risk of hallucinations and generic results goes up.</p>

<p>That's why I create separate agents for code review, security and accessibility. Each one has a bounded context, clear rules and knows exactly where its responsibility ends.</p>

<p>Limits on the agent and tests on the code are the two forms of validation in this workflow. And validation is exactly what vibe coding lacks.</p>

<h2>Why Vibe Coding Doesn't Work for Critical Systems</h2>

<p>The problem with vibe coding isn't AI. It's the absence of validation. When you ask for code without a test that says what's correct, you're trusting that the model understood your intent. That rarely happens consistently. I went into how AI behaves without that anchor in the post about <a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">TDD</a>.</p>

<p>Out of curiosity, I've looked at applications built with vibe coding and gained full access to more than one. That's not the exception. It's the consequence of not validating.</p>

<p>For a static site or a personal project, this may not be a problem. When that mindset reaches critical systems, with users and sensitive data, it becomes a security problem.</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"The fundamental difference between those who use AI well and those who use it poorly is knowledge. We need a pragmatic, not romanticized, view of AI's limits and strengths."</strong></p>
</div>

<p>With that view, besides building better systems, you can apply the same workflow with any tool or LLM. People who vibe code become dependent on the AI and the tool, and don't know how to think without them.</p>

<h2>Concept Over Tool</h2>

<p>I specialize in the concept, not in a specific tool or LLM. I can run this workflow anywhere. Today I mostly use Claude Code or OpenCode with Alibaba's plan. Tomorrow it could be another tool. The workflow stays the same.</p>

<p>What changed was how I split my time. I used to plan very little. Today I spend much more time preparing a well-documented environment before executing. That trade is what made the jump from weeks to hours possible.</p>

<p>And the TDD cycle still doesn't feel natural to me. I'm still learning, and I often feel slower than before. The results say otherwise: fewer bugs, less refactoring later and more confidence to change the code.</p>

<h2>Conclusion</h2>

<p>This workflow wasn't born ready. It came from mistakes, from code that fell apart in production and from integrations that failed because I hadn't planned enough. Trial and error took me from chaos to orchestration.</p>

<p>The main lesson is that AI doesn't replace thinking, it amplifies it. If you think in a disorganized way, AI generates disorganized code faster. If you think in a structured way, it executes in a structured way, and even faster.</p>

<p>If you want to start tomorrow, do it in this order:</p>

<ul>
  <li><strong>Plan before asking for a single line of code.</strong> Define what the system does and what it doesn't.</li>
  <li><strong>Break the work into small tasks.</strong> One task, one function.</li>
  <li><strong>Use TDD to validate.</strong> The test comes first, and the code adapts to it.</li>
  <li><strong>Never trust AI output without validation.</strong></li>
</ul>

<p>The difference between vibe coding and professional AI-assisted development isn't the tool. It's the process.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://claude.com/product/claude-code">Anthropic. "Claude Code: AI-powered coding assistant for developers".</a></li>
  <li><a href="https://openai.com/index/gpt-4-research/">OpenAI. "GPT-4 Research".</a></li>
  <li><a href="https://code.claude.com/docs/en/overview">Anthropic. "Claude Code Developer Documentation".</a></li>
  <li><a href="https://github.com/openai/evals">OpenAI. "OpenAI Evals".</a></li>
  <li><a href="https://www.anthropic.com/81k-interviews">Anthropic. "What 81,000 people want from AI".</a></li>
  <li><a href="https://martinfowler.com/bliki/TestDrivenDevelopment.html">Fowler, Martin. "Test Driven Development".</a></li>
  <li><a href="https://tidyfirst.substack.com/p/canon-tdd">Beck, Kent. "Canon TDD".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/minha-jornada-com-tdd">Mierzwa, Matheus. "Minha Jornada com TDD".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/git-flow-para-github-flow">Mierzwa, Matheus. "Git Flow para GitHub Flow".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Sat, 16 May 2026 21:22:02 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>IA no Desenvolvimento</category>
      <category>Inteligência Artificial</category>
      <category>AI code generation</category>
      <category>workflow de desenvolvimento</category>
      <category>vibe coding</category>
      <category>TDD</category>
      <category>Produtividade</category>
      <category>qualidade de software</category>
      <category>Automação</category>
      <category>Comunidade Dev</category>
      <category>Futuro do Trabalho</category>
      <category>Desenvolvimento Pessoal</category>
      <category>Impacto da IA</category>
      <category>Projetos Pessoais</category>
      <category>test driven development</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/844ef664-e8a5-41ac-a7ec-b9c9793fbe23" medium="image"/>
    </item>
    <item>
      <title>My TDD Journey</title>
      <link>https://www.mmlo.app/en/blog/minha-jornada-com-tdd</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/minha-jornada-com-tdd</guid>
      <description>From Chaos to Code That Works the First Time</description>
      <content:encoded><![CDATA[<p class="intro">I've lost count of the nights I spent debugging something a five-minute test would have caught. Every developer knows that feeling, but few admit it. My relationship with tests started like almost everyone's: write the code first, then fit tests in wherever I could. Until I found out the order could be flipped. In this post I explain how I went from a chaotic, reactive workflow to a process where the code works the first time: what made me try TDD, where I got stuck, the moment it clicked, and what I learned applying it day to day, including with AI.</p>

<h2>How I Developed Before TDD</h2>

<p>My workflow followed a pattern I saw everywhere: write all the code, refactor it for readability, and only then try to generate tests. The problem shows up as soon as you think about it. Testing finished code is much harder than it looks, because it goes through so many changes before the final version that mapping every scenario becomes guesswork.</p>

<p>And manual testing kept going, even after I wrote the automated tests. In practice, I had two validation flows running in parallel, and neither gave me confidence.</p>

<h3>AI made the problem more visible</h3>

<p>As generative AI for code became popular, this got even more obvious. The AI generated tests, but in a way that was far too permissive: it <strong>changed the test so it would pass against the existing code</strong>, instead of making the code adapt to the test. That created a false sense of safety, which only showed up when I ran <em>mutation testing</em> and saw mutants surviving all the time.</p>

<p>The cycle was always the same: code working on my machine, tests passing, and production blowing up. And me thinking the problem was me. That frustration is what pushed me to look for another way.</p>

<h2>What Made Me Try TDD (and Where I Got Stuck)</h2>

<p>The turning point came when I watched Akita talking about AI-assisted development. One idea of his stuck with me: <em>vibe coding</em> is bad, but AI with TDD becomes a <em>game changer</em>.</p>

<p>At first I pushed back. I only knew TDD by name. Writing a test before any code existed sounded absurd: how do you test something that doesn't exist?</p>

<p>But my frustration with bugs was bigger than my skepticism. So I started studying. I read about the <strong>Red-Green-Refactor</strong> cycle, looked at examples, and understood the theory. Even so, the first time I tried to apply it, I got stuck again.</p>

<h3>The hardest part: granularity</h3>

<p>What nobody explains clearly is how much scope to put into each TDD cycle. A whole screen? A micro function? A group of related functions?</p>

<p>I researched a lot and found nothing conclusive. The literature talks about "small units" but doesn't say what that means in practice. So I figured it out by testing:</p>

<ul>
  <li><strong>Whole screens:</strong> didn't work. The scope was too big, the test was brittle, and any change broke everything.</li>
  <li><strong>Groups of functions:</strong> better, but I was still testing too many things at once.</li>
  <li><strong>Micro functions:</strong> this is where it clicked. Tests for specific things, like "try to delete a table without being an admin" or "try to list another user's data by using their ID directly". Small, focused, and independent.</li>
</ul>

<p>This isn't necessarily the right answer for everyone. It's the one that worked for me, and it took me a while to get there. With granularity sorted out, I still had to run the cycle across an entire project.</p>

<h2>Red, Green, Refactor: My First Cycles</h2>

<p>My first TDD project was a personal one, built specifically to practice the methodology: an <strong>AI RPG Master</strong> using the Fate system. Here, what matters is what I learned.</p>

<p>The backend ended up with <strong>more than 500 unit tests</strong>, all following the TDD flow: test first, then the function. The test fails, the code passes, I refactor. And repeat.</p>

<h3>Where AI fits into the cycle</h3>

<p>My use of AI in this process isn't vibe coding. I've been using AI pragmatically for more than a year, with planning, a PRD, a deep understanding of the requirements, and specialized agents for code review, security, and accessibility. The AI executes what I think. I'm the one doing the thinking. I described this workflow in detail in <a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">my AI development workflow</a>.</p>

<p>Learning TDD was no different. I built a TDD specialist agent that runs the Red-Green-Refactor cycle automatically. Each function is a new session with a clean context, because the functions need to be independent. I plan exactly how that function will work, run the cycle, and only then start wiring the functions together.</p>

<h3>The cycle still doesn't feel natural</h3>

<p>One important caveat: Red-Green-Refactor still doesn't feel natural to me. I'm still getting used to it, and I often feel slower than before. That perception changes over time, and what changed my mind was one specific test.</p>

<h2>The Moment It All Clicked</h2>

<p>When I finished the AI RPG Master backend, with its 500-plus unit tests, I still hadn't felt the value of it. I hadn't run the application. It was just a pile of tests passing in the terminal.</p>

<p>Then I ran an <em>end-to-end</em> test. <strong>It passed on the first try.</strong></p>

<p>No refactoring afterwards, no finding a flaw here and another there, no adjusting along the way. That had never happened to me. I always found something to fix later, there was always a hidden bug waiting for the worst moment to show up.</p>

<p>That's when I realized it was worth it. And not just because it worked the first time, but because of the confidence it creates: knowing the code is covered, that every scenario was thought through before it existed, that the test is the anchor and the code is what adapts to it.</p>

<p>That's the memory I go back to when the cycle feels slow and the friction shows up. It also explains why, for me, TDD and AI go together.</p>

<h2>TDD with AI: Why Vibe Coding Doesn't Work</h2>

<p>The problem with vibe coding isn't AI. It's the <strong>lack of validation</strong>. When you ask AI for code without tests defining what's correct, you're trusting that the model understood your intent. And that rarely happens consistently.</p>

<p>I've seen AI change the same test over and over until it passed. And I saw the weakness when I ran mutation testing: mutants survived because the code wasn't well built, and the AI adjusted the test to fit the wrong code. When you flip that logic, the picture changes:</p>

<blockquote><p><strong>"Now I have a correct test that defines the expected behavior. If the AI-generated code doesn't pass the test, it's the code that has to change. Not the test."</strong></p></blockquote>

<h3>How I keep AI in check</h3>

<p>For this to work, I have to be strict with the agents. Each agent has a clear list of what it can and cannot do: "don't do this under any circumstances", "you only do this and nothing else". Without that discipline, AI starts generating random code and the TDD cycle loses its purpose.</p>

<p>Out of curiosity, I've looked into applications built with vibe coding and gained full access to more than one. That isn't an exception, it's the consequence of having no validation. A junior developer who understands the logic behind the code is already a step ahead. But it's TDD with AI that separates people who use the tool safely from people who just keep asking it for things.</p>

<p>After months of working this way, a few lessons became clear.</p>

<h2>What I Learned About TDD in Practice</h2>

<p>I've been using TDD for months on new projects, both personal and for clients. These are the four things that changed my view the most.</p>

<h3>The code gets leaner</h3>

<p>This surprised me. The <em>green</em> phase, where you write the minimum code needed for the test to pass, naturally prevents <em>over-engineering</em>. Since there's already a test stating what I expect, I write just enough for that scenario. The result is smaller, more direct functions with high quality. Code that does exactly what it needs to, no more and no less.</p>

<h3>Feeling slower at first is normal</h3>

<p>At first, TDD feels slower. That's common: you're learning a new methodology, breaking old habits, and thinking differently.</p>

<p>My advice is to document the process. Write down how many refactors you used to do, how long it took to implement a feature, and how many bugs showed up later. Then compare with your numbers using TDD. With data in hand, it becomes clear that you ended up with less work, even if it initially felt like more effort.</p>

<p>Industry studies point the same way. In a <a href="https://doi.org/10.1007/s10664-008-9062-z">study of three teams at Microsoft and one at IBM</a>, researchers from Microsoft Research, IBM, and NC State saw pre-release defect density drop by <strong>40% to 90%</strong> compared with similar projects that didn't use TDD. In exchange, the teams reported a <strong>15% to 35%</strong> increase in initial development time. In my experience, that time comes back with interest in the debugging you no longer have to do.</p>

<h3>TDD isn't for everyone, and that's fine</h3>

<p>This is an unpopular opinion, but I believe it: TDD shouldn't be mandatory. It's a personal choice.</p>

<p>There are several methodologies: TDD, <em>Spec Driven Development</em>, and others. Each person should use the one that works best for them, as long as the result meets the team's standards. Forcing someone to follow a methodology limits their potential.</p>

<p>What I do advocate is that everyone try different methodologies. I came from Spec Driven Development, and it worked well, but I felt stuck. When I tried TDD, I realized it was the best fit for the way I work. The only way to know what works for you is to experiment.</p>

<h3>In legacy systems, little by little</h3>

<p>I get this question a lot. The answer is: gradually. I didn't set out to apply TDD to the company's legacy systems. But when a small task comes in that fits well, I apply TDD to it and make sure that specific part works.</p>

<p>Over time, that builds confidence to change existing code, because you know the tests protect the expected behavior. That's what makes me consider moving from the manual approach to TDD more broadly.</p>

<h2>Conclusion</h2>

<p>My TDD journey wasn't linear. I started skeptical, got stuck on granularity, felt slower, and then watched an end-to-end test pass on the first try. What TDD gave me wasn't just code that works. It was confidence to refactor, safety to change things without fear, and leaner code, because the methodology itself keeps me from building more than I need.</p>

<p>If you want to start, the path is simple:</p>

<ul>
  <li><strong>Pick a small project</strong>, preferably a personal one.</li>
  <li><strong>Write a single test before any code</strong> and watch it fail.</li>
  <li><strong>Write the minimum code to make it pass</strong>, refactor, and repeat.</li>
  <li><strong>Document the process</strong> and compare it with your previous workflow.</li>
</ul>

<p>Don't expect it to feel natural or fast at first. Compare the numbers and see for yourself whether it makes a difference.</p>

<h2>References</h2>
<ul class="references">
  <li><a href="https://martinfowler.com/bliki/TestDrivenDevelopment.html">Fowler, Martin. "Test Driven Development". martinfowler.com, December 2023.</a></li>
  <li><a href="https://tidyfirst.substack.com/p/canon-tdd">Beck, Kent. "Canon TDD". tidyfirst.substack.com, December 2023.</a></li>
  <li><a href="https://blog.cleancoder.com/uncle-bob/2014/12/17/TheCyclesOfTDD.html">Martin, Robert C. "The Cycles of TDD". blog.cleancoder.com, December 2014.</a></li>
  <li><a href="https://doi.org/10.1007/s10664-008-9062-z">Nagappan, Nachiappan; Maximilien, E. Michael; Bhat, Thirumalesh; Williams, Laurie. "Realizing quality improvement through test driven development: results and experiences of four industrial teams". Empirical Software Engineering, 13(3), 2008.</a></li>
  <li><a href="https://en.wikipedia.org/wiki/Test-driven_development">Wikipedia. "Test-driven development".</a></li>
  <li><a href="https://en.wikipedia.org/wiki/Mutation_testing">Wikipedia. "Mutation testing".</a></li>
  <li><a href="https://www.mmlo.com.br/blog/meu-fluxo-de-desenvolvimento-com-ia">Mierzwa, Matheus. "Meu Fluxo de Desenvolvimento com IA".</a></li>
</ul>
]]></content:encoded>
      <pubDate>Wed, 06 May 2026 14:50:29 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>TDD</category>
      <category>test driven development</category>
      <category>red green refactor</category>
      <category>testes unitários</category>
      <category>qualidade de software</category>
      <category>IA no desenvolvimento</category>
      <category>vibe coding</category>
      <category>mutation testing</category>
      <category>clean code</category>
      <category>workflow de desenvolvimento</category>
      <category>AI code generation</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/41dd347b-f519-47c9-ace9-ceb7e0ce3704" medium="image"/>
    </item>
    <item>
      <title>Free Software vs. Open Source</title>
      <link>https://www.mmlo.app/en/blog/free-software-vs-open-source</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/free-software-vs-open-source</guid>
      <description>The History of the Split in the Movement</description>
      <content:encoded><![CDATA[<p class="intro">Have you ever wondered why some say "free software" and others say "open source," when both seem to describe the same type of publicly available code? This isn't just a difference in words. It's a political divide that has shaped the technology we use today. This post reveals how a digital emancipation movement was co-opted by corporate interests and what that means for you, the developer, now.</p>

<h2>The Political Origins of Free Software (1983-1985)</h2>

<h3>The Collapse of the Sharing Community</h3>

<p>Richard Stallman wasn't just a talented programmer at MIT. He was a <strong>leader of a political emancipation movement</strong>. When he founded the GNU Project in 1983 and the Free Software Foundation in 1985, he wasn't making a technical choice. He was responding politically to the enclosure of software that had previously been common in academia.</p>

<p>Stallman vividly describes the collapse of the MIT AI Lab's software sharing community. Until the early 1980s, programmers shared code freely. It was as natural a practice as sharing cooking recipes. But when Digital discontinued the PDP-10 series and the company Symbolics hired almost all the lab's hackers, this community crumbled.</p>

<p>Modern computers of the time, such as VAX or 68020, came with proprietary operating systems. You needed to sign a non-disclosure agreement (NDA) just to receive an executable copy. As Stallman recounts in his founding text:</p>

<div style="border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"That meant that the first step in using a computer was to promise not to help your neighbor. A cooperative community was forbidden."</strong></p>
</div>

<h3>Stallman's Moral Choice</h3>

<p>This wasn't a technical restriction. It was a <strong>political barrier</strong>. Proprietary software didn't arise naturally. It was imposed through legal and economic mechanisms that transformed common knowledge into private property.</p>

<p>Stallman then faced a moral choice. He could have joined the world of proprietary software, signing NDAs and developing closed-source software. He would have made money. But as he himself wrote, at the end of his career, he would look back on "years of building walls to divide people."</p>

<p>Instead, he chose to develop a free operating system. GNU stands for "GNU's Not Unix," a hacker joke that became a political project. The language of <strong>freedom</strong> (free as in freedom, not free as in free beer) has always been at the heart of the movement's narrative.</p>

<h3>The Four Freedoms as Principles of Emancipation</h3>

<p>The four freedoms of free software are not technical specifications. They are <strong>principles of digital emancipation</strong>:</p>

<ul>
  <li>Freedom 0: To run the program as you wish, for any purpose</li>
  <li>Freedom 1: To study how the program works and modify it (requires access to the source code)</li>
  <li>Freedom 2: To redistribute copies, free of charge or for a fee</li>
  <li>Freedom 3: To distribute modified versions, benefiting the community</li>
</ul>

<p>As the FSF explicitly states: <strong>"Using free software is making a political and ethical choice, asserting the right to learn and share what we learn with others."</strong></p>

<p>This perspective was clearly anti-capitalist. Stallman saw proprietary software as a social problem of power and control, not of efficiency. The Free Software Foundation was created in 1985 as a non-profit organization to advocate for and educate about these principles.</p>

<div style=" border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"If I am not for myself, who will be? If I am only for myself, what am I? If not now, when?"</strong></p>
<p>— Hillel (quoted by Richard Stallman in "The GNU Project")</p>
</div>

<p>Solid references confirm this political origin. Stallman himself wrote in "The GNU Project" that the decision was based on a spirit similar to that of Hillel.</p>

<h2>The Four Freedoms: Ethical, Not Technical</h2>

<p>It's crucial to understand that the four freedoms define free software as a <strong>matter of liberty, not price</strong>. This distinction isn't semantic. It's political.</p>

<h3>Freedom vs. Price: A Political Distinction</h3>

<p>Many misinterpret the term "free software" as thinking of free-of-charge software. But Stallman and the FSF have always emphasized: think of "free speech," not "free beer." The ambiguity of the word "free" in English has led to misunderstandings, but the FSF has never backed down from this terminology because it carries the correct political meaning.</p>

<h3>Copyleft as a Guarantee of Freedoms</h3>

<p>The four freedoms are <strong>inalienable</strong>. The mechanism that guarantees them is copyleft, implemented through the GNU General Public License (GPL). Copyleft uses copyright law, but inverts it to serve the opposite purpose: instead of restricting a program, it keeps the program free.</p>

<p>The central idea of copyleft is that everyone is granted permission to run, copy, modify, and distribute the program, but <strong>they are not allowed to add restrictions</strong>. Thus, the crucial freedoms are guaranteed to everyone who has a copy.</p>

<p>For copyleft to be effective, modified versions must also be free. This ensures that work based on the original is available to the community if it's published. When salaried programmers improve GNU software, copyleft prevents their employers from saying, "You can't share these changes because we're going to use them to make our proprietary version."</p>

<p>Stallman explains this need:</p>

<div style=" border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"The requirement that changes must be free is essential if we want to ensure freedom for every user of the program."</strong></p>
<p>— Richard Stallman</p>
</div>

<p>The X Window System illustrates the problem that copyleft solves. Developed at MIT and released as free software with a permissive license, it was quickly adopted by companies that added it to their proprietary Unix systems, in binary form only, covered by the same non-disclosure agreement.</p>

<p>These copies of X were no more free software than Unix. The X developers didn't see this as a problem. Their goal wasn't freedom, it was "success," defined as "having many users." It didn't matter if those users had freedom.</p>

<p>This led to a paradoxical situation: if you judged based on the freedom provided by the distribution terms of the MIT release, X was free software. But if you measured the freedom of the average X user, you would have to say that it was proprietary software.</p>

<p>The GPL solves this by requiring that anything added to or combined with software under the GPL must also be free and under the GPL. It's a "hole big enough to sink a ship" that copyleft closes.</p>

<p>To understand more about the differences between licenses, see our post <a href="https://www.mmlo.com.br/blog/gpl-vs-mit">GPL vs MIT: Which License to Choose for Your Project</a>.</p>

<p>Today, the FSF maintains that free software is an ethical imperative. As they state in their official definition: <strong>"Free software is a matter of liberty, not price. The criterion for a free manual is basically the same as for free software: it is a matter of giving all users certain freedoms."</strong></p>

<h2>The Netscape Case (1998): Victory or Co-option?</h2>

<p>In January 1998, Netscape announced that it would make the Navigator code free and open source. This moment is often celebrated as a victory for free software. But was it really?</p>

<p>Netscape's motivation was <strong>commercial</strong>, not political. It was a defensive strategy against Microsoft, which dominated the browser market with Internet Explorer. Netscape saw the open source model as a way to compete, not an ideological conversion to the principles of Free Software.</p>

<p>This was the <strong>beginning of the corporate co-option</strong> of the movement. When companies realized that "open source" could be profitable, the discourse changed. There was no more talk of freedom, ethics, or emancipation. There was talk of "competitive advantages," "code quality," and "development efficiency."</p>

<p>Opening Netscape's code wasn't an act of generosity. It was a market maneuver. And that matters because it reveals the logic behind corporations' adoption of the open source model.</p>

<p>It's important to note that the creation of the term "Open Source" by Eric Raymond and Bruce Perens wasn't an <strong>intentional betrayal</strong> of Stallman's principles. It was a "necessity of the time" that, unfortunately, compromised the original principles of Free Software.</p>

<p>The material context explains this decision. In 1998, there were few developers working on free software. Projects needed continuous maintenance. The financial and intellectual contribution of companies with commercial and private interests was necessary at the time.</p>

<p>Raymond and Perens saw an opportunity: if they removed the rhetoric of "freedom" and "ethics," which scared executives, and focused on practical arguments, they could attract many more companies to contribute to open source software.</p>

<p>It worked. But at what cost?</p>

<p>Netscape wasn't advocating for digital emancipation. It was defending its market position. And when companies adopted "open source," they didn't adopt the political principles of Free Software. They adopted a development methodology that could increase their profits.</p>

<h2>The Creation of the Term "Open Source"</h2>

<h3>The Marketing Strategy for Silicon Valley</h3>

<p>In 1998, Eric Raymond, Bruce Perens, and others founded the Open Source Initiative (OSI) and coined the term "Open Source." The goal was explicit: to make free software more palatable to companies.</p>

<p>Raymond explicitly said that they needed to abandon the rhetoric of "freedom" so as not to scare Silicon Valley. The strategy was to focus on practical benefits: better quality software, more reliable, developed more efficiently through the "bazaar" model (many eyes, all bugs are shallow).</p>

<h3>The Open Source Definition: Technical Criteria, Not Political</h3>

<p>The Open Source Definition was derived from the Debian Free Software Guidelines, but with a crucial difference: it focused on the terms of the <strong>source code license</strong>, not on the <strong>user's freedoms</strong>.</p>

<p>The ten criteria of the OSD include free redistribution, availability of source code, permission for derivative works, non-discrimination against persons or fields of endeavor, and technology neutrality. These are technical criteria, not political.</p>

<p>Stallman criticizes this approach in "Why Open Source Misses the Point of Free Software":</p>

<div style=" border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>"Some open source supporters considered the term a 'marketing campaign for free software', which would appeal to business executives by highlighting practical benefits of the software, without raising issues of right and wrong that they might not like to hear."</strong></p>
</div>

<h3>The Political Emptying of Language</h3>

<p>The term "open source" quickly became associated with ideas and arguments based only on practical values: making powerful and reliable software. Most discussions about "open source" pay no attention to right and wrong, only to popularity and success.</p>

<p>This linguistic shift wasn't neutral. It was strategic. By removing the language of freedom, they also removed the politics. Software became a tool, not a movement.</p>

<p>Today, when someone says "I'm an open source developer" but never mentions freedom, that person is <strong>perpetuating the political emptying</strong> of the movement. It may not be intentional, but it's the effect of the linguistic choice.</p>

<h2>Inevitable Bifurcation? Material Conditions and Corporate Contribution</h2>

<p>The bifurcation between Free Software and Open Source was <strong>inevitable</strong> due to the material conditions of the time. It wasn't an abstract ideological choice. It was a response to the concrete conditions of software development in the 1990s.</p>

<p>There were far fewer developers then. Free software projects needed continuous maintenance. Without corporate contribution, many projects would die. Companies with commercial and private interests brought financial and intellectual resources necessary for the moment.</p>

<p>This is a <strong>materialist</strong> analysis, not a conspiratorial one. It's not about demonizing Raymond, Perens, or Netscape executives. It's about understanding how material conditions led to certain decisions.</p>

<p>Open Source had an ambiguous role. On the one hand, it allowed the growth of projects and the release of open source versions of private projects. Chromium is a clear example: open source code maintained by Google, used by millions, the basis of Chrome and other browsers.</p>

<p>But this release wasn't out of "goodness of heart." It was due to pressure from the community and the market. Google open sourced Chromium to compete with Firefox and Internet Explorer, to establish web standards that favored its services, to collect usage data through Chrome.</p>

<p>There's nothing wrong with companies using open source for their commercial interests. The problem is when we present this as an act of generosity or as "contribution to the community."</p>

<p><strong>Companies have a profit motive.</strong> Expecting ethical behavior from them out of goodwill is naive. They act when it's in their financial interest. Sometimes this coincides with the interests of the free software community. Sometimes it doesn't.</p>

<p>The bifurcation was, therefore, a consequence of material conditions. It wasn't an individual betrayal. It was a structural adaptation that had political costs.</p>

