I started Forge Code with a simple idea: using AI to plan projects. I soon saw the problem: AI "agrees" with everything, and technical rigor gets lost along the way.
So the project changed direction and became an evidence-driven development system: a guide for taking a project from zero to done, with every step traceable.
The premise is uncomfortable: if there's no verifiable evidence, there's no valid conclusion.
How it works:
1. Numbered steps: each step has an explicit goal and mandatory tests.
2. Evidence as a contract: the developer runs the tests and attaches the evidence (logs, screenshots, outputs).
3. AI as auditor: the AI doesn't give opinions. It checks whether the attached evidence proves the completion criteria, and the answer is binary: ACCEPTED or REJECTED.
That way, "works on my machine" and quick workarounds stop counting as done. Forge Code formalizes the method: demand proof.
The project is open source: the code and the process are open to anyone who wants to audit them.