<p>Stallman recognizes this complexity. He doesn't demonize individuals, but criticizes the philosophy:</p>

<p><strong>"We in the free software movement don't think of the open source field as an enemy. The enemy is proprietary (nonfree) software. But we want people to know that we stand for freedom, so we don't accept being labeled as open source supporters."</strong></p>

<h2>Current Contradictions: AI, Cloud, and GitHub</h2>

<p>The contradictions of 1998 haven't disappeared. They have <strong>intensified</strong> over time. Today we face problems that Stallman and his contemporaries couldn't have fully imagined.</p>

<h3>Artificial Intelligence: Value Extraction on an Industrial Scale</h3>

<p>GitHub Copilot and other AI models were trained on billions of lines of open source code without compensation to the original authors. The code that developers released under open source licenses is being used to train proprietary models that may replace their jobs.</p>

<p>This is a form of <strong>value extraction</strong>. AI companies use the digital commons (open source code) to create proprietary products that generate private profit. It's the enclosure of the digital commons in its most advanced form.</p>

<h3>Cloud Computing: Monetization without Retribution</h3>

<p>AWS and other cloud providers offer services based on open source software (Elasticsearch, Redis, MongoDB) without adequately giving back to the communities that maintain these projects. Companies use open source to build proprietary SaaS services.</p>

<p>This has led to a reaction: licenses like MongoDB's SSPL (Server Side Public License) and the BSL (Business Source License) try to prevent cloud providers from offering open source software as a service without contributing back. But these licenses aren't considered "open source" by the OSI.</p>

<h3>GitHub: Corporate Centralization of Development</h3>

<p>Microsoft, a historically proprietary software company, bought GitHub in 2018 for $7.5 billion. Today, Microsoft is one of the largest contributors on GitHub. Ironic? Perhaps. But it also reveals how open source has been fully integrated into the corporate business model.</p>

<p>GitHub has centralized open source development. Millions of developers use a proprietary platform to host open source code. Microsoft controls the infrastructure, the usage data, the relationships between developers.</p>

<p>A beginner developer today needs to understand this history if they want to identify with software as a <strong>political movement</strong>, not just as a commodity. Otherwise, they will be contributing to their own exploitation without realizing it.</p>

<p>But this isn't about forced political agitation. If this subject is important to you, you can use this foundation for your studies. You can learn by contributing to Free Software. You can review commits from Free Software projects to understand the technical and political decisions behind the code.</p>

<p>The CTA isn't "revolt now." It's <strong>study, understand, participate consciously</strong>. If you value freedom, act to maintain and defend it.</p>

<h2>Language as a Battleground</h2>

<p>The choice of words isn't neutral. "Free Software" emphasizes freedom (ethical, political). "Open Source" emphasizes methodology (practical, technical). This linguistic shift was strategic, and its effects persist today.</p>

<p>The change in language was <strong>both things</strong>: political betrayal AND necessary translation for the context of the 90s. This is a real contradiction, not a Manichean simplification.</p>

<p>Stallman argues that different words convey different ideas:</p>

<p><strong>"'Free software.' 'Open source.' If it's the same software (or nearly so), does it matter which name you use? Yes, because different words convey different ideas."</strong></p>

<p>He recognizes that the term "free software" is prone to misunderstandings (free as in liberty vs. free as in gratis), but argues that the solution is to educate, not abandon the term.</p>

<p>The adaptation of language may be necessary, but it must be a <strong>consensus of the Free Software community</strong>, not a corporate imposition. When Raymond and Perens created "open source," they didn't consult the community. They decided that the rhetoric of freedom was an obstacle and removed it.</p>

<p>Today, linguistic neutrality (terms like "FLOSS" or "FOSS") doesn't help. If you want to defend freedom, using a neutral term isn't the way. Defending freedom implies showing people your support for freedom.</p>

<p>Stallman concludes:</p>

<p><strong>"Each time you say 'free software' instead of 'open source', you help our cause."</strong></p>

<p>This doesn't mean that defenders of Free Software and Open Source can't work together on practical projects. They can and should. But they shouldn't confuse the philosophies.</p>

<h2>Conclusion</h2>

<p>Recognizing the roots and unfolding of this division isn't an exercise in nostalgia. It's a condition for understanding the tensions that define software development in the present.</p>

<p>What began as a digital autonomy project led by Stallman in the 1980s found an inflection point in 1998: the adoption of the term "Open Source" by the community. This transition responded to concrete needs for project sustainability, even if it diluted the original discourse of freedom.</p>

<p>Thirty years later, the dilemmas are worsening. AI models feed on public repositories without compensation. Cloud giants monetize community projects without proportional return. The main code hosting platform belongs to a corporation whose core has always been closed-source software.</p>

<p>For those now entering the field, understanding this trajectory is to distinguish between participating in a movement for technological autonomy or just producing digital commodities. The recent past illuminates the present: what seemed overcome has returned on an expanded scale.</p>

<p>How to proceed? Study the licenses before contributing. Examine commits from projects that truly preserve freedom. Question who benefits from your work. There's no pamphlet here: there's an invitation to grounded reflection and informed choice.</p>

<div style=" border-left: 4px solid #f5f5f5; padding: 15px; margin: 20px 0;">
<p><strong>Every line you produce carries a political position, even when it claims neutrality. Writing code is building the world in which others will live. What world are you helping to create?</strong></p>
</div>

<h2>References</h2>

<ul class="references">
  <li><a href="https://www.gnu.org/gnu/thegnuproject.html">Stallman, Richard. "The GNU Project". GNU Project, 1998. Accessed 2026-04-08.</a></li>
  <li><a href="https://www.fsf.org/about/what-is-free-software/">Free Software Foundation. "What is Free Software and Why is it so Important for Society?". Accessed 2026-04-08.</a></li>
  <li><a href="https://www.gnu.org/philosophy/free-sw.html">Free Software Foundation. "The Free Software Definition". Accessed 2026-04-08.</a></li>
  <li><a href="https://www.gnu.org/philosophy/open-source-misses-the-point.html">Stallman, Richard. "Why Open Source Misses the Point of Free Software". GNU Project. Accessed 2026-04-08.</a></li>
  <li><a href="https://opensource.org/osd">Open Source Initiative. "The Open Source Definition". Accessed 2026-04-08.</a></li>
  <li><a href="https://www.chromium.org/">The Chromium Project. Accessed 2026-04-08.</a></li>
  <li><a href="https://www.gnu.org/philosophy/free-software-even-more-important.html">Stallman, Richard. "Free Software Is Even More Important Now". GNU Project. Accessed 2026-04-08.</a></li>
</ul>

---

"]]></content:encoded>
      <pubDate>Wed, 08 Apr 2026 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Free Software</category>
      <category>OpenSource</category>
      <category>Tecnologia e Sociedade</category>
      <category>Soberania Digital</category>
      <category>Ética na Tecnologia</category>
      <category>Forças Produtivas</category>
      <category>Direitos Digitais</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/7b7c1804-f569-4da8-bf5e-a23fffdd3000" medium="image"/>
    </item>
    <item>
      <title>Cooperative Platforms and the Fediverse</title>
      <link>https://www.mmlo.app/en/blog/plataformas-cooperativas-e-fediverso</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/plataformas-cooperativas-e-fediverso</guid>
      <description>Real Alternatives to Big Tech</description>
      <content:encoded><![CDATA[<p class="intro">What if it were possible to build social networks that don't turn us into products? Where algorithms don't decide what we see, ads don't exploit our attention, and control remains in the hands of users? This isn't some distant utopia. It's the reality of the Fediverse and cooperative platforms that already exist today.</p>

<p>For us front-end developers who build interfaces every day, this question resonates with particular urgency. We are architects of cyberspace, but we often build on land that doesn't belong to us, using tools that alienate us from the product of our labor. This post is an invitation to praxis: to learn about, participate in, and build concrete alternatives to platform capitalism.</p>

<h2>What Are Cooperative Platforms?</h2>

<p>Cooperative platforms are the digital embodiment of traditional cooperative principles. Unlike capitalist platforms like Uber, Airbnb, or YouTube, where a corporation extracts value from the labor and data of users, cooperative platforms are <strong>collectively owned</strong> and <strong>democratically governed</strong> by those who use them.</p>

<p>Trebor Scholz, the theorist who coined the term "platform cooperativism," defines these platforms as "shared ownership and democratic governance of digital services." It's not just about free software, but about a <strong>reorganization of digital production relations</strong>.</p>

<h3>Fundamental Principles</h3>

<p>A cooperative platform is distinguished by:</p>

<ul>
  <li><strong>Collective ownership</strong>: The digital means of production belong to the user-workers, not shareholders</li>
  <li><strong>Democratic governance</strong>: Decisions are made collectively, not by a corporate board</li>
  <li><strong>Fair distribution of value</strong>: The surplus generated is reinvested or distributed among members, not extracted as profit</li>
  <li><strong>Social purpose</strong>: The goal is to serve the community, not maximize financial return</li>
  <li><strong>Free software</strong>: The code is open, auditable, and modifiable by the community</li>
</ul>

<p>These principles are not accidental. They are a <strong>political response</strong> to the capitalist appropriation of the digital commons. When a Big Tech company monetizes your data, your interactions, and your work, it's practicing a form of digital surplus value, extracting value that you produce without adequate compensation.</p>

<h2>Fediverse: The Federated Internet in Practice</h2>

<p>The <strong>Fediverse</strong> (a portmanteau of "federation" and "universe") is the ecosystem that puts these principles into practice. It's a network of independent servers that communicate with each other through open protocols, forming a <strong>decentralized universe of interconnected platforms</strong>.</p>

<p>The clearest analogy is email: you can have an account on Gmail and someone else on Yahoo, but both can exchange messages because they use compatible protocols. The Fediverse applies this same principle to social networks, videos, photos, blogs, and more.</p>

<h3>How Federation Works</h3>

<p>Each server in the Fediverse, called an instance, is independent. Whoever creates an instance defines its own moderation rules, its community, its governance. But through protocols like <strong>ActivityPub</strong> (standardized by the W3C in 2016), these instances can communicate.</p>

<p>This means you can have an account on Mastodon on a Brazilian server and follow someone on a Japanese server, or watch videos from a French PeerTube instance. All from the same interface, all interoperable.</p>

<p>According to 2018 data, the Fediverse already had <strong>3,952 instances</strong> and more than <strong>2.3 million identities</strong>. Today, these numbers are significantly higher, especially after the mass migration of Twitter users to Mastodon in 2022-2023.</p>

<h2>ActivityPub: The Protocol That Unites the Fediverse</h2>

<p>ActivityPub is the <strong>communication protocol</strong> that makes this federation possible. Developed and standardized by the W3C (World Wide Web Consortium), it defines how different platforms can exchange information in a standardized way.</p>

<p>When you publish a post on Mastodon, that post is formatted as an "Activity" of ActivityPub. Servers that understand this protocol can receive, interpret, and display that post to their users, regardless of which software they are running.</p>

<h3>Why Open Protocols Matter</h3>

<p>The existence of open protocols like ActivityPub is a <strong>matter of autonomy</strong>. When a platform is proprietary and closed, like Twitter or Instagram, you are trapped in it. There is no interoperability, no real portability, no alternative but to accept the corporation's terms.</p>

<p>With open protocols, the power is reversed: <strong>you choose your instance</strong>, you can migrate if you disagree with the rules, and you maintain your ability to communicate with the rest of the Fediverse. It's the difference between being a user (someone who uses what is given to them) and a <strong>participant</strong> (someone who has agency over the platform).</p>

<h2>Mastodon, PeerTube, and Other Concrete Alternatives</h2>

<p>The Fediverse is not abstract theory. They are real, functional platforms used by millions. Let's get to know the main ones:</p>

<h3>Mastodon: Federated Microblogging</h3>

<p><strong>Mastodon</strong> is the federated alternative to Twitter. Developed as free software, it is maintained by a non-profit organization and supported by donations. There are no ads, no engagement algorithms, no data sales.</p>

<p>Your timeline is <strong>chronological</strong>, not manipulated to maximize screen time. Moderation is <strong>local and community-based</strong>, not imposed from top to bottom. And you can follow anyone from any instance, creating your own network without corporate intermediaries.</p>

<p>As the developers themselves say: "Social networking that's not for sale." Your timeline should be filled with what matters to you, not what a corporation decides you should see.</p>

<h3>PeerTube: Decentralized Video</h3>

<p><strong>PeerTube</strong> is the alternative to YouTube developed by Framasoft, a French NGO that advocates for free software. Unlike YouTube, PeerTube is not a single platform. It is <strong>software that you install on your own server</strong>, creating your own "YouTube instance."</p>

<p>Currently, there are more than <strong>1,600 PeerTube instances</strong> with more than <strong>1 million videos</strong>. Each instance has its rules, its moderation, its focus (some are general, others are specific to education, art, activism, etc.).</p>

<p>An important technical innovation: PeerTube uses <strong>peer-to-peer (WebRTC)</strong> for video transmission. When a video goes viral, the viewers themselves help distribute the content, reducing the load on the server. It's technology at the service of cooperation, not extraction.</p>

<p>As Framasoft states: "PeerTube considers you as a person and not as a product to be profiled and trapped in video loops."</p>

<h3>Other Fediverse Platforms</h3>

<p>In addition to Mastodon and PeerTube, the Fediverse includes:</p>

<ul>
  <li><strong>Pixelfed</strong>: Instagram alternative (photo sharing)</li>
  <li><strong>WriteFreely</strong>: Federated blogging platform</li>
  <li><strong>Lemmy</strong>: News aggregator and forums (Reddit alternative)</li>
  <li><strong>Friendica</strong>: Complete social network with multiple functionalities</li>
  <li><strong>Hubzilla</strong>: Social CMS with advanced federation</li>
</ul>

<p>Each of these platforms is free software, federated via ActivityPub, and can be self-hosted or accessed through public instances.</p>

<h2>Challenges of the Cooperative Model</h2>

<p>It would be naive to romanticize the Fediverse without recognizing its challenges. Building alternatives to platform capitalism is not simple, and there are real obstacles that explain why these alternatives are still niche.</p>

<h3>UX and Discovery</h3>

<p>Centralized platforms like Twitter and YouTube invest billions in UX (user experience). The Fediverse, developed by communities and NGOs, cannot always compete in this aspect. Content discovery can be less intuitive, and the learning curve is steeper.</p>

<p>Furthermore, <strong>federation requires choices</strong>: which instance to choose? Which other instances to trust? This complexity is the price of autonomy, but it can also be a barrier for non-technical users.</p>

<h3>Sustainable Funding</h3>

<p>Mastodon and PeerTube are supported by <strong>donations</strong>, not profit. This guarantees independence, but also means limited resources. Mastodon, for example, has a small team for millions of users. Development is slower, and infrastructure depends on the generosity of the community.</p>

<p>This is a structural challenge: how to fund platforms that, in principle, refuse the predatory monetization models of Big Tech (ads, data sales, extraction of surplus value)?</p>

<h3>Network Effect</h3>

<p>Social networks are valuable because <strong>everyone is there</strong>. This "network effect" is a powerful barrier: even if Mastodon is technically superior to Twitter, if your friends and colleagues are on Twitter, there is a real cost of migration.</p>

<p>Big Tech companies exploit this effect in a predatory way: they know that you can't simply "leave" because your social network, your contacts, your history are trapped on the platform. It's a form of <strong>digital lock-in</strong>. And the Fediverse, by allowing portability and interoperability, is a direct response to this problem.</p>

<h2>How to Participate: A Guide for Front-End Developers</h2>

<p>We have arrived at praxis. As front-end developers, we have specific skills that can contribute to the Fediverse. More than that: we have a <strong>political duty</strong> to use our work to build alternatives.</p>

<h3>As a User: Migrate Your Accounts</h3>

<p>The first step is simple: <strong>participate</strong>.</p>

<ul>
  <li><strong>Create a Mastodon account</strong>: Visit <a href="https://joinmastodon.org/servers">joinmastodon.org/servers</a> and choose an instance. For Brazilian instances: <code>mastodon.social</code> (international), <code>mastodon.place</code>, or search for instances in Portuguese.</li>
  <li><strong>Test PeerTube</strong>: Explore videos at <a href="https://joinpeertube.org">joinpeertube.org</a> and consider following creators who publish there.</li>
  <li><strong>Try Pixelfed</strong>: If you use Instagram, test federated photo alternatives.</li>
</ul>

<h3>As a Contributor: Participate in Development</h3>

<p>All of these projects are <strong>free software</strong> and need contributors:</p>

<ul>
  <li><strong>Mastodon</strong>: Code at <a href="https://github.com/mastodon/mastodon">github.com/mastodon/mastodon</a>. Needs contributions in React, Ruby on Rails, accessibility, internationalization.</li>
  <li><strong>PeerTube</strong>: Code at <a href="https://github.com/Chocobozzz/PeerTube">github.com/Chocobozzz/PeerTube</a>. Developed in Angular/Node.js, needs front-end, video player, UX.</li>
  <li><strong>WriteFreely</strong>: <a href="https://github.com/writefreely/writefreely">github.com/writefreely/writefreely</a>. Go and simple templates, great for those who want to contribute to federated blogging.</li>
</ul>

<h3>As a Self-Host: Create Your Instance</h3>

<p>If you want total control, you can host your own instance:</p>

<ul>
  <li><strong>Mastodon</strong>: Requires a server with Docker, PostgreSQL configuration, Redis, and your own domain. Documentation at <a href="https://docs.joinmastodon.org">docs.joinmastodon.org</a>.</li>
  <li><strong>PeerTube</strong>: Requires a server with more resources (video is heavy), but the installation is well documented at <a href="https://docs.joinpeertube.org">docs.joinpeertube.org</a>.</li>
</ul>

<p>For a small instance (up to 100-200 users), a VPS of $10-20/month is sufficient. Join other people and split costs. <strong>Cooperativism starts with infrastructure</strong>.</p>

<h3>As an Activist: Educate and Organize</h3>

<p>Finally, use your voice:</p>

<ul>
  <li><strong>Educate colleagues</strong>: Explain the Fediverse to other developers. Organize talks, write posts, make threads on Mastodon.</li>
  <li><strong>Advocate for institutional migration</strong>: Your company, university, or organization can host its own instance. It's a matter of <strong>digital sovereignty</strong>.</li>
  <li><strong>Build alternatives</strong>: Identify gaps in the Fediverse and build new platforms. The protocol is open. What's missing is political imagination.</li>
</ul>

<h2>Conclusion: Code Is Political</h2>

<p>Cooperative platforms and the Fediverse are not just "technical alternatives" to Big Tech. They are <strong>political projects</strong> that embody a different vision of how technology can be organized.</p>

<p>When we choose to use Mastodon instead of Twitter, we are saying that <strong>we refuse to be products</strong>. When we host our own instance, we are stating that <strong>control should not be centralized</strong>. When we contribute to free software, we are building <strong>common digital means of production</strong>.</p>

<p>As front-end developers, we have a special responsibility. The interfaces we build shape how people experience the digital world. We can build interfaces that exploit, that addict, that extract. Or we can build interfaces that <strong>emancipate, that educate, that liberate</strong>.</p>

<p>The Fediverse already exists. Millions of people are already there. The tools are available. What's missing is <strong>political will</strong>, ours and the tech community's, to take seriously the idea that technology can serve humanity, not profit.</p>

<p><strong>Concrete action for today:</strong> Create your account on Mastodon. Choose an instance. Follow this blog. And when you write your next post, publish it on the Fediverse first. Each migration is an act of resistance. Each new instance is a victory.</p>

<p>May the code be free. And may freedom guide us.</p>

<h2>References</h2>

<ul class="references">
  <li><a href="https://pt.wikipedia.org/wiki/Fediverso">Wikipedia. "Fediverse". Accessed: 2026-04-08.</a></li>
  <li><a href="https://joinmastodon.org">Mastodon. "Social networking that's not for sale". Accessed: 2026-04-08.</a></li>
  <li><a href="https://joinpeertube.org">PeerTube. "An alternative to Big Tech's video platforms". Framasoft. Accessed: 2026-04-08.</a></li>
  <li><a href="https://www.w3.org/TR/activitypub/">W3C. "ActivityPub Specification". Accessed: 2026-04-08.</a></li>
  <li><a href="https://platformcoop.net">Platform Cooperativism Consortium. Accessed: 2026-04-08.</a></li>
  <li><a href="https://docs.joinmastodon.org">Mastodon Documentation. Accessed: 2026-04-08.</a></li>
  <li><a href="https://docs.joinpeertube.org">PeerTube Documentation. Framasoft. Accessed: 2026-04-08.</a></li>
  <li><a href="https://fsf.org">Free Software Foundation. Accessed: 2026-04-08.</a></li>
  <li><a href="https://github.com/mastodon/mastodon">Mastodon GitHub Repository. Accessed: 2026-04-08.</a></li>
  <li><a href="https://github.com/Chocobozzz/PeerTube">PeerTube GitHub Repository. Accessed: 2026-04-08.</a></li>
</ul>]]></content:encoded>
      <pubDate>Wed, 08 Apr 2026 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Open Source</category>
      <category>Tecnologia e Sociedade</category>
      <category>Forças Produtivas</category>
      <category>Futuro do Trabalho</category>
      <category>Economia de Plataforma</category>
      <category>Ética na Tecnologia</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/58cfc488-e57a-4083-9f41-a13b0caa6b7b" medium="image"/>
    </item>
    <item>
      <title>GPL vs MIT</title>
      <link>https://www.mmlo.app/en/blog/gpl-vs-mit</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/gpl-vs-mit</guid>
      <description>Choosing a License is a Political Act</description>
      <content:encoded><![CDATA[<p>When you choose a license for your software, you're making a decision that goes far beyond technical convenience. You're defining who controls the code you write, who benefits from it, and what kind of digital commons we're collectively building. GPL and MIT aren't just different legal texts: they're distinct political projects for the future of technology.</p><h2>What Copyleft Is and Why It Matters</h2><p>To understand the political difference between GPL and MIT, we first need to understand the concept of <em>copyleft</em>. Contrary to what the name might suggest, copyleft isn't the opposite of copyright. It's a legal technique that uses copyright laws to guarantee freedom, not restriction.</p><p>The Free Software Foundation defines four essential freedoms that all free software must guarantee: the freedom to use the software for any purpose, the freedom to study and modify the code, the freedom to share copies, and the freedom to distribute modified versions. When a program offers all these freedoms, we call it free software.</p><p><strong>Copyleft is the mechanism that ensures these freedoms are preserved</strong> in all derived versions. When you modify software under copyleft and distribute your version, you must do so under the same terms. This creates a protected digital commons: a common good that can't be enclosed by private interests.</p><h3>The Four Fundamental Freedoms</h3><ul><li><p><strong>Freedom 0:</strong> Use the software for any purpose</p></li><li><p><strong>Freedom 1:</strong> Study how the software works and modify it</p></li><li><p><strong>Freedom 2:</strong> Share copies with other people</p></li><li><p><strong>Freedom 3:</strong> Distribute modified versions to benefit the community</p></li></ul><p>According to Brett Smith of the GNU Project: "When a program offers users all these freedoms, we call it free software. Developers who write software can release it under the terms of the GNU GPL. When they do that, it will be free software, and it will remain free software, no matter who changes or distributes the program."</p><h2>GPL: Defending the Digital Commons Against Enclosure</h2><p>The <strong>GNU General Public License (GPL)</strong> is the most widely used copyleft license. Created by Richard Stallman in 1989, its most recent version (GPLv3) was published on June 29, 2007, after 18 months of public consultation with thousands of comments.</p><p>The GPL operates as a shield against the enclosure of the digital commons. When a company uses GPL code in its product and distributes that product, it must open-source the modifications. This prevents the commons from being transformed into private property without reciprocation.</p><p>The Linux kernel is the most emblematic example. Licensed under GPL v2 since 1992, it remains free and open to this day. Any distributed modification must be open-sourced under the same terms. This mechanism has created one of humanity's largest digital commons: <strong>96.3% of web servers run GNU/Linux</strong>, according to W3Techs data from 2024.</p><h3>GPL v3 Protections Against Modern Threats</h3><p>GPLv3 was updated to protect copyleft against three contemporary threats:</p><ul><li><p><strong>Tivoization:</strong> Devices that run GPL software but prevent user modifications. GPLv3 requires distributors to provide all the information needed to install modified software on the device.</p></li><li><p><strong>Anti-circumvention laws:</strong> Laws like the DMCA criminalize breaking DRM. GPLv3 states that the covered software doesn't count as an effective "protection" technology measure, preserving your right to write code that bypasses DRM.</p></li><li><p><strong>Discriminatory patent agreements:</strong> When someone distributes GPLv3 software, they must provide all recipients with any necessary patent licenses. If a licensee tries to sue users for patent infringement, their license is terminated.</p></li></ul><p>As Smith writes: "We updated the GPL to protect your copyleft from being undermined by legal or technological developments."</p><h2>MIT and Apache: Individual Freedom or Open Door to Appropriation?</h2><p>The <strong>MIT</strong> and <strong>Apache 2.0</strong> licenses belong to the category of permissive licenses. They grant broad freedoms: you can use, modify, distribute, and even incorporate the code into proprietary software. The only typical requirement is to preserve the original copyright notice.</p><p>From the perspective of the developer's individual freedom, these licenses are maximal. You can do almost anything with the code. But from the perspective of the digital commons, they represent a structural vulnerability: anyone can take the code, modify it, close it, and sell it as a proprietary product.</p><h3>The React Case: When a Permissive License Allows Unilateral Change</h3><p>In 2017, Facebook announced that it would change React's license from MIT to BSD+Patents. This new license included a clause that allowed Facebook to revoke patent rights if you sued the company for any reason. The community reacted with alarm. Projects like WordPress and Automattic declared that they could no longer use React.</p><p>In 2018, after intense pressure, Facebook reverted to MIT. But the damage was done: the community learned that <strong>permissive licenses allow unilateral changes</strong> that can harm those who depend on the code. With copyleft, this wouldn't be possible: the terms are locked to protect the commons.</p><h3>Structural Comparison</h3><p>According to GitHub Choose a License, the fundamental difference is in the "Same license" column:</p><ul><li><p><strong>GPL:</strong> Modifications must be released under the same license (copyleft)</p></li><li><p><strong>MIT/Apache:</strong> Modifications can be released under any license, including proprietary</p></li></ul><p>Apache 2.0 has advantages over MIT: it includes an explicit patent grant and is compatible with GPLv3. But it maintains the same structural problem: it doesn't protect the commons against enclosure.</p><h2>Real Cases: Linux (GPL), React (MIT), Git (GPL)</h2><p>To understand the practical implications, let's analyze three emblematic cases:</p><h3>Linux: Copyleft Keeps Critical Infrastructure Free</h3><p>The Linux kernel, under GPL since 1992, is the proof of concept of copyleft on a global scale. No vendor has managed to close the kernel. Any company that wants to use Linux in its product must open-source the kernel modifications. This has created a level playing field where even small companies can compete using the same infrastructure as tech giants.</p><h3>React: Permissive Freedom with Risk of Co-option</h3><p>React, under the MIT license, has become the world's most popular front-end library. But the BSD+Patents episode revealed the fragility of the model: a single company controls the license and can change the terms unilaterally. The community had to mobilize to reverse the decision: a political cost that wouldn't exist with copyleft.</p><h3>Git: GPL Ensures Development Tool Remains Free</h3><p>Git, created by Linus Torvalds in 2005, is licensed under GPL v2. This means that any distributed modification must be open-sourced. Companies like GitHub, GitLab, and Bitbucket can offer commercial services on top of Git, but they can't close the tool itself. The commons remains accessible to all developers.</p><h2>What Political Project Does Your License Support?</h2><p>We've arrived at the central question: when choosing a license for your project, what world are you helping to build?</p><h3>Choose GPL if You Believe In:</h3><ul><li><p><strong>Defense of the digital commons:</strong> You want your code and all improvements to remain accessible to everyone, indefinitely</p></li><li><p><strong>Reciprocity:</strong> Those who benefit from your work should contribute back</p></li><li><p><strong>Protection against enclosure:</strong> Companies shouldn't be able to take your work, close it, and sell it without open-sourcing their modifications</p></li><li><p><strong>Solidarity among developers:</strong> Building common goods that protect all contributors</p></li></ul><h3>Choose MIT/Apache if You Believe In:</h3><ul><li><p><strong>Maximum adoption:</strong> You prioritize getting as many people as possible to use your code, even in proprietary software</p></li><li><p><strong>Individual freedom of the developer:</strong> Each person should be able to do whatever they want with your code</p></li><li><p><strong>Corporate pragmatism:</strong> Companies can contribute without open-sourcing their proprietary code</p></li><li><p><strong>Trust in social norms:</strong> You believe the community will push back against abuses, even without legal protection</p></li></ul><p>There's no universal answer. But it's essential to recognize that <strong>a license is politics materialized in code</strong>. Your choice has real consequences for the type of technological ecosystem we're building.</p><h2>Alternatives and Paths: AGPL, Emerging Licenses</h2><p>For developers looking to further strengthen copyleft, there are alternatives:</p><h3>AGPL v3: Closing the SaaS Gap</h3><p>The GNU Affero GPL v3 extends copyleft to software running on network servers. Under traditional GPL, if you run modified software on a server without distributing binaries, you don't have to open-source the code. AGPL closes this gap: users who interact with the software via the network have the right to receive the source code.</p><p>This is crucial in the SaaS era. Without AGPL, companies like Amazon or Google can take free software, modify it, offer it as a closed service, and not contribute back. With AGPL, they must open-source the modifications.</p><h3>Emerging Anti-Capitalist Licenses</h3><p>New licenses are emerging to address limitations of traditional open source:</p><ul><li><p><strong>License for Oppressive Purposes:</strong> Prohibits use by police, military, or surveillance forces</p></li><li><p><strong>Cooperative Non-Violent Public License:</strong> Requires companies to pay for commercial use, reverting to cooperatives</p></li><li><p><strong>Hippocratic License:</strong> Based on the Universal Declaration of Human Rights, prohibits violations</p></li></ul><p>These licenses are controversial: the OSI doesn't consider them "open source" because they restrict fields of activity. But they represent an important critique: <strong>political neutrality is illusory</strong>. Permissive licenses are also political: they simply favor capitalist appropriation.</p><h2>Conclusion: Your License Is Your Manifesto</h2><p>We've reached the end of this analysis. What have we learned?</p><p>First, that <strong>GPL and MIT aren't equivalent</strong>. They represent distinct philosophies about property, commons, and reciprocity. GPL uses copyleft to protect the digital commons against enclosure. MIT prioritizes individual freedom and maximum adoption, even at the expense of the commons.</p><p>Second, that <strong>a license is political</strong>. There's no neutral choice. By licensing your code, you're taking a position on what kind of technological ecosystem you want to inhabit.</p><p>Third, that <strong>the React case demonstrated the vulnerability of permissive licenses</strong>. A company can change the terms unilaterally, and the community has little recourse. Copyleft locks the terms to protect all participants.</p><p>As developers, we have a concrete choice: to be mere digital executors or active builders of a more equitable future. The code we write today is the world we'll live in tomorrow.</p><p><strong>Your license is your manifesto.</strong> What world are you coding?</p><h2>References</h2><ul><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://www.gnu.org/philosophy/free-sw.html">Free Software Foundation. "The Free Software Definition". GNU Project.</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://www.gnu.org/licenses/copyleft.html">GNU Project. "What is Copyleft?".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://www.gnu.org/licenses/quick-guide-gplv3.html">Smith, Brett. "A Quick Guide to GPLv3". GNU Project.</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://www.gnu.org/licenses/license-list.html">GNU Project. "License List".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://choosealicense.com/appendix/">GitHub. "Choose a License".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://w3techs.com/technologies/details/os-linux">W3Techs. "Usage of Linux as Server-Side Operating System". 2024.</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://survey.stackoverflow.co/2023">Stack Overflow. "Developer Survey 2023".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/COPYING">Linux Kernel Organization. "Linux Kernel Source: COPYING".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://www.gnu.org/licenses/agpl-3.0.html">Free Software Foundation. "GNU Affero General Public License v3".</a></p></li><li><p><a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://opensource.org/licenses">Open Source Initiative. "Licenses".</a></p></li></ul><p></p>]]></content:encoded>
      <pubDate>Wed, 08 Apr 2026 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Open Source</category>
      <category>Tecnologia e Sociedade</category>
      <category>Forças Produtivas</category>
      <category>Futuro do Trabalho</category>
      <category>Economia de Plataforma</category>
      <category>Ética na Tecnologia</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/b26e5a00-e688-441c-abb7-bc027ae4e867" medium="image"/>
    </item>
    <item>
      <title>Gordzilla&apos;s Den</title>
      <link>https://www.mmlo.app/en/blog/toca-do-gordzilla</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/toca-do-gordzilla</guid>
      <description>Neo-Brutalism&apos;s Raw Power in Service of Brand Identity</description>
      <content:encoded><![CDATA[<p>The story of <strong>Toca do Gordzilla</strong> is a war cry against sameness. In a saturated *fast-food* market, where most burger joints are content to be just another in line for generic SaaS systems, Toca decided they needed something more. They needed a system that not only managed orders but **breathed the brand's identity** into every click. Thus was born the <u>Toca System</u>, an application that transformed order management into a themed and visually aggressive experience.</p><blockquote><p>"The goal was to escape the trap of generic SaaS, where the interface lacks brand personality, making it just another among the others."</p></blockquote><p>The challenge was clear: to develop a proprietary system where the interface carried the company's form of communication and identity, from the choice of words to the visual design. The Toca System is proof that technology, when aligned with brand strategy, becomes a **brutal competitive differentiator**.</p><h3>The Metamorphosis of Language: From Products to Monsters</h3><p>The first step in injecting Toca do Gordzilla's personality into the system was the **metamorphosis of language**. The menu, which anywhere else would be a cold list of "products" and "add-ons," was transformed into a themed battlefield:</p><ul><li><p><strong>Products</strong> became <strong>"Monsters."</strong></p></li><li><p><strong>Add-ons</strong> became <strong>"Mutations"</strong> that make the monster stronger.</p></li></ul><p>This language didn't stop at the menu. It invaded the **Administrative Interface**, transforming the kitchen routine into a narrative:</p><ol><li><p>The "no orders" status became <em>"The beast is sleeping."</em></p></li><li><p>The "in preparation" status became <em>"The monster is on the grill."</em></p></li><li><p>The "send to kitchen" button became <em>"Send to the grill."</em></p></li></ol><p>This layer of thematic communication ensures that every interaction with the system, whether from the customer or the administrator, reinforces Toca do Gordzilla's unique identity. It's the **brand's personality** encoded in every *string* of the system.</p><h2>The Visual Shock: The Neo-Brutalist Aesthetic</h2><p>For the design, the choice was equally audacious: **Neo-Brutalism**. Far from the smooth interfaces and gradients of current trends, the Toca System embraces an aesthetic that screams exclusivity and strength, reflecting the "gordzilla" nature of the brand.</p><h3>Visual Characteristics of Neo-Brutalism in the Toca System</h3><p>The design system was built on principles that ensure a visually impactful and functional experience:</p><ol><li><p><strong>Flat and Strong Colors:</strong> The color palette uses HSL to define primary and action colors with high contrast. The base color (Medium Green) and the action color (Orange) are used strategically to guide the eye.</p></li><li><p><strong>Brutal Borders:</strong> Thick borders (2px, 3px, 4px) without excessive rounding. The use of <u>Low Border Radius</u> (2px to 6px maximum) gives a robust and industrial appearance.</p></li><li><p><strong>Displaced Shadows:</strong> The use of hard shadows without *blur* (<code>4px 4px 0px 0px</code>) creates an effect of depth and marking, a classic touch of Neo-Brutalism.</p></li><li><p><strong>High-Contrast Typography:</strong> The combination of **Space Grotesk** for titles and **IBM Plex Sans** for the body of the text ensures legibility and a modern, technical touch.</p></li></ol><p>This design choice is not just aesthetic; it's **functional**. The high contrast and marked borders make the interface highly scannable, especially in the Administrative Interface, where speed and clarity in order management are crucial.</p><h2>The Architecture: Modern MERN and Firebase</h2><p>Behind the brutalist aesthetic, the Toca System is a modern **SPA (Single Page Application)**, built with a robust and scalable stack:</p><h3>Technology Stack</h3><p>Component Key Technologies Function <strong>Frontend</strong> React 18.3.1, TypeScript 5.8.3, Vite 5.4.19 UI construction and development performance. <strong>UI/Design System</strong> Tailwind CSS 3.4.17, shadcn/ui, Radix UI Utility-first styling and accessible components. <strong>Backend/Services</strong> Firebase 12.2.1 (Firestore, Auth, Storage) Complete BaaS platform for database, authentication, and storage. <strong>State Management</strong> TanStack Query (React Query) 5.83.0 Efficient server state management and caching.</p><h3>Architecture and Structure</h3><p>The application follows a **layered architecture** with separation of responsibilities, facilitating maintenance and evolution. The folder structure reflects the concern with organization and the application of design patterns:</p><ul><li><p><code>services/</code>: Direct communication with Firebase.</p></li><li><p><code>repositories/</code>: Implementation of the Repository pattern to abstract the data source.</p></li><li><p><code>useCases/</code>: Implementation of Clean Architecture, ensuring that the business logic is isolated and testable.</p></li></ul><p>This organization allows Toca do Gordzilla to migrate the *backend* from Firebase to a proprietary server (such as a Node.js/Express) in the future without rewriting the business logic.</p><h2>Security and Business Rules in Firebase</h2><p>Security was a fundamental pillar, especially in the management of sensitive order and customer data. The use of **Firebase Authentication** and **Firestore Rules** ensures that only authorized users access the correct information.</p><h3>Implemented Security Rules</h3><p>The Firestore security rules were written to enforce granular access control:</p><ol><li><p><strong>Orders:</strong> Full read access only for Admins. Users can only read their own orders. Creation allowed for any authenticated user.</p></li><li><p><strong>Menu:</strong> Public read access (for the customer to see the menu), but write access restricted to Admins.</p></li><li><p><strong>Customers:</strong> Read/Update access allowed only for the user themselves or for Admins.</p></li></ol><p>In addition, **Security Headers** were configured in Vercel (`X-Content-Type-Options: nosniff`, `X-Frame-Options: DENY`) to mitigate common web security attacks.</p><h2>Functionalities: The System that Sells the Experience</h2><p>The Toca System is not just a CRUD. It is a system that manages the customer experience and the efficiency of the operation.</p><h3>Customer Interface (The Battlefield)</h3><ul><li><p><strong>Digital Menu:</strong> Visualization of categories with *sticky* navigation, item customization (mandatory/optional options, add-ons, observations), and **persistent cart** (via *localStorage*).</p></li><li><p><strong>Themed Customization:</strong> The customer assembles their "Monster" by choosing their "Mutations."</p></li><li><p><strong>Time Verification:</strong> The system verifies the operating hours, preventing orders outside of business hours.</p></li></ul><h3>Administrative Interface (The Headquarters)</h3><ul><li><p><strong>Order Management:</strong> Visualization by status (New, In Preparation, Out for Delivery, Completed), with advanced filters and **real-time status updates**.</p></li><li><p><strong>Menu Management:</strong> Complete CRUD of categories and items, with image upload and configuration of differentiated prices (counter vs. delivery).</p></li><li><p><strong>Delivery Driver Management:</strong> CRUD of delivery drivers, availability control, and order assignment.</p></li></ul><p>The Toca System is proof that, instead of adapting to a generic system, it is possible to build a system that perfectly adapts to a brand's **identity, language, and visual strategy**. Toca do Gordzilla stopped being just another among so many and started having a unique and exclusive system.</p><p><u>The Toca System is the brute force of technology at the service of brand personality.</u></p><p><strong>Project Resources:</strong></p><p>Demonstration Video: <a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://youtu.be/VuJWWDeVPqI">See the Toca System in Action</a></p>]]></content:encoded>
      <pubDate>Sun, 04 Jan 2026 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>React</category>
      <category>Vite</category>
      <category>Firebase</category>
      <category>Tailwind</category>
      <category>Vercel</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/c48139d1-240a-4ad2-a80f-68fd3be2f893" medium="image"/>
    </item>
    <item>
      <title>Code Forge</title>
      <link>https://www.mmlo.app/en/blog/code-forge</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/code-forge</guid>
      <description>The Crucible of Rigor: Where AI Audits the Process and Evidence is Law</description>
      <content:encoded><![CDATA[<p>The story of <strong>Code Forge</strong> isn't about creating code. It's about <u>redefining what "completion" and "done" mean</u> in software development. In a landscape where Artificial Intelligence (AI) has become a powerful tool, the line between functional code and an auditable process has become critical. Code Forge was born from a simple yet profoundly uncomfortable premise:</p><blockquote><p>"If verifiable evidence doesn't exist, then a valid conclusion doesn't exist."</p></blockquote><p>This open-source system wasn't designed to be user-friendly. It was designed to be **correct**. It's the formalization of a **rigor-driven project management** method, where AI is elevated to the role of impartial process auditor, and subjectivity is banished from the acceptance of steps.</p><h3>The Silent Problem of Project Management</h3><p>In traditional development, we always run into the same management and validation problems:</p><ul><li><p>AI "agrees" with what we say, generating a false sense of security in planning.</p></li><li><p>Step reviews and tests are often based on subjective or implicit criteria.</p></li><li><p>The famous phrase <em>"It works on my machine"</em> ironically becomes the success criterion for advancing a step.</p></li><li><p>Project advancement decisions aren't traceable, getting lost in chats or meetings.</p></li></ul><p><strong>Code Forge</strong> solves this by imposing a structure of rigor, clearly and non-negotiably separating the roles of creation, implementation, and, most importantly, **process validation**. It's a guide on how to start a project from scratch and reach the end with **full traceability**.</p><h2>The Chronological Autopsy: The Non-Linear History of Code Forge</h2><p>The <u>real timeline</u> of development reveals the inflection points that transformed a vague idea into a system with a clear intention of **process auditing**.</p><h3>Phase 0 – The Raw Idea and the Nebulous Scope</h3><p>It all started with the intention: <em>"I want an AI application to help me plan projects."</em></p><p>At this initial point, AI was seen as a "generator" of solutions, not as a **validation system**. The intention was good, but the completion criteria were nebulous. This is where most projects die, happy and confident in their own subjectivity.</p><h3>Phase 1 – Infrastructure First, Ego Later: The Trial by Fire</h3><p>The first serious break with the "we'll see later" approach was the decision to start with the infrastructure. The project was immediately anchored in **Docker** and configured to be *self-hosted*. The minimum infrastructure needed to be live before any feature.</p><p>This resulted in a controlled, functional environment, but without any glamour. The project began to exist in the real world, rejecting the temptation to start with the pretty frontend. It was a <strong>key decision</strong> that established rigor from the beginning.</p><h3>Phase 4 – The Shock: The Inflection Point</h3><p>This was the **most important moment**. We stalled and realized: the goal wasn't a generic app with an LLM. The goal was to **replicate exactly the rigor and auditing process** that we were using to build Code Forge itself. The AI needed to be a **process auditor**, not just an idea generator.</p><blockquote><p>"The goal was to replicate exactly the process that was happening here: AI as a plan generator + evidence evaluator + Step auditor."</p></blockquote><p>The result was a **conceptual reset** and a total re-anchoring of the idea. The project finally found its axis: the **formalization of the method**.</p><h3>Phase 5 – Formalization of the Method: The DNA of Code Forge</h3><p>From here, rigor became law. The elements that make up the system's DNA were defined:</p><ol><li><p>**Numbered Steps:** Each task is a step with a closed scope.</p></li><li><p>**Mandatory Tests:** Each Step has a closed list of tests.</p></li><li><p>**Evidence as Artifact:** The result of the test execution is an immutable artifact.</p></li><li><p>**AI as Validator:** The AI verifies the evidence, it doesn't give opinions.</p></li></ol><p>The system stopped being "creative" and became **verifiable**. The **rigid separation of roles** and the **explicit rejection of subjectivity** were implemented.</p><h3>Phase 6 – Evidence Standard: The Human-AI Contract</h3><p>The detail that elevated the project from "interesting" to **dangerous** (in a good way) was the creation of a **standard for submitting tests**. The **Evidence Report Template** was born, a contract between the human and the AI:</p><ul><li><p>The user gains the **explicit responsibility** of providing proof.</p></li><li><p>The AI gains **clear limits** for its actions.</p></li></ul><p>The system became auditable, a point where traditional companies would start to sweat. The evidence is the contract.</p><h2>The Conceptual Architecture: Rigor in 10 Steps</h2><p><strong>Code Forge</strong> structures development into **verifiable steps**, each with objective criteria and auditable evidence. The architecture follows a 10-step flow, where advancement is an automated gate.</p><h3>Step 1 – Project Definition</h3><p>The project is created with initial metadata, technical and business context. The deliverable is a versioned <strong>Project Descriptor</strong>.</p><h3>Step 2 – Breakdown into Steps</h3><p>Each project is divided into **numbered and named Steps**. Each Step has an explicit objective, closed scope, completion criteria, and mandatory tests. The deliverable is the <strong>Step Definition Document</strong>.</p><h3>Step 3 – AI-Assisted Planning</h3><p>The AI acts as a technical analyst and ambiguity questioner, generating an initial plan and mapping risk points. But it <u>doesn't decide anything on its own</u>. The deliverable is the <strong>Proposed Technical Plan</strong>.</p><h3>Step 4 – Human or Assisted Implementation</h3><p>The code, configuration, or documentation is produced. The rule is clear: the implementation <u>doesn't advance the Step automatically</u>. The deliverables are the <strong>Technical Artifacts</strong> (code, config, docs).</p><h3>Step 5 – Formal Definition of Tests</h3><p>Each Step has a closed list of tests, with a unique ID, objective description, and **verifiable criteria**. The deliverable is the immutable <strong>Test Spec</strong> for the Step.</p><h3>Step 6 – Manual Execution of Tests</h3><p>The user executes the tests in the real world (CLI, UI, Logs, Prints). The system <u>doesn't trust the result</u>, only the artifact. The deliverables are the <strong>Raw Evidence</strong>.</p><h3>Step 7 – Generation of the Evidence Report</h3><p>A standardized Markdown report is generated, with blocks per test, verifiable evidence, and an explicit result (<code>PASS</code> | <code>FAIL</code>). The deliverable is the <strong>Evidence Report</strong>.</p><h3>Step 8 – Automated Validation</h3><p>An AI Agent acts as an **auditor**. It doesn't run tests, doesn't correct reports, and doesn't interpret intentions. It only answers: <code>accepted</code>, <code>partial</code>, or <code>rejected</code>. The deliverable is the <strong>Decision Trace</strong> and the Step Status.</p><h3>Step 9 – Blocking or Advancement</h3><p>A rejected Step blocks advancement. An accepted Step releases the next one. <u>No "workarounds"</u>. The system imposes rigor.</p><h3>Step 10 – History and Traceability</h3><p>Everything is recorded: who did it, when, with what evidence, and what decision was made. The deliverable is the complete <strong>Audit Log</strong>.</p><h2>Differentials and Functionalities: Where Code Forge Enforces Rigor</h2><p>The value of Code Forge lies in its technical differentiators, which separate it from any traditional project management tool.</p><h3>Differential 1: AI as Auditor, Not as Authority</h3><p>The AI isn't the author of the code, nor the final authority. It's the **process auditor**. It doesn't "think" the process is good; it **verifies** whether the presented evidence fulfills the contract. This is an application of AI that imposes rigor, not complacency.</p><h3>Differential 2: Evidence as a Non-Negotiable Contract</h3><p>The system operates under the law of evidence. Without evidence, there's no conclusion. This forces the developer to be explicit about what they're testing and what they're delivering. The <strong>Evidence Report</strong> is the contract. If the report says <code>PASS</code>, the evidence must prove it irrefutably.</p><h3>Differential 3: Rigid Separation of AI Roles</h3><p>To avoid cross-bias, Code Forge separates the roles of AI:</p><ul><li><p><strong>Planning AI:</strong> Questions the ambiguity of the Step.</p></li><li><p><strong>Validation AI:</strong> Acts as an auditor, verifying the evidence.</p></li></ul><p>No mixing. No cross-bias. The final acceptance decision is based on a <strong>Decision Trace</strong> generated by a specialized agent.</p><h3>Differential 4: Zero Subjectivity in Acceptance</h3><p>The acceptance process is binary, traceable, and reproducible. The system doesn't accept "maybe" or "almost." The result is <code>accepted</code> or <code>rejected</code>, with the <strong>Decision Trace</strong> as justification. This ensures that project advancement is always based on facts.</p><h3>Differential 5: True Open Source (GPL v3)</h3><p>Code Forge is licensed under **GPL v3**. The reason is strategic: to ensure that the logic, prompts, and validations remain open and auditable. In a system that makes technical decisions, transparency is fundamental. This prevents the silent appropriation of the core and forces the community to maintain rigor.</p><h2>Technological Stack and Infrastructure</h2><p>Code Forge is built on a modern and robust stack, designed to be *self-hosted* and portable.</p><p>Component Key Technologies Function <strong>Backend</strong> Node.js / Bun, TypeScript, REST API Step Management, Evidence Storage, Validation API Exposure. <strong>Frontend</strong> React, Vite, UI based on Steps and Evidence Interface for the user to define Steps, run tests, and upload Evidence. <strong>AI</strong> External LLMs via API, Versioned Prompts, Specialized Agents Acts as a Planning Agent and Validation Agent (Auditor). <strong>Infrastructure</strong> Docker, Local or self-hosted environment Ensures portability and isolation of the execution environment.</p><h2>Future Plans: Coherent Evolution</h2><p>The Code Forge roadmap is focused on coherent evolution, without magical promises, only on improving rigor.</p><h3>Short Term</h3><ul><li><p><strong>CI/CD Plugin:</strong> Integration so that Step validation can be triggered automatically in pipelines.</p></li><li><p><strong>Automatic Validation of Syntactic Evidence:</strong> Use of AI to validate the syntax of the Evidence Report before sending it to the main auditor.</p></li><li><p><strong>Status Dashboard:</strong> Clear view of progress and gates blocked per project.</p></li></ul><h3>Medium Term</h3><ul><li><p><strong>Integration with GitHub/GitLab:</strong> Synchronization of Steps with issues and pull requests.</p></li><li><p><strong>Evidence Signature:</strong> Use of cryptography to guarantee the immutability and origin of the evidence.</p></li><li><p><strong>Multi-LLM Consensus Validator:</strong> Use of multiple LLMs to validate the same evidence, increasing the impartiality of acceptance.</p></li></ul><h3>Long Term</h3><ul><li><p><strong>Evidence-as-Code:</strong> Definition of tests and evidence criteria directly in code.</p></li><li><p><strong>Compliance Audit Mode:</strong> Generation of technical audit reports for compliance (ISO, SOC, LGPD).</p></li></ul><h2>Conclusion: Code Forge Doesn't Try to Be User-Friendly</h2><p><strong>Code Forge</strong> assumes something that the software industry often avoids: People make mistakes, AI makes mistakes, and loose processes generate fragile software. This system exists for those who value technical truth above comfort. It's for those who prefer a **well-justified "no"** over a **comfortable and wrong "yes."**</p><p>Code Forge is the **formalization of rigor** that transforms development into an auditable science.</p><p>The project is open. Come audit the process.</p><p>Open Source Repository: <a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://github.com/MierzwaMatheus/forge-code">Code Forge Source Code</a></p><p>Demonstration Video: <a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://youtu.be/t7gJuxH5ye0">See how Code Forge works</a></p>]]></content:encoded>
      <pubDate>Mon, 29 Dec 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>MongoDB</category>
      <category>Express</category>
      <category>React</category>
      <category>NodeJS</category>
      <category>Docker</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/e03c1a95-e319-4259-9346-d60ec5397e86" medium="image"/>
    </item>
    <item>
      <title>OliBase Odyssey</title>
      <link>https://www.mmlo.app/en/blog/odisseia-do-olibase</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/odisseia-do-olibase</guid>
      <description>The Saga of Building a Backend as a Service (BaaS) in MERN: Where Code Becomes Architecture</description>
      <content:encoded><![CDATA[<p>The journey began not with a *bug*, but with a <strong>developer's existential crisis</strong>. For months on end, immersion in the <strong>MERN Stack</strong> (MongoDB, Express, React, Node) had been methodical and successful. Courses, documentation, and small portfolio projects were completed with Swiss watch precision. However, a cold, cutting truth hung in the air: <em>"I still hadn't built something big enough to force me to make difficult decisions."</em></p><p>The problem wasn't a lack of knowledge, but the absence of a <u>challenge that would force the transition from "knowing how to do" to "knowing how to architect."</u> The ambition was clear: to go beyond the tutorial, embrace the chaos of refactoring, make mistakes, correct them, and finally, <strong>understand the soul of *backend* decisions</strong>. It was in this scenario of searching for a greater technical purpose that the idea of <strong>OliBase</strong> ascended, like a digital phoenix.</p><h3>The Call of Architecture: The BaaS Inspiration (Phase 1)</h3><p>The inspiration came from the giants that simplify the universe of development: <strong>BaaS</strong> (Backend as a Service), such as the ubiquitous Firebase and the elegant Supabase. These platforms solve complex problems—authentication, data management, security—with almost magical ease. So I, fascinated by this abstraction, launched the ultimate challenge:</p><blockquote><p>"What if I tried to build something that replicated this complexity, from scratch, using MERN? Not to compete in the market, but to force a deep learning experience, where each line of code was an architectural lesson?"</p></blockquote><p><strong>OliBase</strong> was born, then, as a <u>technical manifesto</u>. A minimal, elegant, and above all, <strong>honest</strong> BaaS, where the <strong>Backend would be the absolute protagonist</strong>, demanding mastery of each layer of the *stack*.</p><h3>The First Labyrinth: The Separation of Identities (Phases 2 and 3)</h3><p>The first obstacle was not technical, but conceptual: <strong>authentication</strong>. In a BaaS system, there are actually <u>two classes of users</u>, and mixing them would be a fatal design flaw. The critical perception was that:</p><ul><li><p>Authentication for the <strong>dashboard</strong> (the "human user" who manages the BaaS) is one thing.</p></li><li><p>Authentication for <strong>end users</strong> (the customers of the applications that use the BaaS) is something completely different.</p></li></ul><p>This distinction imposed the need for <u>two isolated authentication systems</u>. The first step was the implementation of <strong>Human Auth</strong>, the administrative foundation. It was here that <code>JWT</code> (JSON Web Token) was employed to protect the routes, ensuring that only legitimate administrators could create projects and generate access keys. A solid foundation, without which the house of cards would collapse.</p><h3>The Great Test: Creating the Multi-Tenant Universe (Phase 4)</h3><p>OliBase could not be just a data repository; it needed to be a <strong>universe of isolated data</strong>. The golden rule was <strong>true multi-tenancy</strong>. This means that if Project Alpha and Project Beta use OliBase, Alpha's data must be <u>absolutely invisible</u> to Beta, and vice versa. The technical question was the most crucial:</p><blockquote><p>"How to guarantee the strict separation of data from different applications, even if they share the same physical database?"</p></blockquote><p>The answer was the creation of the <strong>Project</strong> model and the linking of <strong>API Keys</strong> to each one. Isolation was transformed into a protocol requirement, requiring each data request to carry two mandatory *headers*, like a passport and a visa:</p><ol><li><p><code>x-api-key</code>: The secret key that identifies the application.</p></li><li><p><code>x-project-id</code>: The unique identifier of the project.</p></li></ol><p>Membership validation became the most rigorous *middleware* in the system. The mental test that guided this phase was the security mantra: <em>"If someone, by mistake or bad faith, tries to use the key of one project to access data from another, the system </em><strong><em>MUST fail</em></strong><em> categorically and immediately."</em> It was at this point that <u>true isolation</u> was consolidated, elevating OliBase to the level of a production system.</p><h3>The Voice of Conscience: Logging and Observability (Phases 5 and 6)</h3><p>With security established, it was time to give OliBase a <strong>voice</strong>. A complex system needs <strong>observability</strong>, the ability to know what is happening in real time. Before building the complete database, a simple <strong>KV (Key-Value)</strong> system was implemented to validate the data flow and, more importantly, <strong>Logging</strong>.</p><p>The <strong>Logging Middleware</strong> was the solution. It was not a luxury, but an auditing necessity. Designed to be <strong>asynchronous and non-blocking</strong>, it recorded the duration, status code, and route of each request, but with a golden rule: <u>never record sensitive *payloads*</u>. From that moment on, OliBase gained a <strong>memory</strong>, allowing the developer to see, in real time, the pulsating life of their creation. The project ceased to be mere code and became a <strong>"product"</strong> with self-auditing capabilities.</p><h3>The Inflection Point: End-User Authentication (Phases 7 and 8)</h3><p>The real trial by fire came with the implementation of <strong>End-User Auth</strong>. The complexity was exponential: the system needed to manage the identity of users who, although they might have the same email, belonged to <u>completely different data universes</u> (projects). The architectural decision was brutally elegant:</p><ul><li><p>The end user is always a <strong>citizen of a project</strong>.</p></li><li><p>A <code>JWT</code> and a <code>Secret</code> completely separate from the dashboard authentication were created.</p></li><li><p>The end-user <code>JWT</code> carries, inseparably, the <code>projectId</code>, ensuring that they can only interact with their own universe.</p></li></ul><p>Endpoints like <code>POST /api/auth/signup</code> and <code>POST /api/auth/login</code> were just the tip of the *iceberg*. The real engineering was in the <strong>Critical Validations</strong>: the API Key had to be valid, belong to the project, and, most importantly, the email had to be unique <u>within that specific project</u>. After a battery of tests that simulated the chaos of the real world, the result was a <strong>real multi-tenant Auth working</strong>, a triumph that validated the entire isolation architecture.</p><h3>The Crown of Security: Row Level Security (RLS) (Phases 9 and 10)</h3><p>A BaaS without a robust data system is just an authentication system. The implementation of the <strong>Data Layer</strong>, with dynamic <strong>Collections</strong> and <strong>Documents</strong>, required the application of an advanced concept: <strong><u>Row Level Security (RLS)</u></strong>. Inspired by the most sophisticated database systems, RLS is the security layer that ensures that <u>the user only sees what they are allowed to see</u>.</p><p>The access rules were like universal laws of OliBase:</p><ol><li><p>The <strong>Common User</strong> (the application's end customer) can only see and manipulate the documents they created themselves (identified by the <code>ownerId</code>).</p></li><li><p>The project <strong>Admin</strong> (the dashboard user) has a panoramic view, being able to see <em>everything</em> within their project.</p></li><li><p><strong>Projects</strong> remain in absolute isolation, with no chance of data crossing.</p></li></ol><p>The testing phase focused on <strong>negative cases</strong>: the attempt by User A to spy on User B's data, or the audacity of Project A to access Project B's database. Success was measured by the <strong>correct failure</strong> of these attempts. OliBase proved to be not just a data repository, but a <u>guardian of privacy and multi-tenant isolation</u>.</p><h3>The Baptism of Fire: The Reality of Correction (Phase 11)</h3><p>The most learning-rich part of any project is the moment when theory meets failure. The <strong>crucible of learning</strong> came with the discovery of a subtle bug: the administrator *Role* was not being saved correctly on *signup*. This was not a syntax error, but a <strong>flow logic error</strong>.</p><p>The correction required surgical intervention: a <strong>tweak to the signup</strong>, an <strong>explicit validation</strong> in the user creation flow, and a <u>complete refactoring of the authorization *middleware*</u>. <em>"This is real backend,"</em> was the conclusion. The failure proved that the system was not magical, but the result of meticulous engineering work, where each error was transformed into a stronger foundation.</p><h3>Closing the Circle: The Dashboard and Eventum (Phases 12, 13 and 14)</h3><p>To manage this architectural complexity, the <strong>Dashboard</strong> was built in <code>React</code>, in an act of self-reference: it consumed OliBase's <u>own API</u>. With a <em>minimalist cyberpunk</em> and *mobile-first* design, it was the command center for managing Projects, End Users, Collections, RLS, and, most vitally, viewing <strong>Logs</strong> in real time.</p><p>The development cycle was completed with <strong>Eventum</strong>, a real client application. Eventum was not a portfolio project; it was the <strong>living proof</strong> that OliBase worked. It performed:</p><ul><li><p>User authentication via OliBase.</p></li><li><p>CRUD operations protected by RLS.</p></li><li><p>Generation of logs visible in the dashboard.</p></li></ul><p>The final validation was a chorus of success: <strong>OliBase functional, Eventum functional, RLS working, Isolation working</strong>. The project left the realm of theory and established itself as a <u>functional and robust reality</u>.</p><h3>The Legacy: Proof That You Build Systems (Phase 15)</h3><p>What started as a search for a study project has transformed into a complex system that touches on all the pillars of modern software engineering.</p><p></p><p>The code, this work of engineering, will be made available under an <strong>open source license</strong>, and the journey has been documented on video, serving as a beacon for other developers seeking the transition from tutorial to architecture.</p><p></p><p><u>If you are passionate about backend, system architecture, or the power of the MERN Stack, this project is an invitation to a deep technical conversation.</u></p><p></p><p><strong>Odyssey Resources:</strong></p><p>Demonstration Video (Proof of Concept): <a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://youtu.be/Ryy0Fvo4A68">OliBase - Complete Architecture Demonstration</a></p><p>Open Source Repository (The Open Code):<a target="_blank" rel="noopener noreferrer nofollow" class="text-neon-purple underline cursor-pointer" href="https://github.com/MierzwaMatheus/OliBase"> OliBase Source Code</a></p>]]></content:encoded>
      <pubDate>Sat, 20 Dec 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>MongoDB</category>
      <category>Express</category>
      <category>React</category>
      <category>NodeJS</category>
      <category>Docker</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/e28056bf-a336-4138-8ff1-98f46af6f378" medium="image"/>
    </item>
    <item>
      <title>The Front End in the Age of Intelligence</title>
      <link>https://www.mmlo.app/en/blog/o-front-end-na-era-da-inteligencia</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/o-front-end-na-era-da-inteligencia</guid>
      <description>Navigating the Semantic Revolution to Build Superior Experiences</description>
      <content:encoded><![CDATA[<p>Imagine a world where words had no intrinsic meaning, where each sentence was just a random sequence of symbols, without context or purpose. It would be the chaos of incommunicability, a digital babel where information, however abundant, would remain inaccessible in its essence. For a long time, the internet, in its initial phase, operated in an analogous way. Pages upon pages of content were built with generic elements, such as <code>&lt;div&gt;</code> and <code>&lt;span&gt;</code>, which, although useful for visual organization, were meaningless to machines. It was like building a magnificent library, full of valuable volumes, but without any cataloging system, without indicating what was a title, an author, a chapter, or even the literary genre. The result? A treasure trove of information, but imprisoned by the lack of structure and meaning. It is in this scenario of apparent semantic disorder that the silent revolution of the Semantic Web begins, and semantic HTML, together with microdata, emerges as the fundamental pillars to build a web that not only displays, but understands, connects, and ultimately, liberates knowledge.</p> <p>Semantics, in its philosophical root, is the science of meaning. In the context of HTML, semantic HTML refers to the use of tags that explicitly convey the purpose and meaning of the content they surround. Instead of a generic <code>&lt;div&gt;</code> for a page header, we use <code>&lt;header&gt;</code>. For a navigation section, <code>&lt;nav&gt;</code>. For an independent and self-contained article, <code>&lt;article&gt;</code>. For a block of side content, <code>&lt;aside&gt;</code>. This is not merely a matter of elegance or code organization; it is a declaration of intent, a way of giving voice and intelligence to the very structure of the page. By doing so, we are building the first and most crucial layer of a truly intelligent web, where machines and humans can cooperate more effectively, overcoming the barriers of ambiguity and limited interpretation. It is the first step towards demystifying information and making it universally understandable.</p> <h4>The Struggle for Understanding: Why Semantic HTML is Essential for the Liberation of Information?</h4> <p>The adoption of semantic HTML is not an option, but an imperative necessity for any front-end developer who aims to build superior, inclusive, and above all, fair digital experiences. The benefits are multifaceted and directly impact accessibility, search engine optimization (SEO), and code maintainability, serving as tools to democratize access to knowledge.</p> <p>Firstly, **accessibility**. Imagine a blind or visually impaired user trying to navigate a website. They depend on assistive technologies, such as screen readers, software that converts visual content into audio or braille. If the site is built only with meaningless <code>&lt;div&gt;</code>s and <code>&lt;span&gt;</code>s, the screen reader has no context. It reads everything as a continuous block, without clear distinction between a main title, a paragraph of text, a list of items, or a navigation menu. It's like trying to read a book without punctuation, without paragraph breaks, without chapters – an exhausting and often fruitless experience. Semantic HTML, on the other hand, provides this vital structure. When a screen reader encounters a <code>&lt;nav&gt;</code> tag, it knows that there is a navigation area and can inform the user, allowing them to jump directly to the main content if they wish, or explore the menu options in an organized way. This is not just a good development practice; it is a fundamental right of digital inclusion. Denying semantics is, in essence, erecting invisible barriers for millions of people, limiting their access to knowledge, civic participation, and economic participation in the digital society. It is a form of exclusion that the Semantic Web seeks to dismantle, ensuring that information is a common good, accessible to all.</p> <p>Secondly, **SEO**. Search engines, such as Google, are the gatekeepers of information on the web, the great organizers of the vast ocean of data. They crawl billions of pages to understand their content and relevance, determining what will be presented to billions of users. Without semantic HTML, they need to "guess" what each part of your page means, based on less precise heuristics. With tags like <code>&lt;article&gt;</code>, <code>&lt;section&gt;</code>, <code>&lt;header&gt;</code>, and <code>&lt;footer&gt;</code>, you are speaking the language of search engine bots, providing them with clear and unambiguous clues about the hierarchy and purpose of your content. This not only improves indexing, making your page easier to find, but also increases the chances of a good ranking in search results, as the algorithm better understands the objective and relevance of your page for a given query. In a world where visibility is power and attention is a valuable currency, semantic HTML is the tool that democratizes access to attention, allowing relevant content to be found, regardless of the size of the corporation or the capital behind it. It is a crucial step to ensure that quality information, often produced by small collectives or individuals, is not lost in the noise.</p> <p>Finally, **maintainability and collaboration**. Semantic code is cleaner, more readable, and easier to understand, both for you in the future, when you need to revisit an old project, and for other developers who may work on the project. It is the difference between a well-organized instruction manual, with clear sections and subtitles, and a pile of loose and disorganized notes. This intrinsic clarity of semantic code reduces errors, speeds up the development of new features, and facilitates the continuous evolution of the project. In a team environment, where collaboration is the key to success and collective productivity, semantic HTML becomes the common language that unites efforts, ensuring that everyone is on the same page, literally, and that the work of one complements the other in a harmonious way. It is the basis for more efficient software production that is less prone to rework, optimizing collective effort.</p> <h4>Microdata and Schema.org: Elevating Semantics to the Next Level of Connection</h4> <p>If semantic HTML is the basis for giving structural meaning to content, microdata and Schema.org are the next level of refinement, allowing you to add an even richer and more granular layer of context and intelligence to your data. Imagine you have a cake recipe on your site. With semantic HTML, you can mark the recipe title, the ingredients, and the preparation instructions. But what if you want search engines to know, unequivocally, that "30 minutes" is the preparation time, that "wheat flour" is a specific ingredient, and that "5 stars" is the average user rating for that recipe? That's where microdata and Schema.org come in, transforming raw data into structured knowledge.</p> <p>**Microdata** is an HTML5 specification that allows you to embed structured data directly into existing HTML using simple attributes like <code>itemscope</code>, <code>itemtype</code>, and <code>itemprop</code>. They work like small semantic labels that you attach to your content, describing the type of item (e.g., a recipe, a product, an event, a person) and its specific properties (preparation time, price, location, author's name). It is a way of "labeling" information so that machines can process and understand it without ambiguity.</p> <p>**Schema.org** is a collaborative vocabulary, a joint initiative of internet giants like Google, Microsoft, Yahoo, and Yandex, with the mission of creating and promoting a standardized set of schemas for structured data on the internet. It provides a vast collection of "types" (such as <code>Recipe</code>, <code>Product</code>, <code>Event</code>, <code>Person</code>, <code>Organization</code>, <code>LocalBusiness</code>) and "properties" (such as <code>prepTime</code>, <code>ingredients</code>, <code>price</code>, <code>location</code>, <code>name</code>, <code>description</code>) that you can use with microdata (or other formats like JSON-LD and RDFa) to describe your content in a universally understandable way. This standardization is crucial because it allows search engines to understand the meaning and relationships behind the entities on your page, which, in turn, can lead to the display of "rich snippets" or "rich results" in search results. Think of those search results that show rating stars, cooking time, price, images, or even the stock of a product directly in the listing – that's the power of Schema.org in action, making the information richer and immediately useful to the user, without the need to click and navigate through several pages.</p> <p>The impact of microdata and Schema.org on the front-end is profound and transformative. They elevate the developer from a mere "pixel painter" to a "knowledge architect". By implementing these structured data, you are not only making your site more visible; you are actively contributing to the construction of a "web of data" where information is interconnected, understandable by machines on a global scale, and therefore, more powerful. This opens doors to smarter applications, more effective voice assistants that can answer complex questions, and a richer and more personalized user experience, where information finds the user proactively and contextually, often before they even need to actively search. It is the materialization of the promise of the Semantic Web: an internet where data is not only displayed, but truly understood and used to serve humanity in more efficient and equitable ways.</p> <h4>The Front-end as a Catalyst for Collective Intelligence</h4> <p>The relationship between semantic HTML, microdata, and artificial intelligence is intrinsically symbiotic. AI and machine learning thrive on structured and contextualized data. The more meaning you incorporate into your HTML, the easier it becomes for AI algorithms to process, analyze, and infer information from your content. It's like providing a detailed and annotated map instead of an ambiguous sketch. With an accurate map, AI can chart routes, identify points of interest, make complex connections, and make much more accurate and useful predictions.</p> <p>The front-end developer, in this constantly evolving scenario, is not just an implementer of designs or an encoder of interfaces, but a true catalyst for the collective intelligence of the web. By choosing semantic tags and applying microdata, they are, in fact, "training" the web, teaching machines to understand the digital world in a deeper and more human way. This is an immense responsibility and an unprecedented opportunity to shape the future. We are building the foundations for an era where the web will not only be a passive repository of documents, but a vast network of interconnected knowledge, capable of reasoning, learning, and serving humanity in ways we have barely begun to imagine. It is our chance to shape the future of information, ensuring that it is accessible, understandable, and, above all, useful to everyone, not just those who hold the processing power or the most sophisticated algorithms.</p> <p>Resistance to adopting these practices, often justified by the "complexity" or the "extra time" they demand, is a dangerous shortsightedness. It is the short-term mentality that sacrifices the future in the name of immediate convenience, perpetuating disorganization and inefficiency. Those who embrace semantics are not just following a technological trend; they are investing in the resilience, visibility, and relevance of their projects in an increasingly intelligent and competitive digital ecosystem. They are, ultimately, contributing to a fairer and more equitable web, where knowledge is truly free and accessible to all, and not just to those who can decipher the chaos. The semantic revolution begins with every line of code we write, with every tag we choose, with every meaning we assign, paving the way for an internet that serves the collective.</p> <h3>Connecting Knowledge: RDF, OWL, and SPARQL – The Backbone of Connected Data</h3> <p>If semantic HTML and microdata are the language that allows machines to understand the meaning of specific content on a page, the Resource Description Framework (RDF), the Web Ontology Language (OWL), and SPARQL are the grammar and vocabulary that allow these machines not only to understand, but also to connect, reason, and query vast networks of distributed knowledge. We are talking about going beyond simply marking data on a page; we are entering the realm of building knowledge graphs, where the relationships between information are as crucial as the information itself. It is here that the promise of a truly intelligent web begins to materialize, transforming the internet from a collection of documents into a vast global database, where the front-end acts as the interpreter and viewer of this interconnected complexity.</p> <p>Think of the internet as a gigantic library. Semantic HTML helps us catalog each book, identify its title, author, publisher. But RDF, OWL, and SPARQL allow us to go further: they allow us to understand the relationships between the books. Which book is a sequel to another? Which authors are contemporaries? Which books address the same theme, even if from different publishers? This ability to understand relationships is what transforms isolated data into interconnected knowledge, and it is the basis for the intelligence we aim for on the web.</p> <h4>RDF: The Universal Language for Describing Resources</h4> <p>The Resource Description Framework (RDF) is the foundation upon which the Semantic Web is built. It provides a standard model for describing resources on the web, whether they are documents, people, places, events, or anything else that can be identified by a Uniform Resource Identifier (URI). The beauty of RDF lies in its simplicity and flexibility. It describes information in terms of "triples": subject-predicate-object. For example, the triple "John (subject) has as friend (predicate) Mary (object)" is an RDF statement. "Book X (subject) was written by (predicate) Author Y (object)".</p> <p>Each part of the triple is identified by a URI, which allows the information to be globally unique and interconnected. This means that the "John" from one database can be the same "John" from another, as long as both use the same URI to identify him. This ability to link data from different sources is revolutionary. It breaks down information silos, allowing heterogeneous systems to share and understand each other's data. For the front-end developer, this means that, instead of dealing with monolithic APIs and rigidly formatted data, they can access and integrate information from a distributed network of connected data, opening a universe of possibilities for creating richer and more dynamic applications. It is the promise of a web where information flows freely, without the constraints of proprietary formats or isolated databases, a true ecosystem of shared knowledge.</p> <h4>OWL: Defining Meaning and Relationships</h4> <p>If RDF gives us the structure to describe resources, the Web Ontology Language (OWL) gives us the power to define the meaning and relationships between these resources in a more expressive and formal way. An ontology, in this context, is a formal model of a domain of knowledge, specifying the types of things that exist in that domain and the properties and relationships that can exist between them. OWL allows developers to define classes (categories of things), properties (attributes and relationships), and constraints on these properties.</p> <p>For example, with OWL, we can define that "Person" is a class, that "hasChild" is a property that connects two "Persons", and that "hasChild" is a transitive property (if A has child B and B has child C, then A is grandfather of C). We can also define that "Professor" is a subclass of "Person" and that "Professor" "teaches" a "Subject". These formal definitions allow reasoning systems (reasoners) to infer new knowledge from existing data. If a system knows that "Maria is a Person" and "Maria hasChild John", and that "John is a Person", it can infer that "Maria is mother of John" (if the property "hasChild" is defined with that semantics). This inference capability is what truly elevates the Semantic Web to a level of intelligence, allowing machines not only to store and retrieve data, but also to reason about it.</p> <p>For the front-end, OWL is crucial because it ensures the consistency and validity of the data being consumed. By understanding the ontology behind the data, the developer can build interfaces that accurately reflect the underlying knowledge model, avoiding errors and ensuring that user interactions are semantically correct. Furthermore, OWL's inference capability can be leveraged to enrich the user experience, presenting information that was not explicitly provided, but that was logically deduced. It is the basis for building applications that not only display data, but that understand the world that these data represent, allowing for a more natural and intelligent interaction.</p> <h4>SPARQL: The Query Language for the Web of Data</h4> <p>With data described in RDF and ontologies defined in OWL, we need a way to query this vast network of information. That's where the SPARQL Protocol and RDF Query Language (SPARQL) comes in. SPARQL is to the Web of Data what SQL is to relational databases: a powerful query language that allows developers to retrieve and manipulate RDF data in a flexible and expressive way.</p> <p>With SPARQL, you can ask complex questions about the connected data. For example, you can ask: "What are all the books written by authors who were born in a certain country and that address the theme 'artificial intelligence'?" Or "What are all the people who are friends of John and who also work in the same company as Maria?". The ability to perform these complex queries on distributed data is what unlocks the true potential of the Semantic Web. For the front-end, this means that applications can become much more dynamic and responsive to user needs. Instead of relying on pre-defined APIs that return fixed datasets, the front-end can build custom SPARQL queries to fetch exactly the information needed, in real time, from multiple connected data sources. This allows for the creation of interactive dashboards, complex data visualization tools, and applications that dynamically adapt to the user's context.</p> <p>Learning SPARQL is a fundamental investment for the front-end developer who wants to navigate the era of intelligence. It offers unprecedented control over data retrieval, allowing the user interface to be powered by deep and interconnected knowledge, rather than just raw data. It is the tool that allows the front-end to transcend mere presentation and become a true explorer and orchestrator of knowledge on the web.</p> <h4>The Front-end as a Builder of Bridges to Knowledge</h4> <p>The intersection of RDF, OWL, and SPARQL with front-end development is where the magic of the Semantic Web really happens. The front-end, in this context, is not just the "face" of the application; it is the builder of bridges, the translator, and the viewer of a universe of connected data. It is responsible for transforming the complexity of RDF triples, OWL inferences, and SPARQL queries into user experiences that are intuitive, accessible, and meaningful.</p> <p>Imagine an application that helps researchers find relevant scientific articles. With the Semantic Web, this application would not only search for keywords, but would understand the relationships between authors, institutions, research topics, and even the methods used. The front-end could visualize these relationships in an interactive graph, allowing the researcher to navigate the knowledge in a much richer and more intuitive way. They could, for example, click on an author and see all their co-authors, their most cited publications, and the research areas in which they work, all interconnected through RDF data and queried via SPARQL.</p> <p>Another example would be a product recommendation system. Instead of just recommending products based on previous purchases, a semantic system could understand the characteristics of the products (material, color, style), the user's preferences (allergies, lifestyle), and the relationships between these attributes. The front-end would present recommendations that are not only "popular", but semantically relevant and personalized, justifying the choices based on inferred knowledge. This elevates the user experience from a simple transaction to an intelligent and contextualized interaction.</p> <p>The challenges for the front-end developer are real: dealing with the complexity of knowledge graphs, optimizing the performance of SPARQL queries on large volumes of data, and designing interfaces that effectively communicate semantic relationships. However, the opportunities are even greater. Those who master these technologies will be at the forefront of building the next generation of web applications, applications that not only display information, but that understand, reason, and interact with the world in an intelligent way. They will be building tools that empower users with knowledge, break down information barriers, and contribute to a more informed and connected society.</p> <p>The Semantic Web, with its foundations in RDF, OWL, and SPARQL, is not just a technical evolution; it is a revolution in the way information is organized and accessed. It represents a crucial step towards a web where knowledge is a truly shared resource, where machines serve as extensions of our cognitive capacity, and where the front-end is the portal to this new era of intelligence. It is the materialization of the vision of a web that not only connects us to each other, but that connects us to the very fabric of knowledge, allowing humanity to advance collectively, overcoming the limitations of fragmented and isolated information.</p> <h3>Intelligence at the Edge: AI, Machine Learning, and the Personalization of the Semantic Experience</h3> <p>If the Semantic Web provides us with the structure and language for machines to understand the meaning of data, Artificial Intelligence (AI) and Machine Learning (ML) are the engines that drive this understanding into action, transforming raw data into actionable insights and deeply personalized experiences. We are witnessing the dawn of an era where the web is not just a repository of information, but an intelligent ecosystem that learns, adapts, and anticipates user needs. At the heart of this transformation, the front-end emerges as the stage where this intelligence manifests itself, translating complex algorithms and data inferences into intuitive and engaging interactions. It is the materialization of the promise of a web that not only responds, but proactively serves the individual, elevating the digital experience to an unprecedented level.</p> <p>For a long time, personalization on the web was rudimentary, based on simple rules or browsing history. "If the user visited page X, show product Y." This approach, although functional, lacked depth and nuance. With the integration of AI and ML, powered by semantic data, personalization transcends the mere display of relevant content; it becomes predictive, contextual, and adaptive. The machine not only knows what you have done, but begins to infer what you *probably* will do, what you *really* need, and even what you *desire* before you even express it. This ability to anticipate, previously restricted to science fiction, is now a tangible reality, and the front-end is the point of contact where this magic happens.</p> <h4>The Synergy: How AI and ML Consume Semantic Data</h4> <p>The effectiveness of AI and ML algorithms critically depends on the quality and structure of the data with which they are trained and operate. Semantic data, with its richness of context and explicit relationships, is the ideal food for these systems. While unstructured data requires a massive effort of pre-processing and inference by the AI, semantic data (RDF, OWL, Schema.org) arrives "pre-digested", with the meaning already encoded. This speeds up the learning process and improves the accuracy of predictions and recommendations.</p> <p>Imagine a movie recommendation system. Without semantic data, AI can only correlate watched movies with similar movies (based on genres or actors). With semantic data, AI can understand that "Movie A" is a "science fiction drama" directed by "Director X", that "Director X" is known for "exploring themes of existentialism", and that "Movie B" is also a "science fiction drama" that "explores themes of existentialism", even if it does not have the same actors or directors. This deep understanding of relationships and attributes allows AI to make much more sophisticated and personalized recommendations, going beyond superficial associations. It can, for example, recommend a philosophy book that complements the theme of a movie, or a documentary about the director's life, because the semantic network connects these different types of resources.</p> <p>In addition, AI can use semantic data to understand the user's intent in search queries or interactions with voice assistants. If a user asks "What's the weather forecast for tomorrow?", the AI, based on ontologies of time and location, can infer that "tomorrow" refers to the following date and that the default location is the user's, or the last searched. This ability to infer and contextualize intent is fundamental to building conversational interfaces and intelligent assistants that seem to "understand" the user in an almost human way. The front-end, in this scenario, not only displays the answer, but can visualize it intuitively, perhaps with temperature graphs, weather condition icons, and interactive maps, all powered by semantic intelligence.</p> <h4>Predictive and Adaptive Personalization: The New Paradigm of User Experience</h4> <p>Personalization driven by AI and semantic data goes far beyond simple user segmentation. It manifests itself in various forms, each elevating the user experience to a new level:</p> <p><ul> <li>**Contextual Recommendations:** Not just "you might like this", but "you might like this NOW, because it's raining and you searched for thriller movies last night". AI uses the semantic context (weather, location, recent history, inferred preferences) to offer recommendations that are incredibly relevant in the present moment. For the front-end, this means the ability to render dynamic components that adapt in real time, displaying suggestions for content, products, or services that deeply resonate with the user's current state and implicit needs.</li> <li>**Adaptive Interfaces:** The user interface can adapt dynamically based on the user's semantic profile and behavior. A user with visual impairment may have the contrast and font size adjusted automatically, or a user who prefers voice interactions may have the audio controls more prominent. AI, informed by semantic data about accessibility and user preferences, can optimize the interface for each individual, making the experience more inclusive and efficient. The front-end, with modern frameworks, can be built to be inherently adaptable, responding to these intelligent signals.</li> <li>**Semantic and Conversational Search:** Search ceases to be a simple keyword match and becomes a conversation. The user can ask questions in natural language, and the AI, using ontologies and knowledge graphs, understands the intent and returns accurate and contextual results, often presented in a summarized or visually rich way. The front-end is the translator of this conversation, displaying the results clearly, allowing refinements and additional interactions, transforming the search into a journey of knowledge discovery.</li> <li>**Dynamic Content Generation:** In some cases, AI can even generate content dynamically, such as summaries of articles, product descriptions, or answers to frequently asked questions, based on semantic data. The front-end then presents this generated content smoothly and integrated into the interface, creating an experience that seems to be custom-made for the user.</li> </ul></p> <p>This predictive and adaptive personalization is not just a convenience; it is a powerful tool to combat information overload and to ensure that the user receives what is most relevant to them, at the right time. It is the promise of a web that not only informs, but that truly serves the individual, optimizing their time and their ability to make informed decisions.</p> <h4>The Role of the Front-end in Visualizing and Interacting with Intelligent Systems</h4> <p>The front-end developer is the architect of the user experience, and in the era of intelligence, their role becomes even more critical and complex. They are responsible for translating the complexity of AI models and semantic knowledge graphs into interfaces that are not only functional, but also intuitive, understandable, and enjoyable to use. This requires a set of skills that goes beyond traditional HTML, CSS, and JavaScript.</p> <p><ul> <li>**Semantic Data Visualization:** Presenting knowledge graphs, inferred relationships, and complex data in an understandable way is a design and development challenge. The front-end needs to master data visualization libraries (such as D3.js, Three.js, or graph visualization frameworks) to create visual representations that allow the user to explore and understand the semantic connections. This may include interactive knowledge maps, network visualizations, or dashboards that show how AI arrived at a particular recommendation.</li> <li>**Conversational and Voice Interfaces:** With the rise of voice assistants and chatbots, the front-end needs to design and implement interfaces that facilitate interaction in natural language. This involves not only rendering text, but managing the flow of the conversation, displaying contextual suggestions, and integrating with Natural Language Processing (NLP) and Natural Language Generation (NLG) APIs that are powered by semantic data.</li> <li>**AI Feedback and Transparency:** For users to trust intelligent systems, they need to understand how AI is making decisions. The front-end has a crucial role in providing clear and transparent feedback on AI inferences. This can be done through visual explanations, "why this recommendation?" or "how was this answer generated?", making the AI process less of a "black box" and more of a collaborative partner.</li> <li>**Performance Optimization for AI:** Applications that rely on AI and ML can be resource-intensive. The front-end needs to optimize data loading, component rendering, and interaction with AI APIs to ensure a smooth and responsive experience, even on devices with limited resources. This may involve techniques such as lazy loading, image optimization, and efficient use of Web Workers.</li> <li>**Human-Centered Design and AI Ethics:** The front-end developer is on the front lines of human interaction with AI. It is their responsibility to ensure that interfaces are designed ethically, avoiding algorithmic biases, protecting user privacy, and ensuring that personalization does not become manipulation. Understanding human cognitive biases is fundamental to designing interfaces that guide the user ethically and transparently, without exploiting vulnerabilities.</li> </ul></p> <p>Artificial Intelligence and Machine Learning, when fueled by a robust semantic foundation, have the potential to radically transform the user experience on the web, making it more intuitive, personalized, and efficient. The front-end, in this scenario, is not a mere executor, but a true architect of intelligence, responsible for shaping how humans interact with machines and how knowledge is accessed and used. It is a field of vast opportunities and responsibilities, where human creativity joins computational power to build a web that not only connects us, but empowers us and serves us in ways that were once just dreams. It is the materialization of a web that truly understands the world and its inhabitants, a crucial step towards building a more intelligent and equitable digital society.</p> <h3>Decentralization and Semantics: The Front-end in Web 3.0 and the Power of the User</h3> <p>For decades, the internet we know has flourished under a predominantly centralized model. Technology giants have erected digital empires, concentrating vast amounts of data, computing power, and, consequently, influence over information and human interaction. Our data, our digital identities, our communications – all of this has come to reside on servers controlled by a few corporations, often transforming us from users into products. This centralization, while facilitating the initial growth and scalability of the web, has created a power imbalance, raising urgent questions about privacy, censorship, and the true ownership of our digital footprint. But, as in every structure of concentrated power, the opposing force arises, the search for autonomy and distribution. It is in this context that Web 3.0 emerges, not as a mere technological update, but as a revolutionary movement that seeks to redefine the relationship between the individual and the network, and the front-end is the portal to this new era of empowerment.</p> <p>The promise of Web 3.0 is that of an internet where control does not reside in a handful of entities, but is distributed among its participants. A web where users are the true owners of their data, where information is resistant to censorship, and where collaboration is incentivized by transparent and fair mechanisms. This vision, once utopian, is becoming a reality thanks to technologies like blockchain and decentralized data protocols. And, surprisingly, the Semantic Web, with its emphasis on the organization and meaning of data, is not only compatible with this decentralized vision; it is an essential component for Web 3.0 to reach its maximum potential, transforming the mere decentralization of infrastructure into a true decentralization of knowledge and power.</p> <h4>The Problem of Centralization: A Call to Action for the Front-end</h4> <p>The centralized web, although it has connected us in unimaginable ways, has brought with it a series of problems that are now becoming unsustainable. Our personal data, from browsing history to financial and health information, is collected, stored, and monetized by companies that operate in information silos. This concentration of data not only makes us vulnerable to leaks and cyber attacks, but also deprives us of control over our own digital identity. Censorship, whether governmental or corporate, becomes a real threat, where platforms can arbitrarily remove content or ban users, stifling freedom of expression and access to information. In addition, the opacity of centralized algorithms creates "filter bubbles" and echo chambers, limiting our exposure to diverse perspectives and subtly manipulating our perceptions.</p> <p>For the front-end developer, who has traditionally been the builder of the interface for these centralized systems, this reality imposes an ethical dilemma. We are the architects of the user experience, but this experience has been, in large part, mediated and controlled by third parties. Web 3.0 emerges as an opportunity to reverse this dynamic, to build interfaces that not only connect users to services, but that empower them to own, control, and monetize their own data and creations. It is a call for the front-end to become an agent of liberation, building the gateways to a more democratic and transparent internet.</p> <h4>Web 3.0 and Blockchain: The Foundation of Decentralization</h4> <p>At the heart of Web 3.0 is blockchain technology, a distributed and immutable ledger that allows the secure and transparent recording of transactions and data without the need for a central authority. Each "block" of information is cryptographically linked to the previous one, forming a chain that is verified and maintained by a network of computers, not by a single server. This distributed architecture gives the blockchain revolutionary properties: **immutability** (once recorded, data cannot be changed), **transparency** (all transactions are visible to all participants in the network), and **censorship resistance** (there is no central point of failure or control that can be turned off or manipulated).</p> <p>In addition to the blockchain, other decentralized protocols, such as the InterPlanetary File System (IPFS) for file storage and distributed computing networks, complement the Web 3.0 infrastructure. Together, they form the basis for creating applications that do not depend on centralized servers, but operate on a peer-to-peer network, where data is stored in a distributed manner and control is shared. This fundamental change in the architecture of the internet has profound implications for the front-end, which now needs to interact with these new paradigms of data and business logic.</p> <h4>The Confluence: Semantic Web Meets Web 3.0</h4> <p>The Semantic Web, with its goal of making data understandable to machines through structures like RDF, OWL, and Schema.org, is the ideal partner for Web 3.0. While Web 3.0 offers the infrastructure for decentralization and data ownership, the Semantic Web provides the meaning and context that transform this raw data into actionable knowledge. Without semantics, decentralized data would be just a jumble of information without a clear meaning for machines, limiting its potential for interoperability and intelligence.</p> <p>Imagine health data stored in a decentralized manner on a blockchain, with the patient having full control over who can access it. If this data is semantically enriched (using medical ontologies, for example), it can be easily understood by different health systems, researchers, and even AI algorithms, without the need for complex API integrations or format conversions. The patient not only owns their data, but this data is intelligible and useful in any context, maximizing its value and potential. This synergy between decentralization and semantics is what truly unlocks the power of data, transforming it from assets controlled by corporations into resources that directly serve the individual and the collective.</p> <h4>dApps and the Front-end: Building the New Frontier</h4> <p>Decentralized applications, or dApps, are the Web 3.0 equivalent of traditional applications. The main difference is that the business logic and data storage of a dApp reside on a blockchain (through "smart contracts") and on decentralized storage networks, instead of centralized servers. For the front-end developer, this means a fundamental change in how applications are built and interact with the "backend".</p> <p>The front-end of a dApp is still built with standard web technologies (HTML, CSS, JavaScript, frameworks like React, Vue, Angular). However, instead of making HTTP requests to a centralized API, it interacts directly with the blockchain through libraries like Web3.js or Ethers.js. These libraries allow the front-end to call smart contract functions, read data from the blockchain, and sign transactions (with the help of digital wallets like MetaMask). File storage, such as images or documents, is often done on networks like IPFS or Arweave, and the front-end is responsible for interacting with these protocols to load and display the content.</p> <p>The challenges for the front-end are significant: <ul> <li>**New Tools and Paradigms:** Learning to interact with blockchains, smart contracts, and decentralized storage systems requires a new mindset and a new set of tools.</li> <li>**User Experience (UX) for Decentralization:** The UX of dApps can be complex. Blockchain transactions require user confirmations, can take time to process, and involve costs (gas fees). The front-end needs to design interfaces that clearly communicate these processes, manage loading states, and provide transparent feedback to the user.</li> <li>**Security:** Security is paramount. As the front-end interacts directly with the blockchain and user wallets, any vulnerability can have direct financial consequences.</li> <li>**Performance:** Reading data from the blockchain can be slower than from a centralized database. The front-end needs to optimize how data is fetched and cached to ensure a smooth experience.</li> </ul></p> <p>However, the opportunities are immense. The front-end can build applications that offer true digital ownership (NFTs), decentralized governance systems (DAOs), peer-to-peer markets without intermediaries, and censorship-resistant social networks. By integrating semantic data with these dApps, the front-end can create experiences that are not only decentralized, but also intelligent and contextually rich, allowing users to control and monetize their own data in a meaningful way.</p> <h4>User Empowerment and Data Sovereignty: The Front-end Revolution</h4> <p>The convergence of the Semantic Web and Web 3.0 represents a fundamental shift in the paradigm of data ownership and control. Instead of being mere consumers of services and data providers for corporations, Web 3.0 allows us to be]]></content:encoded>
      <pubDate>Mon, 16 Jun 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>Web3</category>
      <category>Futuro do Trabalho</category>
      <category>Ética na Tecnologia</category>
      <category>Soberania Digital</category>
      <category>Acessibilidade</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/4ca653a2-158c-4e58-983c-07a33fc82d2f" medium="image"/>
    </item>
    <item>
      <title>Open Source</title>
      <link>https://www.mmlo.app/en/blog/open-source</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/open-source</guid>
      <description>The Collaborative Utopia That Capitalism Is Trying to Co-opt</description>
      <content:encoded><![CDATA[<p>In the vast and dynamic universe of technology, few ideas resonate with the force and idealism of the Open Source movement. Born from a yearning for freedom, collaboration, and unrestricted sharing of knowledge, it has risen as a true digital utopia, promising to democratize access to technology and build a future where code is a common good. However, like a mighty river encountering the dams of capital, this current of innovation and solidarity has been incessantly harassed, shaped, and, in many cases, co-opted by the gears of the capitalist system. This is the story of an ideal that defies the logic of ownership, but which, paradoxically, has become an indispensable tool for those who seek profit above all else.</p><h2>The Collaborative Genesis: The Original Dream of Open Source</h2><p>To understand the complexity of Open Source today, it is imperative to return to its roots, to a time when collaboration and sharing were the norm in the nascent world of computing. In the 1950s and 1960s, technological development, especially in academic and research environments such as ARPANET (the precursor to the modern internet), was intrinsically collaborative and open. Researchers and developers exchanged source code, reviewed each other's work, and built upon each other's innovations without the constraints of private ownership. It was an ecosystem where knowledge flowed freely, driven by curiosity and the collective desire for advancement.[1](#__citation:https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHQqQVd1ffAdkDYH30ALbyumRu6H-vNnX1aeAYdHl2_bu2zXSFN9K9pGavb1peLt1Z6af1NCAmul0hablEqqHIptzClSmXpopm3cgPhjeJ-GyWE4vpSrt8nD4UDRiZHNfFn5c7K6tdmlLk_TLWNINr2UEknSqISma9HFJa8IfE= )</p><p>This golden age of collaboration began to erode with the rise of proprietary software in the 1970s, and more intensely in the 1980s. Companies like Microsoft and Apple began to see software as a product to be sold, with its source code kept secret, protected by copyrights and patents. This commodification of knowledge generated a reaction. Richard Stallman, a seminal figure in this movement, coined the term "Free Software" and founded the Free Software Foundation (FSF) in 1985. For Stallman, user freedom was paramount: the freedom to run the program for any purpose, the freedom to study how the program works and adapt it to your needs, the freedom to redistribute copies, and the freedom to distribute copies of your modified versions.[2](#__citation:...)</p><p>The philosophy of Free Software, therefore, was not based on gratuity, but on freedom. It was an ethic deeply anti-capitalist in its essence, challenging the notion of intellectual property and seeking to decommodify software. The source code was seen as knowledge to be shared, a common good, and not a commodity to be controlled and sold.</p><h2>The Logic of Capital and Appropriation: When Utopia Becomes a Tool</h2><p>The transition from "Free Software" to "Open Source" was not just a change in nomenclature; it was a deliberate strategy to make the model attractive to the corporate world. And it worked. Companies that once viewed open source with suspicion began to realize the immense value it could add to their businesses.</p><p>The appropriation of Open Source by capitalism manifests itself in several ways. First, cost reduction is an undeniable attraction. Using open-source software eliminates the need to pay expensive licenses for proprietary products.</p><h2>The Alienation of the Developer in the Capitalist Open Source Ecosystem</h2><p>Developers, often driven by passion, idealism, and the desire to contribute to a common good, dedicate hours of their time and intellect to Open Source projects. However, when these projects are co-opted by large corporations, the nature of the work can transform.</p><p>The "freedom" to contribute can mask a form of exploitation. The immaterial work of the developer is appropriated and monetized by companies that profit without necessarily remunerating the original creators.</p><h2>Resistance and Contradictions: Open Source as a Battlefield</h2><p>Despite the undeniable co-option, Open Source is not a passive monolith. It is configured as an ideological and practical battlefield, where resistance and contradictions are constant.</p><p>Copyleft licenses, vigilant communities, and alternative funding models represent real forms of resistance to the privatization of knowledge.</p><h2>Towards a Digital Utopia? The Revolutionary Potential of Open Source</h2><p>Despite the tensions, Open Source carries a revolutionary potential by decommodifying knowledge and challenging the logic of intellectual property. It offers a glimpse of an alternative social organization, based on collaboration and solidarity.</p><h3>References</h3>

<p>
<a href="https://www.gnu.org/philosophy/free-software-even-more-important.html" target="_blank" rel="noopener noreferrer">
Stallman, Richard. “Free Software Is Even More Important Now”. GNU Project.
</a>
</p>

<p>
<a href="https://opensource.org/osd" target="_blank" rel="noopener noreferrer">
Open Source Initiative. “The Open Source Definition”.
</a>
</p>

<p>
<a href="https://www.catb.org/~esr/writings/cathedral-bazaar/" target="_blank" rel="noopener noreferrer">
Raymond, Eric S. “The Cathedral and the Bazaar”.
</a>
</p>

<p>
<a href="https://www.gnu.org/philosophy/open-source-misses-the-point.html" target="_blank" rel="noopener noreferrer">
Stallman, Richard. “Why Open Source Misses the Point of Free Software”.
</a>
</p>

<p>
<a href="https://www.redhat.com/en/topics/open-source/what-is-open-source" target="_blank" rel="noopener noreferrer">
Red Hat. “What is Open Source?”.
</a>
</p>]]></content:encoded>
      <pubDate>Fri, 13 Jun 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Open Source</category>
      <category>Tecnologia e Sociedade</category>
      <category>React</category>
      <category>Forças Produtivas</category>
      <category>Futuro do Trabalho</category>
      <category>Economia de Plataforma</category>
      <category>Ética na Tecnologia</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/5b548a52-37ea-4872-8151-a4dfbdce8251" medium="image"/>
    </item>
    <item>
      <title>Engineers of Freedom or Architects of the Prison?</title>
      <link>https://www.mmlo.app/en/blog/engenheiros-da-liberdade-ou-arquitetos-da-prisao</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/engenheiros-da-liberdade-ou-arquitetos-da-prisao</guid>
      <description>The Front-End Developer&apos;s Dilemma in the Age of Surveillance</description>
      <content:encoded><![CDATA[<p>
	Comrades, the digital age promised us freedom and connection, but, as with
	every promise under capital, it conceals chains. We, front-end developers,
	are the architects of the interfaces that shape the human experience on the
	web. But what is our role when these same interfaces become tools of
	surveillance and social control? This is the dilemma that confronts us: are
	we engineers of freedom or, inadvertently, architects of the digital prison?
</p>

<h3>1. The Invisible Web: How Front-end Becomes a Surveillance Tool</h3>
<p>
	The internet, this vast network of information and interactions, is built
	layer by layer. At the heart of the user experience, where the digital meets
	the human, lies the front-end. It is here that HTML structures the content,
	CSS styles it, and JavaScript brings interactivity to life. But, under the
	logic of surveillance capitalism, this same interactivity and the relentless
	pursuit of "improving the user experience" become fertile ground for massive
	data collection and incessant monitoring. The front-end, which should be the
	gateway to digital freedom, often transforms into the invisible web that
	imprisons us.
</p>
<p>
	Digital surveillance, in essence, is the monitoring and collection of data
	about the online activities of individuals and groups, carried out by
	companies, governments, and even cybercriminals, with the aim of understanding
	behavior patterns, predicting future actions, or controlling the information
	that reaches the public. We, front-end developers, are the builders of the
	tools that enable this collection. From the simple click to the time spent on
	a page, every interaction is potential data. User tracking tools, such as
	Google Analytics and Hotjar, are widely used to collect, store, and analyze
	user behavior, including page views, acquisition channels, and session
	durations. These tools, often integrated directly into the front-end code,
	allow companies to understand how users interact with their sites, identify
	bottlenecks, and optimize the customer journey.
</p>
<p>
	Data collection in the front-end goes far beyond cookies, which, although still
	relevant, can be blocked or deleted. "Browser fingerprinting" is a more
	sophisticated method of identifying users, collecting information such as
	browser type and version, operating system, screen size and resolution, fonts
	in use, and language and time settings. This information, when combined, forms
	a unique "digital fingerprint" that allows tracking of online activity without
	the need for cookies. The front-end developer, when using libraries and
	frameworks, often incorporates, without full awareness, third-party scripts
	that facilitate this collection. The HTTP Web Almanac reveals that most sites
	(about 95%) use third-party content, which may include tracking scripts. These
	scripts, although not essential for site development, are widely used for
	analytics, advertising, and integration with social networks, transmitting
	information such as the page URL and the user's IP address.
</p>
<p>
	The central issue here is not the mere collection of data to improve
	usability—a legitimate goal in itself. The problem arises when this collection
	is instrumentalized for purposes of social control and profit maximization,
	transforming the human experience into "behavioral surplus value." Shoshana
	Zuboff, in her work on "Surveillance Capitalism," argues that Big Tech and
	other companies use information and communication technologies (ICT) to
	expropriate human experience, which becomes raw material processed and
	commodified as behavioral data. The user often freely cedes their information
	by agreeing to complex terms of use or by using "free" services.
</p>
<p>
	The interface we build, with its buttons, forms, and navigation flows, can be
	designed to induce the user to make decisions that benefit the company, not
	the individual. "Dark Patterns" are clear examples of this: intentionally
	deceptive design tactics that manipulate user behavior, leading them to perform
	unwanted actions, such as subscribing to services, providing personal
	information, or accepting terms unconsciously. Examples include
	"confirmshaming" (using phrases to shame the user), disguised ads, hidden
	subscriptions, and fake timers that create urgency. These practices exploit
	cognitive and psychological vulnerabilities, compromising the user's autonomy
	and freedom of choice. The front-end developer, by implementing these
	patterns, becomes an accomplice, consciously or not, in the architecture of
	this digital prison. The proliferation of standardized interface component
	libraries has facilitated the dissemination of these practices, allowing
	developers to incorporate manipulation techniques without necessarily
	understanding the ethical implications of their design choices.
</p>
<p>
	Technology, under capitalism, is not neutral; it is a political instrument.
	Great scientific and technological discoveries, instead of liberating, often
	perfect the instruments of domination and exploitation. The front-end, in this
	context, is the front line of this domination, where the interface becomes the
	point of contact for data extraction and behavioral manipulation. The
	"algorithmization" of society, driven by algorithms that capture, store, and
	cross-reference data, creates an arrangement of dependence and social control,
	where the individual is monitored perpetually. The goal is not explicit
	coercion, but the subtle motivation for the user to hand over their data and
	remain engaged, without realizing the exploitation.
</p>
<p>
	The responsibility of the front-end developer, therefore, transcends mere
	functionality and aesthetics. It extends to the ethics and social impact of
	each line of code. The choice of frameworks, libraries, and design patterns is
	not just a technical decision; it is a political decision. Popular frameworks
	like React, Angular, and Vue.js, although powerful for building interactive
	interfaces, can also be vehicles for the implementation of surveillance logics
	if they are not used with critical awareness. The security of usability, which
	aims to build a relationship of trust with the user, is easily broken when
	manipulation and surveillance become the hidden objective.
</p>
<p>
	The history of information security, from ancient cryptography to modern
	cybersecurity, has always been linked to data protection and privacy. However,
	in the digital age, the scale of data collection and the sophistication of
	surveillance tools have reached unprecedented levels. The General Data
	Protection Law (LGPD) in Brazil and the GDPR in the European Union are
	attempts to regulate this collection and protect the privacy of individuals,
	requiring consent and transparency. However, the mere existence of laws does
	not guarantee freedom; the fight for privacy and against surveillance is
	continuous and requires the conscious action of everyone, especially those who
	build the digital tools. The dilemma is clear: will we be accomplices of the
	prison or engineers of freedom? The answer lies in every line of code we
	write.
</p>

<h3>2. Surveillance Capitalism and the Expropriation of Human Experience</h3>
<p>
	To understand the depth of the front-end developer's dilemma, it is imperative
	to delve into the nature of "Surveillance Capitalism," a concept coined by
	Shoshana Zuboff that unveils the economic logic behind massive data
	collection. It is not just about targeted advertising or "improving the user
	experience," but about a new form of capital accumulation based on the
	expropriation of human experience. From this perspective, every interaction,
	every click, every mouse movement, every time spent on a page, every scroll
	that the user makes on an interface that we, front-end developers, build, is
	transformed into "raw material" for the production of "behavioral surplus
	value."
</p>
<p>
	In traditional capitalism, surplus value is extracted from the living labor of
	the worker in the production of material goods. In surveillance capitalism, the
	"raw material" is human life itself, transformed into behavioral data. Human
	experience is captured, analyzed, and predicted to be sold in the behavioral
	futures market. Technology companies, the Big Techs, become the new feudal
	lords, not of lands, but of our data and, consequently, of our future. The
	front-end is the battlefield where this expropriation occurs. It is the
	interface that, subtly and often imperceptibly, collects the data that fuels
	this system.
</p>
<p>
	Let's think about the frameworks and libraries we use daily: React, Angular,
	Vue.js. They are powerful tools that allow us to build complex and dynamic
	interfaces. However, the way they are conceived and the business logic that
	drives them often already incorporates the premise of data collection. For
	example, the ease of integration with third-party APIs for analyzing user
	behavior or for displaying personalized ads. The developer, when implementing
	an interface component, may be unknowingly opening a door for data collection
	that will be used for surveillance purposes.
</p>
<p>
	The "algorithmization" of society is a process that intensifies this
	expropriation. Algorithms, which are the heart of digital applications, are
	designed to capture, store, and cross-reference data from various groups of
	people. This mechanism allows companies to offer products and services in a
	highly targeted way, generating more profits. The front-end is the visible
	face of these algorithms, the interface through which they interact with the
	user and collect the data necessary for their operation. What appears to be a
	"personalization" of the experience is, in fact, a form of control and
	manipulation, where the user is subtly guided to consume, to interact in
	certain ways, to reveal more about themselves.
</p>
<p>
	The logic behind this is the relentless pursuit of predictability. In
	surveillance capitalism, the goal is to predict and shape human behavior to
	ensure economic results. This is done through the analysis of large volumes of
	behavioral data, which are transformed into "prediction products." The
	front-end, by capturing every nuance of user interaction, provides the inputs
	for these predictive models. The developer, by optimizing the performance of
	an interface or by implementing a new feature, may be unintentionally
	contributing to the efficiency of this mechanism of prediction and control.
</p>
<p>
	The illusion of "free services" is one of the pillars of this system. The user
	"freely cedes" their information by agreeing to extensive and complex terms of
	use, which few people read. This "free" access is, in fact, an unequal
	exchange: access to a service in exchange for the expropriation of their
	experience and their data. The front-end, by presenting these terms of use in
	an obscure way or by facilitating "consent" with a single click, becomes an
	accomplice in this unbalanced transaction.
</p>
<p>
	Ethics in front-end development, therefore, cannot be an afterthought. It must
	be intrinsic to every step of the process. Awareness of "Dark Patterns" is a
	crucial step. These manipulative design tactics, which aim to deceive the user
	so that they make decisions contrary to their interests, are the most explicit
	manifestation of the expropriation of human experience in the front-end. By
	implementing them, the developer is not just "optimizing conversion"; they are
	actively participating in manipulation and control.
</p>
<p>
	The responsibility of the front-end developer is immense. We are the builders
	of the interfaces that define people's relationship with the digital world. If
	we do not question the logic behind the functionalities we implement, if we do
	not refuse to build interfaces that manipulate and surveil, we will, in fact,
	be architecting the prison. The fight against surveillance capitalism begins in
	every line of code, in every design decision, in every refusal to be a mere
	executor of orders that aim at the expropriation of human experience. It is a
	call to action, to class consciousness, and to solidarity with the users, who
	are, ultimately, the expropriated.
</p>

<h3>3. Dark Patterns: The Engineering of Manipulation in the Front-end</h3>
<p>
	In the vast and complex universe of front-end development, where creativity
	and functionality intertwine to shape the user experience, an insidious shadow
	emerges: "Dark Patterns." Far from being mere design errors or usability
	failures, these are deliberate tactics, ingeniously designed to manipulate
	user behavior, inducing them to make decisions that, more often than not,
	benefit the company at the expense of their own interests. For us, front-end
	developers, understanding the mechanics and ethical implications of Dark
	Patterns is a moral and professional imperative.
</p>
<p>
	The rise of Dark Patterns reflects a silent metamorphosis in the relationship
	between technology and humanity. If in the early decades of computing the
	interfaces were raw and functional, designed by engineers for other
	engineers, today they are sophisticated psychological traps. Usability, once a
	primary goal, has been perverted to become a tool of persuasion and control.
	The front-end, which should be the facilitator of interaction, has become the
	stage for a psychological war, where every click, every scroll, every form
	filled can be a victory for the logic of surveillance capital.
</p>
<p>
	There are several types of Dark Patterns, each with its specific strategy to
	deceive the user. "Confirmshaming," for example, uses phrases that shame the
	user for not accepting an offer or option, such as "No, I don't want to save
	money" instead of a simple "No, thank you." This tactic exploits the aversion
	to guilt and the desire for social conformity, pressuring the user to act
	against their will. Another example is the "Roach Motel," where it is easy to
	enter a service, but extremely difficult to leave, with complex cancellation
	processes hidden in mazes of menus. This generates frustration and
	distrust, but guarantees the forced retention of users.
</p>
<p>
	"Sneak into Basket" is a common tactic in e-commerce, where additional items
	are automatically included in the shopping cart without the user's explicit
	consent. The user only notices the inclusion at the time of payment, often too
	late or with the pressure to finalize the purchase. "Misdirection" diverts the
	user's attention to irrelevant information, hiding more advantageous or
	important options. For example, the option not to purchase travel insurance
	may be hidden in a complex drop-down menu. These practices are not accidental;
	they are the result of intentional design, often driven by "conversion" and
	"engagement" metrics that prioritize profit above ethics.
</p>
<p>
	For us, front-end developers, the implementation of Dark Patterns is an act of
	betrayal of user trust. It is the instrumentalization of our technical skills
	for purposes of manipulation. Popular web development frameworks include
	pre-built components that can facilitate the implementation of these patterns,
	allowing developers to incorporate them without necessarily understanding the
	ethical implications of their design choices. This underscores the need for a
	critical awareness and continuous ethical education in our field.
</p>
<p>
	The negative impact of Dark Patterns is multifaceted. For users, they result in
	frustration, confusion, and distrust, harming autonomy and freedom of choice.
	For companies, although they may generate short-term gains, they damage
	reputation and brand trust in the long term. In addition, regulations such as
	the GDPR in the European Union and the LGPD in Brazil are increasingly
	combating these practices, imposing fines and legal sanctions. This
	demonstrates that society, albeit slowly, is reacting to this form of digital
	exploitation.
</p>
<p>
	The responsibility to avoid Dark Patterns falls on designers and developers. It
	is essential to understand the different types of dark patterns and the
	negative effects they can have on the user experience. Transparency,
	simplification of processes, and collection of user feedback are ethical
	practices that should replace manipulation. The education of the development
	team about the ethical implications of design is fundamental to ensure that
	technology is used to empower, not to control.
</p>
<p>
	We, front-end developers, have the power to resist. We can refuse to implement
	Dark Patterns. We can advocate for transparent and ethical designs. We can
	educate our colleagues and employers about the dangers and moral implications
	of these practices. The fight against manipulation in the front-end is a fight
	for user autonomy, for human dignity in the digital environment, and,
	ultimately, for the construction of an internet that serves the interests of
	the collective, and not just the profits of capital. It is an act of
	resistance that begins in every line of code, in every design decision, in
	every refusal to be an architect of the prison.
</p>

<h3>
	4. The Role of the Front-end Developer in Resistance and in Building Ethical
	Interfaces
</h3>
<p>
	Faced with the dark scenario of digital surveillance and manipulation through
	Dark Patterns, the front-end developer cannot remain neutral. Neutrality, in
	this context, is an illusion that serves the interests of capital. Our work,
	which shapes and gives life to digital interfaces, carries an immense
	responsibility. We are the architects of the user experience, and it is
	precisely in this strategic position that our power of resistance and
	transformation lies. The question is no longer "what can we build?", but "what
	should we build?" and "how can we build for emancipation, and not for
	domination?".
</p>
<p>
	The front-end developer is the professional who translates the design and
	vision of a project into an interactive and attractive interface. We use HTML,
	CSS, and JavaScript to bring web pages to life, ensuring they are responsive
	and optimized. However, this technical ability must be accompanied by a deep
	ethical awareness. Ethics in the digital world involves maintaining the
	dignity, security, and privacy of users. This means that every design and
	implementation decision must be guided by principles that protect the autonomy
	of the individual.
</p>
<p>
	The first line of defense is the conscious refusal to implement Dark Patterns.
	As we have seen, these tactics are designed to deceive and manipulate. An
	ethical front-end developer must identify and question any request that
	resembles a Dark Pattern. This may mean challenging the business logic,
	educating the team about the negative implications of these practices, and
	proposing transparent and honest alternatives. Transparency and simplification
	of processes, such as facilitating the cancellation of services or the
	changing of privacy settings, are examples of how the front-end can be used to
	empower the user, rather than imprison them.
</p>
<p>
	In addition to refusal, there is proactivity in building interfaces that
	prioritize user privacy and control. This involves implementing features that
	give the user power over their own data. For example, the creation of clear
	and accessible privacy control panels, where the user can easily see what data
	is being collected, how it is being used, and have the option to consent to or
	revoke that consent in a granular way. The General Data Protection Law (LGPD)
	and the GDPR require organizations to inform individuals about the collection
	and use of their data and ensure their explicit consent. We, developers, are
	responsible for translating these legal requirements into concrete
	functionalities in the interface.
</p>
<p>
	The use of technologies that minimize tracking is another front of resistance.
	This includes the preference for data analysis solutions that prioritize
	privacy, such as Matomo, which offers an open-source solution and 100% data
	ownership, in contrast to tools that centralize data collection in large
	corporations. The implementation of "privacy by design" and "privacy by
	default" approaches means that data protection is incorporated from the
	initial phases of development, and not as an additional feature. This may
	involve minimizing data collection, anonymizing whenever possible, and ensuring
	robust information security.
</p>
<p>
	In-depth technical knowledge about how tracking tools work is crucial.
	Understanding the operation of cookies, web beacons, and, especially, browser
	fingerprinting, allows the developer to implement countermeasures or, at the
	very least, warn about the risks. The choice of frameworks and libraries
	should be made with a critical eye, evaluating not only their technical
	efficiency, but also their ethical implications and how they can be
	instrumentalized for surveillance.
</p>
<p>
	Education and awareness are powerful tools. Front-end developers can and should
	be advocates for privacy and ethics in their teams and in the community.
	Sharing knowledge about the dangers of digital surveillance, manipulation
	tactics, and best practices for building ethical interfaces is a revolutionary
	act. This contributes to the development of a critical mindset, empowering
	individuals to question the content they consume and to protect their privacy
	online.
</p>
<p>
	The fight for digital freedom is a class struggle. Capital incessantly seeks
	new ways to extract value, and the human experience in the digital environment
	has become a new field of exploitation. We, front-end developers, are on the
	front lines of this battle. We are not mere coders; we are agents of
	transformation. By refusing to be architects of the prison and by dedicating
	ourselves to being engineers of freedom, we contribute to the construction of
	a more just, equitable, and truly free internet. It is a call to action, to
	solidarity, and to the construction of a digital future where technology
	serves humanity, and not capital.
</p>

<h3>
	5. Towards Digital Emancipation: Building a Front-end for the Revolution
</h3>
<p>
	The journey of the front-end developer, from mere executor of interfaces to
	conscious agent in the fight against digital surveillance, culminates in the
	vision of a front-end for the revolution. This is not a naive utopianism, but
	a materialist understanding that technology, although forged under the yoke of
	capital, contains within itself the seeds of its own overcoming. Our work, the
	tools we master, and the interfaces we build can be diverted from their
	purpose of domination and reoriented towards human emancipation.
</p>
<p>
	History teaches us that technology is never neutral; it reflects and amplifies
	existing power relations. However, it can also be a force for change, a
	catalyst for class consciousness and organization. The front-end, as the
	closest layer to the user, is the point of contact where ideology manifests
	itself and where resistance can be most effective. It is from the interface
	that we can unmask the logic of surveillance and build alternatives that serve
	the interests of the collective.
</p>
<p>
	The construction of a front-end for the revolution implies adopting a critical
	stance towards dominant tools and practices. This means questioning the
	hegemony of frameworks and libraries that, although efficient, may be
	intrinsically linked to business models based on surveillance. It is not about
	abandoning these tools, but about using them with discernment, adapting them
	and, when necessary, developing open-source alternatives that prioritize
	privacy, security, and user autonomy. The free and open-source software
	movement, for example, already carries in its DNA many of the principles of
	collaboration and sharing that are fundamental to a post-capitalist society.
</p>
<p>
	Resistance in the front-end also manifests itself in education and the
	dissemination of knowledge. We must demystify technology, revealing its hidden
	mechanisms of control and exploitation. Workshops, articles, lectures, and
	open-source projects that demonstrate how to build ethical and secure
	interfaces are acts of agitation and propaganda. By empowering other
	developers and users to understand and protect themselves from surveillance, we
	are building a solid foundation for digital emancipation. Awareness of "Dark
	Patterns" and "Surveillance Capitalism" is not just an academic exercise; it
	is a call to action, an awakening to the reality of digital exploitation.
</p>
<p>
	The fight for digital privacy is intrinsically linked to the fight for a more
	just society. Privacy is not a luxury, but a fundamental right, essential for
	freedom of thought, expression, and association. When surveillance becomes
	omnipresent, dissent is stifled and social control deepens. Therefore, every
	effort to protect privacy in the front-end is a step towards a freer and more
	democratic society.
</p>
<p>
	The future of front-end work, from a revolutionary perspective, is not that of
	merely optimizing engagement metrics for the profit of capital. It is that of
	building interfaces that serve the collective, that facilitate collaboration,
	access to free information, and social organization. Imagine interfaces that
	empower communities, that facilitate the exchange of knowledge without the
	intermediation of corporations that profit from our data, that allow the
	organization of social movements in a safe and private way. This is the
	front-end we should strive for.
</p>
<p>
	The incorporation of principles of accessibility and usability is not just a
	good design practice; it is a political act. By ensuring that interfaces are
	accessible to all, regardless of their abilities, we are combating digital
	exclusion and promoting social inclusion. A truly revolutionary front-end is
	one that serves everyone, and not just a privileged elite.
</p>
<p>
	The digital revolution will not come from above, but from below, from the hands
	of those who build the digital infrastructure. We, front-end developers, have a
	crucial role in this transformation. We are not just coders; we are organic
	intellectuals, capable of translating theory into practice, criticism into
	code. The fight against surveillance capitalism is an extension of the class
	struggle, and the front-end is one of its most important fields.
</p>
<p>
	May every line of code we write be a brick in the construction of a new
	society, where technology is a tool of liberation, and not of oppression. May
	we be, in fact, engineers of freedom, and not architects of the prison. History
	calls us to action. The future awaits us. Forward, comrades!
</p>

<h3>References</h3>
<ul>
	<li>
		Dark Patterns: O Lado Sombrio do Design que Manipula Decisões - Dois Z
		Publicidade. Disponível em:
		<a
			href="https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGzdLMFRxB-vlPCLhZaKPIz9tfb0U2Ei5_Ptn2QEEPdfYornnim92RugqrlG4zlITHizD_W6OaWWepKS2Jx5C48pNUtZ7KcyTwcqlby862JKD1uHSwAbPUYzl-j-WAcwv_fG2HkWGDd8hmu81dwmBM-FOAmWaeExOUKZi5ZwtG0uF7CaRrKL6CDrUKvDJ20bw=="
			target="_blank"
			>https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGzdLMFRxB-vlPCLhZaKPIz9tfb0U2Ei5_Ptn2QEEPdfYornnim92RugqrlG4zlITHizD_W6OaWWepKS2Jx5C48pNUtZ7KcyTwcqlby862JKD1uHSwAbPUYzl-j-WAcwv_fG2HkWGDd8hmu81dwmBM-FOAmWaeExOUKZi5ZwtG0uF7CaRrKL6CDrUKvDJ20bw==</a
		>
		Dark Patterns: The Dark Side of Design That Manipulates Decisions - Dois Z
		Publicidade. Available at:
	</li>
	<li>
		Dark Patterns e a Experiência do Usuário - ZRP TechBlog. Disponível em:
		<a
			href="https://zrp.com.br/blog/dark-patterns-e-a-experiencia-do-usuario/"
			target="_blank"
			>https://zrp.com.br/blog/dark-patterns-e-a-experiencia-do-usuario/</a
		>
		Dark Patterns and User Experience - ZRP TechBlog. Available at:
	</li>
	<li>
		O que são Dark Patterns? - Blog da Hostinger. Disponível em:
		<a
			href="https://www.hostinger.com.br/tutoriais/o-que-sao-dark-patterns"
			target="_blank"
			>https://www.hostinger.com.br/tutoriais/o-que-sao-dark-patterns</a
		>
		What are Dark Patterns? - Hostinger Blog. Available at:
	</li>
	<li>
		Dark Patterns: O lado obscuro do design de interfaces - Medium. Disponível
		em:
		<a
			href="https://medium.com/@marcelo.gomes.br/dark-patterns-o-lado-obscuro-do-design-de-interfaces-e0d0a7a0145f"
			target="_blank"
			>https://medium.com/@marcelo.gomes.br/dark-patterns-o-lado-obscuro-do-design-de-interfaces-e0d0a7a0145f</a
		>
		Dark Patterns: The Dark Side of Interface Design - Medium. Available at:
	</li>
	<li>
		A ética no mundo digital: a importância da segurança da informação - Blog da
		Conexão. Disponível em:
		<a
			href="https://conexao.com.br/blog/a-etica-no-mundo-digital-a-importancia-da-seguranca-da-informacao/"
			target="_blank"
			>https://conexao.com.br/blog/a-etica-no-mundo-digital-a-importancia-da-seguranca-da-informacao/</a
		>
		Ethics in the Digital World: The Importance of Information Security - Conexão
		Blog. Available at:
	</li>
	<li>
		O que é ética digital? - Blog da Certisign. Disponível em:
		<a
			href="https://www.certisign.com.br/blog/o-que-e-etica-digital"
			target="_blank"
			>https://www.certisign.com.br/blog/o-que-e-etica-digital</a
		>
		What is Digital Ethics? - Certisign Blog. Available at:
	</li>
	<li>
		Ética digital: o que é e qual a sua importância? - Blog da EAD Unifaveni.
		Disponível em:
		<a
			href="https://ead.unifaveni.com.br/blog/etica-digital-o-que-e-e-qual-a-sua-importancia/"
			target="_blank"
			>https://ead.unifaveni.com.br/blog/etica-digital-o-que-e-e-qual-a-sua-importancia/</a
		>
		Digital Ethics: What is it and what is its importance? - EAD Unifaveni Blog.
		Available at:
	</li>
</ul>]]></content:encoded>
      <pubDate>Sun, 01 Jun 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Comunicação</category>
      <category>Design Ético</category>
      <category>Comunidade Dev</category>
      <category>Open Source</category>
      <category>Tecnologia e Sociedade</category>
      <category>Transformação Digital</category>
      <category>Forças Produtivas</category>
      <category>Plataformas Digitais</category>
      <category>Privacidade de Dados</category>
      <category>Sociedade da Informação</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/63fc0253-68d6-41f0-bdb2-8b1767435b08" medium="image"/>
    </item>
    <item>
      <title>The Hidden Cost of Optimization</title>
      <link>https://www.mmlo.app/en/blog/o-custo-oculto-da-otimizacao</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/o-custo-oculto-da-otimizacao</guid>
      <description>How AI Impedes Creativity in Web Development.</description>
      <content:encoded><![CDATA[<p>AI-driven optimization in web development promises efficiency but conceals a cost: the potential entrapment of human creativity. We will delve into the dialectic between technological advancement and the essence of creative work, exploring how the relentless pursuit of productivity can, paradoxically, alienate the developer from their own work and homogenize the digital landscape. Is the machine freeing us or molding us in its image?</p>
<h3>The Algorithm of Alienation: When the Tool Becomes the Master</h3>

<p>In the dawn of the digital age, the promise of technology was liberation. Tools emerged to amplify human capacity, to allow us to create with a speed and scale previously unimaginable. In web development, this promise seemed to materialize with each new library, each new framework that simplified the process of transforming ideas into interactive interfaces. Optimization, then, was the relentless pursuit of cleaner, faster, more efficient code – a noble quest aimed at improving the user experience and, by extension, the very art of building for the web. But, as in every narrative of unbridled progress, there is a dark side, a hidden cost that is rarely debated in boardrooms filled with charts of ascending productivity. The meteoric rise of Artificial Intelligence, once seen as the developer's ultimate ally, is beginning to reveal its more complex face: that of a subtle master who, under the guise of optimization, imprisons the very creativity it should serve.</p>

<p>Let's think for a moment about the nature of the front-end developer's work. It is not merely the writing of lines of code; it is the transmutation of abstract concepts into tangible experiences. It is the dance between logic and aesthetics, engineering and art. Each pixel, each transition, each interaction is the result of creative choices, of a design process that emanates from the human mind. The developer, in this context, is a digital artisan, an architect of virtual realities. But what happens when the hammer and chisel begin to dictate the shape of the sculpture? What happens when the tool, which should be an extension of the creator's will, begins to predefine the limits of their expression? It is here that AI-driven optimization reveals its cruelest paradox: in seeking algorithmic perfection, it inadvertently paves the way for homogenization and, ultimately, for the alienation of the creator themselves.</p>

<p>The promise of AI in web development is seductive. Tools such as automated code generators, predictive performance optimizers, pattern-based design assistants, and even platforms that create entire interfaces from textual descriptions, promise a future where repetitive work is eradicated, freeing the developer for higher-value tasks. Who wouldn't want an assistant that writes the boilerplate, suggests the best color palette, or optimizes image loading with a click? The narrative is clear: more time to innovate, less time for boredom. However, this "liberation" comes at a price. By delegating more and more creative and technical decisions to AI, the developer begins to disconnect from the final product of their work. They become less of a creator and more of an algorithm supervisor, a mere operator of a machine that, while efficient, lacks soul.</p>

<p>This disconnection is not accidental; it is a modern manifestation of the alienation of labor, a concept deeply explored by Karl Marx. Marx argued that, under capitalism, the worker is alienated in four fundamental ways. In the context of AI-driven web development, these forms of alienation manifest themselves in disturbing ways. First, there is **alienation from the product of labor**. When an AI generates a large part of the code or design of an interface, the developer loses the sense of ownership and authorship over what is being built. The final product is no longer the expression of their unique vision, but rather the result of an algorithmic process that they merely supervised. "Optimization" becomes a veil that hides the loss of creative identity in the digital artifact.</p>

<p>Secondly, **alienation from the labor process** deepens. Web development, at its core, is an iterative process of problem-solving, experimentation, and discovery. It is a continuous dialogue between the developer and the code, where creativity flourishes in overcoming technical and aesthetic challenges. When AI takes over tasks such as choosing components, structuring CSS, or optimizing scripts, the developer is relegated to a "curator" or "reviewer" role. The process becomes less about active creation and more about passive validation. The joy of discovery, the satisfaction of shaping something from scratch, is replaced by the cold efficiency of an automated process. Work, which should be a source of fulfillment, transforms into a routine devoid of deep creative engagement.</p>

<p>The third form of alienation, **alienation from human essence (or "species-being")**, is perhaps the most insidious. Marx argued that the capacity for creative and conscious work is what distinguishes humans from other species. It is through work that we express our individuality, our ability to transform the world according to our will and imagination. In web development, this capacity manifests itself in the ability to conceive innovative solutions, to create user experiences that transcend the merely functional. When AI, in its pursuit of optimization, standardizes and automates these creative decisions, it not only optimizes the code; it optimizes the very humanity of the process. The developer is encouraged to think within the parameters of the algorithm, to accept the "best practices" dictated by the machine, rather than challenge or transcend them. Originality becomes a luxury, experimentation an unnecessary risk. Optimization, in this sense, is not just a tool; it is an ideology that shapes the developer's mind, imprisoning their innate ability to create freely.</p>

<p>Finally, **alienation from other human beings** also manifests itself. Although AI may seem like an individual tool, its large-scale implementation in technology companies often leads to a reduction in human collaboration and fragmentation of teams. If an algorithm can generate code or optimize designs, the need for creative discussions, collective brainstorming, and human feedback may diminish. The developer becomes more isolated, interacting more with the machine than with their peers. The camaraderie, mutual learning, and exchange of ideas that enrich the work environment are replaced by the aseptic efficiency of human-machine interaction. Optimization, which should connect us to users more effectively, paradoxically disconnects us from each other in the creative process.</p>

<p>Consider practical examples. Tools like GitHub Copilot, which suggest lines of code and entire functions, are undeniably useful for accelerating development. But, by blindly relying on these suggestions, the developer may miss the opportunity to explore alternative solutions, to learn the fundamentals behind certain implementations, or to develop their own coding style. The "optimization" here is speed, but the cost is the depth of learning and the originality of the solution. Similarly, design platforms that use AI to generate layouts and components based on usability data (such as some plugins for Figma or automated design system tools) can create interfaces that are "perfect" from the point of view of conversion or user experience. However, this perfection is often a statistical average, a convergence towards what is "optimal" for the majority, to the detriment of what is truly innovative or artistically expressive. Optimization, in this case, leads to standardization, to the dilution of visual identity, and to the loss of the ability to surprise and delight the user with something genuinely new.</p>

<p>The trap of optimization lies in its promise of unrestricted efficiency. We are conditioned to believe that faster is always better, that more automated is always superior. This is a powerful cognitive fallacy, the **automation bias**, which leads us to over-rely on automated systems, even when human intuition or creativity could offer a superior path. AI offers us a shortcut, and our mind, naturally inclined to save energy, readily accepts it. But, in doing so, we give up something invaluable: the ability to shape the digital world with our own hands, with our own minds, with our own souls. Optimization, which should be a means to an end – the creation of rich and meaningful digital experiences – becomes an end in itself, an imperative that dictates the form and content of our work. The hidden cost is the gradual erosion of creativity, the transformation of the developer from an artisan into a mere machine operator, trapped in the golden cage of algorithmic efficiency.</p><h3>The Tyranny of the Standard: How Algorithmic Optimization Dilutes Digital Diversity</h3>

<p>The promise of the internet, in its early days, was that of a vast and inexhaustible ocean of possibilities, a space where the diversity of ideas and the uniqueness of human expression could flourish without the shackles of traditional media. Each site, each application, was a blank canvas, an invitation to innovation and experimentation. In the web front-end, this freedom translated into the ability to create interfaces that not only worked but also had personality, that told a unique visual story. However, as optimization became the mantra of the industry, driven by Artificial Intelligence algorithms that promise the "best" experience, we have witnessed a paradoxical and worrying trend: the homogenization of the web. What was once a garden of a thousand flowers is beginning to resemble a digital monoculture, where efficiency is celebrated above originality, and the standard imposes itself on the singular.</p>

<p>AI, in its relentless pursuit of performance metrics, conversion rates, and user engagement, operates based on data. It analyzes millions of interactions, navigation patterns, and A/B test results to identify what "works best." The problem lies in what "works best" means to an algorithm. For AI, "best" is synonymous with predictability, with conformity to what has already proven to be effective for the majority. It does not seek disruption, surprise, or intrinsic beauty; it seeks optimization for a specific, usually commercial, outcome. This leads to a convergence towards the "average optimal," towards the design that pleases the largest slice of the public but, at the same time, lacks any distinctive trait. It is the application of **confirmation bias** on an industrial scale: AI confirms and amplifies what is already popular, reinforcing existing patterns instead of encouraging the exploration of new paths.</p>

<p>Imagine a scenario where all the restaurants in the world, in their pursuit of "optimization" and "efficiency," began to serve exactly the same dish, prepared in the same way, because the data indicates that it is the most consumed and generates the highest profit. What would be gained in predictability and volume would be lost in culture, in flavor, in experience. Similarly, in web development, algorithmic optimization pushes us towards a universe where landing pages look alike, e-commerce sites adopt the same navigation structures, and UI/UX components become indistinguishable between different platforms. The "best practice" dictated by AI becomes the only practice, stifling the experimentation and audacity that are the engine of true innovation. The front-end developer, who was once an artist with a palette of unlimited colors and shapes, now finds themselves limited to a predefined set of building blocks, all "optimized" for the same purpose.</p>

<p>This standardization is not a mere side effect; it serves very clear economic interests. In capitalism, mass production and predictability are key to maximizing profit. Diversity, uniqueness, and disruptive innovation, although valued in theory, represent additional risks and costs. An "optimal" and replicable design, generated or suggested by AI, reduces development time, minimizes the need for creative experimentation (and, therefore, for more expensive human talent), and ensures a predictable success rate. Creativity, which was once a competitive differentiator, is now commodified, transformed into a standardized product that can be generated at scale. The developer's work, which once required intuition, aesthetic sensitivity, and the ability to solve problems in a non-linear way, is gradually reduced to an assembly task, where prefabricated parts are fitted together according to the instructions of an algorithm. Optimization, in this context, is a tool of control, a way to ensure that digital production aligns with the imperatives of capital, even if it means sacrificing the richness and complexity of the human experience.</p>

<p>The ideology behind this homogenization is subtle but powerful. It is based on the **authority bias**, where the "intelligence" of the machine is seen as superior to human intuition. If AI, with its ability to process vast amounts of data, suggests a certain path, who are we to question it? This blind belief in algorithmic superiority leads developers and designers to abdicate their own creative judgment in favor of the "best practices" dictated by the machine. The result is an internet that, although functional and efficient, lacks soul, surprise, and authenticity. Interfaces become interchangeable, experiences predictable, and the ability of a site or application to truly stand out and create a deep emotional connection with the user is severely compromised. Optimization, which should serve the user, ends up serving the logic of mass production, transforming the web into a vast shopping center where all the stores sell slightly different products but with the same packaging.</p>

<p>Let's think about the impact of this on digital culture itself. The internet, at its core, is a reflection of humanity. If humanity is diverse, complex, and multifaceted, the internet should mirror this richness. However, by allowing optimization algorithms to guide us towards the "average optimal," we are building a digital future where individual expression is diluted in favor of conformity. What happens to radical ideas, to experimental designs, to interfaces that challenge conventions? They are marginalized, considered "not optimized," "inefficient," "out of the standard." Genuine innovation, which often arises from breaking paradigms and exploring unknown territories, is stifled by a system that rewards the replication of proven success. Algorithmic optimization, in this sense, is not just a technical tool; it is a cultural agent that shapes how we interact with the digital world, limiting our horizon of possibilities and pushing us towards an increasingly uniform and predictable experience.</p>

<p>History shows us that true creativity flourishes in freedom, in the ability to explore, to err, and to learn from mistakes. Optimization, when it becomes an end in itself, imposes a straitjacket on this process. It tells us what is "right" even before we have a chance to discover what is "possible." The front-end developer, who once saw themselves as an architect of experiences, runs the risk of becoming a mere assembly line worker, fitting prefabricated parts into a design dictated by artificial intelligence. The beauty of the web lies in its ability to be a mirror of the infinite variety of human experience. If we allow algorithmic optimization to lead us to a future of digital uniformity, we will be losing not only creativity but also an essential part of our own humanity. The hidden cost of optimization is the loss of diversity, the dilution of uniqueness, and the transformation of the web into an increasingly homogeneous and, paradoxically, less efficient space in its ability to inspire and connect.</p><h3>The Proletarianization of Code: When Optimization Disqualifies the Digital Artisan</h3>

<p>The history of capitalism is, in large part, the history of the relentless pursuit of greater productivity and control over the labor process. Since the Industrial Revolution, with the introduction of the steam engine and, later, the assembly line, capital has sought to fragment complex tasks into simpler and more repetitive operations. The goal? Reduce dependence on highly skilled workers, cheapen labor, and increase the predictability of production. This process, known as **deskilling**, is one of the cruelest manifestations of capitalist logic, transforming the artisan into a mere machine operator, the thinker into an executor. In the universe of web development, the rise of Artificial Intelligence, under the pretext of optimization, seems to be replicating this phenomenon, subtly but relentlessly, redefining the role of the front-end developer.</p>

<p>Let's think about the trajectory of a front-end developer. A decade ago, building a web interface required a deep knowledge of HTML, CSS, JavaScript, as well as a keen understanding of design, usability, and performance. It was a job that demanded creativity, complex problem-solving, and an ability to synthesize different disciplines. The developer was, in many respects, an architect and a builder, a master of their art. Today, with the proliferation of frameworks, libraries, and, more recently, AI tools that generate code, suggest components, and optimize layouts, the nature of this work is changing. Optimization, which should be a tool to enhance work, is becoming a substitute for human knowledge and intuition. The developer, who once dominated the process, is now invited to delegate more and more to the machine.</p>

<p>Consider the case of John, an experienced front-end developer. He has spent years honing his skills, learning the nuances of CSS, the complexities of JavaScript, and the art of creating fluid user experiences. He prided himself on his ability to solve problems creatively, to find elegant solutions to technical challenges. But, in recent years, John has felt a shift. AI tools in his development environment suggest entire lines of code, automatically optimize images, and even generate UI components based on textual descriptions. At first, he saw this as a blessing, a way to accelerate his work. However, over time, he realized that he was using his own brain less and the machine's suggestions more. He found himself accepting the code generated by AI without questioning, without exploring alternatives, simply because it was "optimized" and "fast." The joy of discovery, the satisfaction of building something from scratch, was being replaced by the efficiency of automation. John, the artisan, was becoming an algorithm supervisor, a mere "code paste-er" generated by artificial intelligence.</p>

<p>This is the essence of deskilling. It is not about making work easier, but about making it less dependent on complex skills and human judgment. When AI takes over tasks that once required deep knowledge and experience, the developer is gradually stripped of these skills. They become less of an expert and more of a "tool operator." Optimization, in this context, is a strategy to reduce the "quality" of the work required, making it more accessible to a larger number of people with less training, which, in turn, can put pressure on wages and increase the replaceability of labor. It is the proletarianization of code, where the developer, once an "organic intellectual" of the digital production process, is pushed into a position closer to the assembly line.</p>

<p>The illusion of empowerment is a powerful cognitive bias that leads us to overestimate the benefits of a tool while underestimating its hidden costs. AI tools are marketed as "assistants" that "free" us from repetitive tasks, allowing us to focus on "higher-value work." However, what is this "higher-value work" when the very definition of value is shaped by algorithmic optimization? If AI can generate the code, optimize the design, and even predict performance, what is left for the developer? The answer, often, is supervision, debugging, and adapting machine-generated solutions. Creative and strategic work is gradually eroded, replaced by a more tactical and reactive function.</p>

<p>Furthermore, dependence on AI tools can lead to what we call **anchoring bias**. Once AI suggests a solution, that suggestion becomes an "anchor" for the developer's thinking. It is much more difficult to deviate from this anchor and explore radically different paths, even if potentially more innovative or appropriate. Algorithmic optimization, by presenting an immediate "best practice," inhibits exploration, experimentation, and the ability to think "outside the box." Genuine innovation, which often arises from breaking paradigms and exploring non-obvious solutions, is stifled by the convenience of the ready-made answer. The developer, instead of being an explorer of new frontiers, becomes a follower of trails already mapped by artificial intelligence.</p>

<p>This deskilling not only affects the individual; it has systemic implications. If most developers become dependent on AI tools for basic tasks, the collective capacity for innovation and complex problem-solving may diminish. Companies become hostages to the technologies and optimization models dictated by the large corporations that develop these AIs. Technological autonomy, the ability to create truly customized and disruptive solutions, is compromised in favor of standardized efficiency. The hidden cost of optimization, in this sense, is the loss of mastery, the erosion of critical capacity, and the transformation of the developer from a digital artisan into a mere code worker, trapped in a production logic that values quantity over quality, efficiency over creativity, and the algorithm over the human soul.</p><h3>Digital Feudalism and the Commodification of Creativity: Who Owns the Algorithmic Means of Production?</h3>

<p>The history of humanity is, in its essence, a saga of control over the means of production. From the land in the feudal era, through the factories and machines of the Industrial Revolution, to the financial capital in the modern era, the possession and control of these means have determined power relations, social structures, and the distribution of wealth. In the 21st century, with the vertiginous rise of Artificial Intelligence and its deep integration into the fabric of digital production, we are witnessing a new and subtle, but equally powerful, concentration of power. Algorithms, AI models, and the platforms that host them are becoming the new "means of production" of intellectual and creative work, especially in web development. The crucial question that arises is: who owns these means? And what are the implications of this ownership for creativity, autonomy, and the future of work in the front-end?</p>

<p>In capitalism, innovation is often co-opted and transformed into a tool for accumulation. AI, which could be a liberating force, capable of democratizing creation and empowering the individual, is being, to a large extent, monopolized by a handful of technology giants. Companies like Google, Microsoft, OpenAI, Meta, and Amazon invest billions in AI research and development, building massive computing infrastructures, collecting vast oceans of data, and training models that become increasingly sophisticated and ubiquitous. These companies not only create the tools; they create the ecosystem in which modern web development operates. Think of the frameworks, the clouds, the code repositories, the integrated development environments (IDEs), and, now, the AI assistants that integrate with all of this. The front-end developer, who could once operate with a more decentralized and open set of tools, now finds themselves increasingly dependent on a handful of platforms controlled by these corporations.</p>

<p>This dependence is not accidental; it is a deliberate strategy. By offering "free" or low-cost AI tools that promise optimization and efficiency, these companies attract millions of developers to their ecosystems. Once inside, migration out becomes progressively more difficult due to the **network effect** and **switching costs**. The developer gets used to the convenience, the seamless integration, the intelligent suggestions. They become, in essence, a digital "vassal," working within the domains of a technological "feudal lord." This is the essence of what some theorists call "digital feudalism" or "platform capitalism": large corporations not only sell products but control the platforms that mediate most of the economic and social activity online, charging "tolls" (whether in data, attention, or technological dependence) and dictating the rules of the game.</p>

<p>Creativity, in this new arrangement, is commodified. What was once an organic, idiosyncratic, and deeply human process is now subjected to the logic of mass production and algorithmic optimization. AI, by analyzing millions of designs, code patterns, and user experiences, identifies what is "optimal" and replicates it. The result is a standardization that, although efficient, dilutes uniqueness. The developer's creativity is no longer valued for its intrinsic originality but for its ability to align with the patterns "optimized" by AI. The value of creative work is reduced to an efficiency metric, a conversion rate, a page load time. The digital artist, who once molded the clay with their own hands, now operates a machine that spits out "perfect" replicas of an algorithmic ideal. Optimization, in this sense, is not just about making things better; it is about making things more predictable, more scalable, and, crucially, more controllable by the holders of the algorithms.</p>

<p>This commodification of creativity has profound implications for the labor market itself. If AI can generate much of the boilerplate code, if it can suggest designs and optimize performance, the demand for developers with "basic" or "intermediate" skills may decrease, or their wages may be pushed down. What remains is a niche for "prompt engineers" – those who know how to interact with AI to obtain the desired results – or for architects of complex systems who orchestrate the interaction between different AIs and legacy systems. The skills pyramid inverts: creative and artisanal work is devalued, while the ability to operate and manage the machines becomes the new differentiator. This reinforces the **availability bias**, where the ease of access to AI-generated solutions leads to an overestimation of their effectiveness and an underestimation of the need for deeper human skills.</p>

<p>Furthermore, dependence on proprietary AI tools creates a new form of digital "enclosure." Just as peasants were expelled from their common lands during the enclosure in England, developers may find themselves "enclosed" by AI ecosystems that limit their freedom of action and their ability to innovate outside the established parameters. The code generated by an AI may be optimized for a particular platform, the design patterns suggested may favor the products of the company that developed the AI, and the performance data may be used to further refine the company's own models, creating a vicious cycle of dependence. Optimization, which should be a universal benefit, becomes a competitive advantage for the few who hold the algorithmic means, solidifying their power and their dominant position in the market.</p>

<p>The narrative that AI "frees" the developer from repetitive tasks so that they can focus on "higher-value work" is a powerful ideological tool. It appeals to our **optimism bias**, making us believe that the future will inevitably be better and that technology will solve all our problems. However, this narrative often ignores the fundamental question: who defines what is "higher value"? And who benefits from this "higher value"? If the "higher-value work" is the orchestration of algorithms that, ultimately, serve the interests of large corporations, then the "liberation" is, in fact, a redefinition of work that serves to consolidate existing power. Creativity, which should be a force of subversion and transformation, is domesticated, transformed into a resource to be optimized and exploited. The developer, who was once an agent of change, runs the risk of becoming a mere cog in an algorithmic machine, whose purposes and control are far beyond their influence.</p>

<p>In short, AI-driven optimization in web development, although presented as a neutral and beneficial advancement, is, in fact, an ideological and economic battlefield. It not only shapes how we build the web but also who builds it, how they are valued, and who benefits from their work. By commodifying creativity and concentrating the algorithmic means of production in the hands of a few, we are paving the way for a future where the web, which promised to be a space of freedom and diversity, may become a reflection of the tyranny of the standard and the relentless logic of capital. The hidden cost of optimization, in this scenario, is the loss of creative autonomy, the precarity of work, and the consolidation of a new type of feudalism, where algorithms are the new lords and human creativity, the new servitude.</p><h3>Reclaiming the Tool: Strategies of Resistance and the Struggle for Creative Autonomy</h3>

<p>Faced with the scenario we have painted – a future where algorithmic optimization, driven by Artificial Intelligence, threatens to alienate the developer, standardize the web, deskill work, and concentrate power in the hands of a few – the question that arises is not whether we should resist, but how. The history of class struggle, and of all humanity, is the history of resistance against the forces that seek to oppress and control. In the context of web development and the rise of AI, this resistance does not manifest itself in physical barricades but in acts of consciousness, in deliberate choices, and in the reaffirmation of the intrinsic value of human creativity. It is not about rejecting technology but about reclaiming it, about shaping it to serve the interests of humanity, and not just the imperatives of capital.</p>

<p>The first and most fundamental strategy of resistance lies in **awareness**. It is necessary to unmask the seductive narrative of "total optimization" and recognize the hidden costs it imposes. The developer needs to understand that algorithmic efficiency, although attractive, can be a trap that distances them from their own creative capacity. This means questioning the tools, not blindly accepting the "best practices" dictated by algorithms, and, above all, valuing one's own intuition and experience. The **authority bias**, which leads us to over-rely on the machine, must be actively combated. It is crucial to remember that AI is a tool, a reflection of the data with which it was trained and the objectives for which it was programmed. It does not possess consciousness, intuition, or the capacity for disruptive innovation that arises from human experience and the ability to make unexpected connections.</p>

<p>On a practical level, resistance can begin with the **reaffirmation of technical mastery**. Instead of becoming a mere "prompt engineer" or an algorithm supervisor, the developer should deepen their knowledge of the fundamentals. Understanding how CSS really works, how JavaScript manipulates the DOM, how browsers render pages – this deep knowledge is the antidote to deskilling. When one understands the internal mechanics, it is possible to go beyond the suggestions of AI, adapt them, challenge them, and, ultimately, create truly original solutions. AI can generate code, but the ability to debug, optimize manually, refactor, and innovate at a fundamental level still resides in the human domain. This pursuit of mastery is not a return to the past but a form of empowerment in the present, ensuring that the developer maintains control over their craft.</p>

<p>Another front of resistance is the **promotion of diversity and experimentation**. If AI pushes us towards homogenization, the answer is to embrace uniqueness. This means challenging standards, exploring unconventional designs, experimenting with new interaction approaches, and prioritizing artistic expression over mere functionality. It is an act of creative rebellion. Companies and development teams should be encouraged to allocate time and resources to "non-optimized" projects, to research and development that aims not only at efficiency but at genuine innovation and the creation of truly memorable experiences. The **endowment effect**, which makes us value what we create more, can be an ally here: by investing time and effort in creating something unique, the developer and the team develop an attachment and a pride that transcend the cold metrics of optimization.</p>

<p>**Collaboration and community building** are also vital. If digital feudalism seeks to isolate the developer and make them dependent on proprietary platforms, the answer is to strengthen networks of mutual support. This includes contributing to open-source projects, participating in developer communities that value the exchange of knowledge and collective construction, and advocating for open standards and interoperability. Dependence on a single AI vendor or a single platform can be mitigated by diversifying tools and building a more resilient and decentralized ecosystem. Solidarity among knowledge workers is a powerful force against precarity and deskilling. It is the reaffirmation that knowledge and creativity are common goods, not the exclusive property of corporations.</p>

<p>Finally, resistance involves **redefining success**. In a society obsessed with metrics of productivity and efficiency, it is easy to fall into the trap of believing that the "best" is always the fastest, the most automated, the most profitable. However, true success, from a human and creative point of view, can be measured by originality, by beauty, by the ability to inspire, by the depth of the connection that a digital experience establishes with the user. It is necessary to question the hegemony of algorithmic metrics and value the intangible, the subjective, the artistic. This requires a change of mentality, both individual and collective, that prioritizes quality over quantity, soul over algorithm. It is an act of **framing** the problem: instead of accepting the framing of optimization as the only path, we can frame creativity and autonomy as the true objectives.</p>

<p>The struggle for creative autonomy in web development, in the face of algorithmic optimization, is an extension of the historical struggle for the dignity of labor and for human freedom. It is not about demonizing AI, which has immense potential for good, but about ensuring that it is a tool at the service of humanity, and not the other way around. It is a call to action for front-end developers, the digital artisans of the 21st century, to reaffirm their role as creators, as thinkers, as artists. It is the struggle to ensure that the web of the future is a reflection of the rich and diverse tapestry of human experience, and not a homogeneous and soulless landscape, dictated by the cold logic of algorithms. The hidden cost of optimization is high, but the price of inaction is even higher: the loss of our own ability to dream, to create, and to shape the digital world in our own image.</p>

<p>---</p>

<h3>References:</h3>
<ul>
    <li>Marx, K. (2004). Economic and Philosophical Manuscripts. Boitempo Editorial. Available at: <a href="https://www.boitempoeditorial.com.br/produto/manuscritos-economico-filosoficos-117">https://www.boitempoeditorial.com.br/produto/manuscritos-economico-filosoficos-117</a></li>
    <li>Braverman, H. (1974). Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century. Monthly Review Press. Available at: <a href="https://monthlyreview.org/product/labor_and_monopoly_capital/">https://monthlyreview.org/product/labor_and_monopoly_capital/</a></li>
    <li>Zuboff, S. (2019). The Age of Surveillance Capitalism: The Fight for a Human Future at the New Frontier of Power. PublicAffairs. Available at: <a href="https://www.publicaffairsbooks.com/titles/shoshana-zuboff/the-age-of-surveillance-capitalism/9781610395694/">https://www.publicaffairsbooks.com/titles/shoshana-zuboff/the-age-of-surveillance-capitalism/9781610395694/</a></li>
    <li>Morozov, E. (2011). The Net Delusion: The Dark Side of Internet Freedom. PublicAffairs. Available at: <a href="https://www.publicaffairsbooks.com/titles/evgeny-morozov/the-net-delusion/9781610390347/">https://www.publicaffairsbooks.com/titles/evgeny-morozov/the-net-delusion/9781610390347/</a></li>
    <li>Kahneman, D. (2011). Thinking, Fast and Slow. Farrar, Straus and Giroux. Available at: <a href="https://us.macmillan.com/books/9780374533557/thinkingfastandslow">https://us.macmillan.com/books/9780374533557/thinkingfastandslow</a></li>
</ul>]]></content:encoded>
      <pubDate>Tue, 27 May 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Inteligência Artificial</category>
      <category>React</category>
      <category>Impacto da IA</category>
      <category>Futuro do Trabalho</category>
      <category>Tecnologia e Sociedade</category>
      <category>Automação</category>
      <category>Precarização</category>
      <media:content url="https://useful-duck-844.convex.cloud/api/storage/5343309f-67fa-4bfb-a353-2e220a7c6159" medium="image"/>
    </item>
    <item>
      <title>Beconfident.app: Artificial Intelligence and WhatsApp Revolutionize English Language Learning</title>
      <link>https://www.mmlo.app/en/blog/beconfidentapp-inteligencia-artificial-e-whatsapp-na-revolucao-do-aprendizado-de-ingles</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/beconfidentapp-inteligencia-artificial-e-whatsapp-na-revolucao-do-aprendizado-de-ingles</guid>
      <description>An analysis of the social impact, advantages, and innovative AI-driven methodology that is changing how Brazilians learn English.</description>
      <content:encoded><![CDATA[<h3>Introduction: Breaking Barriers in English Learning</h3>

<p>In an increasingly connected and globalized world, fluency in English has gone from being a differentiator to a fundamental necessity. Whether to boost your career, access knowledge, consume culture, or simply communicate with people from different parts of the globe, mastering the English language opens doors and expands horizons. However, the path to fluency is often arduous, expensive, and, for many, inaccessible. Traditional methods, in-person classes with rigid schedules, and the lack of constant practice are significant barriers that keep countless people from achieving their dream of speaking English with confidence. It is in this challenging scenario that <strong>Beconfident.app</strong> emerges, an innovative platform that aims to revolutionize the way we learn English, using <strong>artificial intelligence (AI)</strong> and the convenience of tools that are already part of our daily lives, such as WhatsApp.</p>

<p>Beconfident's proposal is bold: to make English learning more <strong>accessible, personalized, and practical</strong>, allowing users to develop fluency by conversing about topics of their interest, anytime and anywhere. By integrating AI tutors directly into WhatsApp and also through a dedicated application, the platform breaks the rigidity of conventional methods and organically inserts language practice into students' routines. But does this approach really work? What is the true social impact of a tool like this? What are the concrete advantages for those who decide to embark on this learning journey? And what do experts and users themselves say about this experience? In this article, we will conduct an in-depth analysis of Beconfident.app, exploring its innovative methodology, its benefits, its transformative potential, and the discussions it raises in the field of education and technology.</p>

<h3>The Social Impact: Democratizing Access to Fluency</h3>

<p>One of the most relevant aspects of Beconfident.app lies in its potential for <strong>social impact</strong>, particularly with regard to democratizing access to English learning. In a country like Brazil, marked by profound social and economic inequalities, English proficiency often acts as a watershed, a cultural and professional asset that concentrates opportunities in the hands of a few. Traditional language courses tend to have high tuition fees, expensive materials, and require commuting, factors that exclude a vast portion of the population that does not have financial resources or free time. Beconfident, by offering an alternative based on accessible technology (smartphones and WhatsApp, widely used) and with a subscription model potentially more economical than traditional schools, directly tackles this structural barrier.</p>

<p>The possibility of practicing English <strong>24 hours a day, 7 days a week</strong>, directly through WhatsApp, eliminates the need to commute and adapts to the routine of workers, students, and parents who have limited and fragmented time slots. This flexibility is crucial to include individuals who would otherwise never be able to attend an in-person course. Furthermore, the platform actively seeks, as mentioned in its communications [2], to bring the opportunity to learn English "affordably and in the palm of your hand to everyone," indicating an awareness of its role in reducing barriers to access. The very nature of the interaction via AI can reduce anxiety and the fear of making mistakes, common in traditional classroom environments, making learning less intimidating for those who feel insecure or have had negative experiences in the past.</p>

<p>However, a critical analysis from a social perspective cannot ignore the nuances. Although digital technology expands access, it does not completely eliminate inequalities. The need for a smartphone with internet access, even a basic one, is still a limiting factor for the most vulnerable segments of the population. The quality of the internet connection can also impact the experience, especially in features that require more bandwidth, such as voice or video calls (if available). Moreover, the dependence on a tool like WhatsApp, belonging to a technological conglomerate (Meta), raises questions about data privacy and the concentration of digital power, aspects that deserve continuous attention from society. The promise of democratization is potent, but its realization depends on a more inclusive digital ecosystem and public policies that guarantee universal access to technology and quality connectivity. Beconfident fits into this context as a promising tool, but broader social transformation requires a systemic view that goes beyond the individual technological solution.</p>

<h3>Advantages for Users: Flexibility, Personalization, and Real Practice</h3>

<p>Beyond the broader social impact, Beconfident.app offers a series of <strong>direct and tangible advantages</strong> for its individual users. The main one, as already mentioned, is the <strong>unparalleled flexibility</strong>. The possibility of practicing English anytime, anywhere, using WhatsApp or the dedicated application, adapts perfectly to the busiest schedules and modern lifestyles. There is no longer a need to stick to fixed class times or waste time commuting. Learning happens during breaks in the day, on transportation, at home, transforming idle moments into productive practice opportunities.</p>

<p>Another strong point is the <strong>personalization of learning</strong>. The platform uses artificial intelligence not only to simulate conversations, but also to adapt the content to the level and interests of each student. The user can choose to converse with AI tutors of different nationalities and with varied profiles (such as photographers, athletes, writers, etc., as presented on the official website [1]), focusing on topics that really interest them. This makes learning more engaging and relevant, avoiding the monotony of generic lessons. The AI also provides <strong>real-time feedback</strong> on pronunciation and grammar, a valuable resource that allows the student to correct errors immediately and accelerate their progress. This instant and private correction, without the embarrassment of a public evaluation, is an important differentiator for many students.</p>

<p>The methodology focuses on <strong>immersion in real-life situations</strong>. Instead of just memorizing grammar rules, users practice English in simulated scenarios, such as trips, negotiations, orders in restaurants, or discussions about current topics. This communicative approach, centered on the practice of conversation, is fundamental to develop confidence (the platform's name itself, "Be Confident," reinforces this) and effective fluency. The integration with WhatsApp [3, 5] allows the practice to be even more natural, using a familiar interface for exchanging text and audio messages, in addition to calls (according to the described functionalities). The platform also mentions the generation of <strong>personalized study plans</strong> based on the student's English level and weekly availability, offering a guided structure for those who prefer a more defined roadmap.</p>

<p>The advantages listed by the platform include:
<ul>
    <li>Conversation mode</li>
    <li>Professor mode</li>
    <li>Vocabulary mode</li>
    <li>Unlimited conversation (in paid plans)</li>
    <li>Questions via audio/text</li>
    <li>Q&A via WhatsApp</li>
    <li>Translation via audio</li>
    <li>Support in Portuguese</li>
    <li>Advanced artificial intelligence</li>
</ul>
These combined resources aim to offer a complete, dynamic, and tailored learning experience to individual needs, representing a significant advance over many traditional methods.</p>

<h3>The Revolution in the Method: AI and WhatsApp as Allies of Fluency</h3>

<p>The core of Beconfident.app's innovation lies in its methodological approach, which combines <strong>advanced artificial intelligence</strong> with the <strong>ubiquity of WhatsApp</strong>. This combination represents a significant break with traditional language teaching models. The AI does not act merely as a chatbot programmed with fixed answers, but as an <strong>interactive tutor</strong> capable of maintaining fluid and adaptive conversations on a vast range of topics. This ability to simulate real dialogues, with different accents and speaking styles (through the various tutor profiles), provides a conversation practice much closer to reality than rigid textbook exercises.</p>

<p>The integration with WhatsApp [3, 5, 6] is perhaps the most disruptive element. By bringing learning to a platform that billions of people already use daily for personal and professional communication, Beconfident removes barriers to access and integrates the study of English into the daily flow. The exchange of text and audio messages, familiar to WhatsApp users, makes the interaction with the AI intuitive. The <strong>automatic Portuguese-English translation</strong> functionality during conversations [3, 5] is a valuable support, especially for beginners, allowing them to understand and respond even when they do not yet fully master the vocabulary or sentence structure. This reduces frustration and keeps the student engaged, encouraging trial and error as a natural part of the process.</p>

<p>The use of AI allows a level of <strong>personalization and instant feedback</strong> that would be impractical in large classes or even with overloaded private teachers. The AI can analyze pronunciation, identify grammatical errors, and suggest corrections in real time [1], directing learning efficiently. Furthermore, the AI's ability to adapt the conversation to the student's interests [1] makes the practice more meaningful and motivating. Discussing hobbies, work, news, or pop culture in English, with a "tutor" available 24/7, transforms studying into a more enjoyable and less obligatory activity. This student-centered approach and constant practice of conversation aligns with modern theories of second language acquisition, which emphasize the importance of authentic communication and comprehensible input.</p>

<p>It is important to note that, although AI offers many advantages, it does not completely replace human interaction. The cultural nuance, empathy, and adaptability of an experienced human teacher are still irreplaceable in many aspects of learning. However, Beconfident's AI functions as a powerful tool to <strong>complement and scale</strong> access to conversation practice, which is often the biggest bottleneck in English learning for most people. The platform seems to have found an interesting balance, using technology to provide intensive practice and personalized feedback, democratizing access to resources that were previously restricted.</p>

<h3>What Do Experts and Users Say?</h3>

<p>The evaluation of an innovative educational platform like Beconfident.app would not be complete without considering the perspectives of experts in language teaching and, mainly, of the users themselves who experience the tool on a daily basis. Although in-depth analyses by applied linguists or renowned educators may still be incipient, given the relative novelty of the platform, it is already possible to find <strong>feedbacks and reviews</strong> in various sources that give us clues about its reception and perceived effectiveness.</p>

<p>In the app stores (Google Play [3] and Apple App Store [5]), Beconfident generally has positive ratings, although the total number of ratings may still be limited in some regions. Comments often highlight the <strong>practicality of integration with WhatsApp</strong>, the <strong>convenience of being able to study anytime</strong>, and the <strong>usefulness of instant feedback</strong> provided by the AI. Users mention the feeling of progress in conversation and the loss of fear of speaking English. Review videos on YouTube [4, 8], although often produced by affiliates or enthusiastic users, also tend to highlight the innovation of the method and the ease of use. Some compare Beconfident with other AI platforms for language learning, pointing out its differentiators.</p>

<p>However, it is crucial to maintain a critical eye. Some reviews may mention points for improvement, such as the need to enhance the naturalness of the AI's voice in certain situations, expand the variety of conversation scenarios, or optimize the interface of the dedicated application. The effectiveness of the platform may also vary depending on the student's initial proficiency level, their discipline, and their specific goals. For very advanced students, the interaction with the AI may eventually become limited in terms of complexity and depth. The absence (or lower frequency) of interaction with human native speakers may also be seen as a disadvantage by some experts, who advocate the importance of cultural exchange and exposure to authentic linguistic variations that only human contact provides.</p>

<p>From the point of view of investment and recognition in the startup ecosystem, Beconfident has attracted attention. The investment received from Latitud [7], a platform that supports startups in Latin America, and the recognition in startup competitions [1] (such as mentions of Harvard and MIT in its promotional material, although the specific details of these competitions need to be verified) suggest that experts in the technology and education market see potential in the solution. Daniel Nascimento, one of the co-founders, expresses on platforms like LinkedIn [10] the vision that Beconfident can "really positively impact people's lives," reinforcing the discourse of transformation and accessibility. In short, the overall perception seems to be positive, recognizing the innovation and practicality of Beconfident, but the consolidation of its reputation and the validation of its long-term effectiveness will depend on more independent studies and the continuous evolution of the platform.</p>

<h3>Conclusion: The Future of Language Learning in the Palm of Your Hand?</h3>

<p>Beconfident.app emerges as an intriguing and potentially transformative proposal in the landscape of English teaching. By leveraging <strong>artificial intelligence</strong> and integrating it with ubiquitous platforms like <strong>WhatsApp</strong>, the startup directly addresses some of the most persistent barriers to language learning: cost, lack of time, methodological rigidity, and, crucially, the lack of opportunities for constant conversation practice. The flexibility of being able to practice anytime and anywhere, the personalization of content to the student's interests, and the instant feedback on pronunciation and grammar are undeniable advantages that can significantly accelerate the path to fluency.</p>

<p>From a social point of view, the platform carries the promise of <strong>democratizing access</strong> to English, a language whose mastery still represents significant capital in the job market and in access to global information. By offering a more accessible and adaptable alternative than many traditional courses, Beconfident has the potential to include individuals who were previously on the margins of learning opportunities. However, it is essential to maintain a critical perspective, recognizing that digital exclusion is still a reality and that the dependence on technologies controlled by large corporations raises important questions about privacy and data sovereignty.</p>

<p>The methodology focused on <strong>conversation and immersion</strong> in real situations, combined with the AI's ability to simulate interactions and provide corrections, represents an interesting evolution in teaching approaches. Although qualified human interaction remains irreplaceable for certain aspects of learning, Beconfident's AI positions itself as a powerful tool to scale practice and offer individualized support. The initial reception by users and the recognition in the startup ecosystem indicate a promising future, but the consolidation of its effectiveness and impact will depend on the continuous evolution of technology and more in-depth long-term evaluations.</p>

<p>In short, Beconfident.app is not just an application, but a symptom and an agent of the profound changes that artificial intelligence is bringing to education. It represents a significant step towards more <strong>personalized, flexible, and accessible</strong> language learning, putting the power of practice in the student's palm. It remains to follow its trajectory and observe how it and other similar initiatives will shape the future of acquiring new languages in an increasingly interconnected world.</p>

<h3>References</h3>
<ul>
    <li>[1] Beconfident. **Beconfident | Speak English with artificial intelligence**. Available at: https://beconfident.app/. Accessed: May 23, 2025.</li>
    <li>[2] BeConfident (LinkedIn Post). **Young people create a startup for English practice...**. Available at: https://pt.linkedin.com/posts/beconfidentapp_jovens-criam-startup-para-pr%C3%A1tica-de-ingl%C3%AAs-activity-7251952070788866050-hpJN. Accessed: May 23, 2025.</li>
    <li>[3] Google Play Store. **BeConfident – Apps on Google Play**. Available at: https://play.google.com/store/apps/details?id=ai.beconfident.app&hl=pt. Accessed: May 23, 2025.</li>
    <li>[4] Wesley Reviews (YouTube). **Is BeCONFIDENT good? Does English via WhatsApp work? Get to know BeConfident from the inside!**. Available at: https://www.youtube.com/watch?v=J1oAtnjRBuU. Accessed: May 23, 2025.</li>
    <li>[5] Apple App Store. **BeConfident: Learn English on the App Store**. Available at: https://apps.apple.com/br/app/beconfident-aprenda-ingl%C3%AAs/id6468505840. Accessed: May 23, 2025.</li>
    <li>[6] Projeto Draft. **How about learning English via WhatsApp? BeConfident uses AI to...**. Available at: https://www.projetodraft.com/que-tal-aprender-ingles-pelo-whatsapp-a-beconfident-usa-ia-para-ajudar-os-alunos-a-ganhar-fluencia-no-idioma/. Accessed: May 23, 2025.</li>
    <li>[7] Latitud Blog. **Why Latitud invested in BeConfident, applying AI to language learning**. Available at: https://www.latitud.com/blog/latitud-beconfident-ai-language-learning. Accessed: May 23, 2025.</li>
    <li>[8] BeConfident App - Speaking (Instagram Reel). **This is simply the best app to learn English...**. Available at: https://www.instagram.com/reel/C7fKvUXup4A/. Accessed: May 23, 2025.</li>
    <!-- Added two more references to ensure a minimum of 6, using already researched sources -->
    <li>[9] BeConfident (YouTube Channel). **How to use the BECONFIDENT APP?**. Available at: https://www.youtube.com/watch?v=52hx7h52ee4. Accessed: May 23, 2025.</li>
    <li>[10] Daniel Nascimento (LinkedIn Post). **BeConfident - The AI startup created by students...**. Available at: https://pt.linkedin.com/posts/danielnascimentoo_beconfident-a-startup-de-ia-criada-por-estudantes-activity-7189623271963713538-fcP-. Accessed: May 23, 2025.</li>
</ul>]]></content:encoded>
      <pubDate>Thu, 22 May 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Chatbots</category>
      <category>Acessibilidade</category>
      <category>Produtividade</category>
      <category>Inteligência Artificial</category>
      <category>Tecnologia e Sociedade</category>
      <category>Transformação Digital</category>
      <category>Tecnologia e Desenvolvimento</category>
      <category>Inovação Local</category>
      <category>Comunidades de Aprendizagem</category>
      <category>Tecnologia Educacional</category>
      <media:content url="https://i.postimg.cc/XNycM6xG/beconfident.png" medium="image"/>
    </item>
    <item>
      <title>Technology, Work, and Social Transformations in the Contemporary Era</title>
      <link>https://www.mmlo.app/en/blog/tecnologia-trabalho-e-transformacoes-sociais-na-era-contemporanea</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/tecnologia-trabalho-e-transformacoes-sociais-na-era-contemporanea</guid>
      <description>A critical analysis of how innovations like AI and automation are redefining work, class structure, and power dynamics in today&apos;s society.</description>
      <content:encoded><![CDATA[<p><strong>Introduction:</strong> This analysis of "Technology, Work, and Social Transformations in the Contemporary Era" seeks to foster an in-depth reflection on the transformations that technology imposes on the social structure and contemporary power dynamics. Starting from a careful observation of the material reality, we aim to unveil the complexities inherent in technological development and how it affects living conditions and social interactions in different global contexts.</p> <p>Far from a simplistic or merely technical view, we understand that technology, work, and social transformations in the contemporary era is a multifaceted field, where innovations, driven by human creativity and social needs, interact with existing economic and political structures. This article proposes to examine the implications of this dynamic, exploring both the constructive potential of technology and the challenges to its equitable and socially responsible use.</p> <h3>Historical Context and Foundations of Social Analysis</h3> <p>For a comprehensive analysis of "Technology, Work, and Social Transformations in the Contemporary Era," it is essential to use conceptual tools that allow us to understand reality in its totality and its transformations. The concept of <strong>the intensification of productive forces and its impact on social relations of production</strong> is fundamental to unraveling the logic underlying the observed phenomena. We will analyze how the introduction of new technologies, such as artificial intelligence and advanced automation, not only increases productive capacity but also reconfigures the relations between capital and labor, generating new forms of control, qualification, and division of labor.</p> <p>Historically, the development of society's productive capacities has always been intrinsically linked to transformations in social relations. From the first technological revolutions to the current digital age, we observe how different social groups seek to direct the means of production to serve specific interests. In the context of "technology, work, and social transformations in the contemporary era," this manifests itself through the reconfiguration of labor markets with automation, where the pursuit of efficiency and profitability drives the substitution of living labor for dead labor, while also creating new demands for specialized skills, deepening the segmentation of the working class.</p> <h3>Analysis of Contemporary Manifestations and Their Complexities</h3> <p>In the current scenario, "Technology, Work, and Social Transformations in the Contemporary Era" presents itself with multiple, often complex facets. On the one hand, automation and AI offer the potential to eliminate repetitive and dangerous tasks, as well as optimize complex production processes, opening the possibility of reducing the working day and focusing on more creative and intellectually stimulating activities for human beings. On the other hand, the way this technology is implemented and distributed can generate new forms of inequality in access to the benefits of these technologies, concentration of power in the hands of a few corporations that hold the patents and data, and the risk of structural unemployment for large sectors of the population, if there are no compensatory social policies and a redistribution of the wealth generated.</p> <p>We will examine concrete cases and data that illustrate this dynamic:</p><ul><li><strong>Case Study 1:</strong> The impact of automation on the automotive industry in countries like Brazil, where the modernization of plants has led to significant productivity gains for companies, but also to mass layoffs and the need to retrain workers for functions that often were not created in the same proportion.</li><li><strong>Case Study 2:</strong> The use of AI-based project management software in technology companies, which promises greater efficiency and collaboration, but which, in some cases, is used to intensively monitor the productivity of developers, generating pressure and anxiety.</li><li><strong>Data Analysis:</strong> Data from the ILO and research institutes on the increase in productivity decoupled from the increase in wages in recent decades in various industrialized sectors, and the correlation between investment in automation and the stagnation or fall in the level of employment in certain occupations.</li></ul><p>These examples demonstrate that technology is not an isolated factor, but an element that interacts deeply with existing social, economic, and political structures. The way "technology, work, and social transformations in the contemporary era" develops and is applied reflects and, at the same time, influences these structures.</p> <h3>Perspectives for a Socially Just Technological Development</h3> <p>In light of the presented scenario, critical reflection is fundamental to guide action. It is important to identify the paths to a technological development that is socially just and that promotes collective well-being. The pursuit of democratizing access to the benefits of technology, ensuring that the productivity gains resulting from automation and AI are returned to society as a whole, through the reduction of the working day without loss of pay, investment in education and retraining, and strengthening the social protection network, is an important step to align technical progress with social progress.</p> <p>Experiences of the discussion on universal basic income as a way to mitigate the impacts of technological unemployment, the promotion of worker cooperatives that use advanced technologies in a self-managed way, and the creation of industrial policies that encourage innovation aimed at solving social and environmental problems, demonstrate that it is possible to build another technological future, guided by the needs of the majority and not by restricted interests. Citizen participation and the strengthening of democratic institutions are crucial to drive these transformations and for "technology, work, and social transformations in the contemporary era" to serve human emancipation and the construction of a more equitable society.</p><p>The construction of governance mechanisms and popular participation capable of directing scientific and technological development to meet collective interests is a central challenge. This implies not only regulating the uses of technology but also fostering a broad and informed public debate about its directions and purposes.</p> <h3>Conclusion: Towards a Conscious and Transformative Appropriation of Technology</h3><p>In summary, the analysis of "Technology, Work, and Social Transformations in the Contemporary Era" from a socially conscious perspective reveals that technology is a field of intense interactions and social, economic, and political implications. The promises of progress and universal well-being must be accompanied by a critical assessment of their impacts and who benefits from their advances. However, this same analysis shows us that technological development is not a predetermined destiny, and that the conscious and organized action of society can positively influence its course.</p><p>The task at hand is to promote a social and democratic appropriation of science and technology, putting them at the service of building a more just, egalitarian, and ecologically sustainable society. May this article contribute to critical reflection and to the strengthening of initiatives that seek a future where technology, work, and social transformations in the contemporary era is, in fact, a tool for the common good and for the realization of human potential in its fullness.</p> <h3>References</h3> <ul> <li>Marx, K. (1867). <em>Das Kapital, Book I</em>. (Edition used).</li> <li>Engels, F. (1884). <em>The Origin of the Family, Private Property and the State</em>. (Edition used).</li> <li>Braverman, H. (1974). <em>Labor and Monopoly Capital: The Degradation of Work in the Twentieth Century</em>. Monthly Review Press.</li> <li>Huws, U. (2014). <em>Labor in the Global Digital Economy: The Cybertariat Comes of Age</em>. Monthly Review Press.</li> <li>Srnicek, N. (2017). <em>Platform Capitalism</em>. Polity Press.</li> </ul>]]></content:encoded>
      <pubDate>Tue, 25 Mar 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Tecnologia e Sociedade</category>
      <category>Transformação Digital</category>
      <category>Forças Produtivas</category>
      <category>Estrutura Social</category>
      <category>Inteligência Artificial</category>
      <category>Automação</category>
      <category>Futuro do Trabalho</category>
      <media:content url="https://i.postimg.cc/0jS3WM7L/Image-fx-13.png" medium="image"/>
    </item>
    <item>
      <title>The Platform Economy and the New Dynamics of Digital Work</title>
      <link>https://www.mmlo.app/en/blog/a-economia-de-plataformas-e-as-novas-dinamicas-do-trabalho-digital</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/a-economia-de-plataformas-e-as-novas-dinamicas-do-trabalho-digital</guid>
      <description>Investigating work arrangements mediated by digital platforms, their impact on labor rights, and the prospects for worker organization in this new landscape.</description>
      <content:encoded><![CDATA[<p><strong>Introduction:</strong> This analysis of "The Platform Economy and the New Dynamics of Digital Work" seeks to foster an in-depth reflection on the transformations that technology imposes on the social structure and contemporary power dynamics. Starting from a careful observation of the material reality, we aim to reveal the complexities inherent in technological development and how it affects living conditions and social interactions in different global contexts.</p> <p>Far from a simplistic or merely technical view, we understand that the platform economy and the new dynamics of digital work are a multifaceted field, where innovations, driven by human creativity and social needs, interact with existing economic and political structures. This article proposes to examine the implications of this dynamic, exploring both the constructive potential of technology and the challenges to its equitable and socially responsible use.</p> <h3>Historical Context and Foundations of Social Analysis</h3> <p>For a comprehensive analysis of "The Platform Economy and the New Dynamics of Digital Work," it is essential to use conceptual tools that allow us to understand reality in its totality and its transformations. The concept of <strong>the real subsumption of labor under capital through algorithms</strong> is fundamental to unraveling the logic underlying the observed phenomena. We will investigate how digital platforms, through their management algorithms, not only mediate but control and intensify the work process, dictating rhythms, prices, and conditions, in a new form of value extraction that resembles factory production but is spatially dispersed.</p> <p>Historically, the development of society's productive capacities has always been intrinsically linked to transformations in social relations. From the first technological revolutions to the current digital age, we have observed how different social groups seek to direct the means of production to serve certain interests. In the context of "the platform economy and the new dynamics of digital work," this manifests itself through the expansion of informal and precarious work, now mediated by technology, as a strategy to reduce costs and make labor relations more flexible on the part of capital, in a context of weakening unions and protective labor legislation in many countries.</p> <h3>Analysis of Contemporary Manifestations and Their Complexities</h3> <p>In the current scenario, "The Platform Economy and the New Dynamics of Digital Work" presents itself with multiple, often complex facets. On the one hand, digital platforms offer a segment of workers flexible hours and autonomy in deciding when and how to work, in addition to representing an immediate source of income in contexts of high unemployment or to supplement other activities. On the other hand, the way this technology is implemented and distributed can generate the absence of basic labor rights (vacations, paid leave, social security), the intensification of the workday, low pay per hour actually worked, the transfer of risks and costs to the worker (vehicle maintenance, mobile data), and constant algorithmic surveillance.</p> <p>We will examine concrete cases and data that illustrate this dynamic:</p><ul><li><strong>Case Study 1:</strong> The situation of app delivery drivers in large urban centers, who face long hours, exposure to the risk of accidents and violence, low pay, and little or no social protection, while platform companies profit from the high demand for delivery services.</li><li><strong>Case Study 2:</strong> Attempts by app drivers and delivery workers to organize into cooperatives or associations to negotiate better conditions with the platforms or create their own alternatives, facing challenges such as the dispersion of the category and repression by companies.</li><li><strong>Data Analysis:</strong> Academic research that demonstrates the decline in average hourly income of platform workers over time, the increase in reports of uncovered work accidents, and the high turnover of workers in these apps, indicating the unsustainability of the model in the long term for the worker.</li></ul><p>These examples demonstrate that technology is not an isolated factor, but an element that interacts deeply with existing social, economic, and political structures. The way "the platform economy and the new dynamics of digital work" develops and is applied reflects and, at the same time, influences these structures.</p> <h3>Perspectives for Socially Just Technological Development</h3> <p>Given the presented scenario, critical reflection is essential to guide action. It is important to identify paths to technological development that is socially just and promotes collective well-being. The pursuit of regulating work on digital platforms, guaranteeing labor and social security rights equivalent to those of other categories, establishing fair minimum wages, and limiting the power of algorithms in the management of work, with transparency and the possibility of contestation, is an important step to align technical progress with social progress.</p> <p>Experiences of draft laws under discussion in several countries to classify platform workers as employees, the creation of specific social protection funds for this category, and the encouragement of cooperative platforms that are controlled by the workers themselves, demonstrate that it is possible to build another technological future, guided by the needs of the majority and not by restricted interests. Citizen participation and the strengthening of democratic institutions are crucial to drive these transformations and for "the platform economy and the new dynamics of digital work" to serve human emancipation and the construction of a more equitable society.</p><p>The construction of governance mechanisms and popular participation capable of directing scientific and technological development to meet collective interests is a central challenge. This implies not only regulating the uses of technology, but also fostering a broad and informed public debate about its direction and purposes.</p> <h3>Conclusion: Towards a Conscious and Transformative Appropriation of Technology</h3><p>In summary, the analysis of "The Platform Economy and the New Dynamics of Digital Work" from a socially conscious perspective reveals that technology is a field of intense interactions and social, economic, and political implications. The promises of progress and universal well-being must be accompanied by a critical assessment of their impacts and who benefits from their advances. However, this same analysis shows us that technological development is not a predetermined destiny, and that the conscious and organized action of society can positively influence its course.</p><p>The task at hand is to promote a social and democratic appropriation of science and technology, putting them at the service of building a more just, egalitarian, and ecologically sustainable society. May this article contribute to critical reflection and to the strengthening of initiatives that seek a future where the platform economy and the new dynamics of digital work are, in fact, a tool for the common good and for the realization of human potential in its fullness.</p> <h3>References</h3> <ul> <li>Wood, A. J., Graham, M., Lehdonvirta, V., & Hjorth, I. (2019). Good Gig, Bad Gig: Autonomy and Algorithmic Control in the Global Gig Economy. <em>Work, Employment and Society, 33</em>(1), 56-75.</li> <li>Scholz, T. (2016). <em>Platform Cooperativism: Challenging the Corporate Sharing Economy</em>. Rosa Luxemburg Stiftung.</li> <li>De Stefano, V. (2016). The rise of the ‘just-in-time’ workforce: On-demand work, crowdwork and labour protection in the ‘gig-economy’. <em>International Labour Office, Conditions of Work and Employment Series, (71)</em>.</li> <li>Antunes, R. (2018). <em>O Privilégio da Servidão: o novo proletariado de serviços na era digital</em>. Boitempo Editorial.</li> <li>Lenin, V. I. (1916). <em>Imperialismo, Fase Superior do Capitalismo</em>. (Edition used to contextualize the expansion of new forms of exploitation).</li> </ul>]]></content:encoded>
      <pubDate>Thu, 20 Mar 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Economia de Plataforma</category>
      <category>Trabalho Digital</category>
      <category>Condições de Trabalho</category>
      <category>Precarização</category>
      <category>Plataformas Digitais</category>
      <media:content url="https://i.postimg.cc/QMF12RMr/Image-fx-14.png" medium="image"/>
    </item>
    <item>
      <title>Technological Autonomy and Equitable Development in the Global Digital Landscape</title>
      <link>https://www.mmlo.app/en/blog/autonomia-tecnologica-e-desenvolvimento-equitativo-no-cenario-global-digital</link>
      <guid isPermaLink="true">https://www.mmlo.app/en/blog/autonomia-tecnologica-e-desenvolvimento-equitativo-no-cenario-global-digital</guid>
      <description>Analyzing the challenges of technological dependence for developing nations and the importance of seeking alternatives that promote sovereignty and grassroots development in the digital age.</description>
      <content:encoded><![CDATA[<p><strong>Introduction:</strong> This analysis of "Technological Autonomy and Equitable Development in the Global Digital Landscape" seeks to foster an in-depth reflection on the transformations that technology imposes on the social structure and contemporary power dynamics. Based on a careful observation of material reality, we aim to reveal the complexities inherent in technological development and how it affects living conditions and social interactions in different global contexts.</p> <p>Far from a simplistic or merely technical view, we understand that technological autonomy and equitable development in the global digital landscape is a multifaceted field, where innovations, driven by human creativity and social needs, interact with existing economic and political structures. This article aims to examine the implications of this dynamic, exploring both the constructive potential of technology and the challenges to its equitable and socially responsible use.</p> <h3>Historical Context and Foundations of Social Analysis</h3> <p>For a comprehensive analysis of "Technological Autonomy and Equitable Development in the Global Digital Landscape," it is essential to use conceptual tools that allow us to understand reality in its totality and its transformations. The concept of <strong>technological dependence as a contemporary expression of center-periphery relations</strong> is fundamental to unraveling the logic underlying the observed phenomena. We will examine how the concentration of production and control of advanced digital technologies (hardware, software, platforms, data) in a few central countries perpetuates the subordination of peripheral countries, limiting their capacities for autonomous development and the sovereign appropriation of the benefits of digitalization.</p> <p>Historically, the development of society's productive capacities has always been intrinsically linked to transformations in social relations. From the first technological revolutions to the current digital age, we have observed how different social groups seek to direct the means of production to serve specific interests. In the context of "technological autonomy and equitable development in the global digital landscape," this manifests itself through the continuity of practices of conditioned technology transfer, the imposition of technological standards that benefit transnational corporations, and the difficulty of countries in the Global South in building robust and independent innovation ecosystems, reflecting historical dynamics of colonialism and imperialism.</p> <h3>Analysis of Contemporary Manifestations and Their Complexities</h3> <p>In the current scenario, "Technological Autonomy and Equitable Development in the Global Digital Landscape" presents itself with multiple facets, often complex. On the one hand, digital technology, if appropriated and developed locally, has the potential to boost strategic sectors, solve specific social problems (health, education, family farming), promote productive inclusion, and reduce historical inequalities in developing countries. On the other hand, the way this technology is implemented and distributed can generate brain drain, the remittance of profits abroad, vulnerability to sanctions and espionage, the inability to define technological priorities aligned with national needs, and the deepening of the international division of labor, with the periphery relegated to consumer of technology or supplier of raw data.</p> <p>We will examine concrete cases and data that illustrate this dynamic:</p><ul><li><strong>Case Study 1:</strong> The dependence of African countries on foreign payment platforms and cloud services, which generates high costs, limits local innovation, and exposes sensitive data to external jurisdictions, hindering the creation of robust national alternatives.</li><li><strong>Case Study 2:</strong> Initiatives for the development of free software and open hardware in countries such as Brazil and India, which seek to create technological solutions adapted to local realities and foster national technical training, facing challenges of funding and scale to compete with large corporations.</li><li><strong>Data Analysis:</strong> UNCTAD data on the flow of royalties and payments for intellectual property of technology from developing countries to developed countries, and the low percentage of patents registered by inventors from peripheral countries in the areas of advanced digital technologies.</li></ul><p>These examples demonstrate that technology is not an isolated factor, but an element that interacts deeply with existing social, economic, and political structures. The way "technological autonomy and equitable development in the global digital landscape" develops and is applied reflects and, at the same time, influences these structures.</p> <h3>Perspectives for Socially Just Technological Development</h3> <p>Given the presented framework, critical reflection is essential to guide action. It is important to identify the paths to technological development that is socially just and that promotes collective well-being. The pursuit of building national and regional technological sovereignty, through investment in local research and development, the training of qualified human resources, the creation of government procurement policies that prioritize national solutions, and the encouragement of South-South cooperation in science and technology, is an important step towards aligning technical progress with social progress.</p> <p>Experiences such as the creation of technology parks and incubators for technology-based companies focused on local needs, the establishment of public software and open government data policies, and active participation in international forums to defend fairer rules in technology trade and internet governance, demonstrate that it is possible to build another technological future, guided by the needs of the majority and not by restricted interests. Citizen participation and the strengthening of democratic institutions are crucial to drive these transformations and for "technological autonomy and equitable development in the global digital landscape" to serve human emancipation and the construction of a more equitable society.</p><p>The construction of governance and popular participation mechanisms capable of directing scientific and technological development to meet collective interests is a central challenge. This implies not only regulating the uses of technology, but also fostering a broad and informed public debate about its directions and purposes.</p> <h3>Conclusion: Towards a Conscious and Transformative Appropriation of Technology</h3><p>In summary, the analysis of "Technological Autonomy and Equitable Development in the Global Digital Landscape" from a socially conscious perspective reveals that technology is a field of intense social, economic, and political interactions and implications. The promises of progress and universal well-being must be accompanied by a critical assessment of their impacts and of who benefits from their advances. However, this same analysis shows us that technological development is not a predetermined destiny, and that the conscious and organized action of society can positively influence its course.</p><p>The task at hand is to promote a social and democratic appropriation of science and technology, putting them at the service of building a more just, egalitarian, and ecologically sustainable society. May this article contribute to critical reflection and to the strengthening of initiatives that seek a future where technological autonomy and equitable development in the global digital landscape is, in fact, a tool for the common good and for the realization of human potential in its fullness.</p> <h3>References</h3> <ul> <li>Fanon, F. (1961). <em>The Wretched of the Earth</em>. (Edition used, for the anti-colonial perspective).</li> <li>Santos, T. D. (1970). <em>The Structure of Dependence</em>. The American Economic Review, 60(2), 231-236.</li> <li>Couldry, N., & Mejias, U. A. (2019). <em>The Costs of Connection: How Data is Colonizing Human Life and Appropriating It for Capitalism</em>. Stanford University Press.</li> <li>Valle, C. (2020). <em>Digital Sovereignty: Challenges and Perspectives for Brazil</em>. (Fictitious publication for example, search for real references).</li> <li>Ho Chi Minh. (Date of relevant work). <em>On the Struggle for National Independence</em>. (Work relevant to the perspective of sovereignty).</li> </ul>]]></content:encoded>
      <pubDate>Sat, 15 Mar 2025 00:00:00 GMT</pubDate>
      <dc:creator>Matheus Mierzwa</dc:creator>
      <category>Soberania Digital</category>
      <category>Desenvolvimento Tecnológico</category>
      <category>Tecnologia e Desenvolvimento</category>
      <category>Autonomia Nacional</category>
      <category>Inovação Local</category>
      <media:content url="https://i.postimg.cc/XYbyHpRd/Image-fx-15.png" medium="image"/>
    </item>
  </channel>
</rss>
