AI tools are powerful, capable of handling deep implementation work, refactoring of entire subsystems, and catch things you'd miss at 2am. Signet is built with AI tools. AI helps us translate ideas and concepts into code faster than we ever could alone, and we'd be fools not to use it.
However, the human behind the tool is responsible for understanding every decision, every architectural choice and why the implementation of the code works the way that it does. And moreover, how their code fits into the broader codebase as a whole. Powerful AI tools are not an excuse to stop thinking.
If you can't explain why a change exists, how it works, and what breaks if you remove it, it doesn't belong in your PR. Full stop. You are already screwed, and should open an issue or start a discussion instead of burning your tokens.
Contributors should use AI freely. Claude Code, Cursor, Codex, anything is fine.
But for the love of god, AI is your hands, not your brain.
Read what it writes and Disclose it. All AI usage must be disclosed in your PR.
Not because it's shameful, but because it's honest and helps maintainers
understand contributions better. Include Assisted-by tags in your commit
messages:
Assisted-by: Claude-Code:claude-opus-4-6
Assisted-by: Cursor:claude-sonnet-4-5 biome
Every PR is reviewed by humans. Sending unreviewed AI output to a maintainer wastes their time validating work you didn't bother to understand. That's not a contribution, that's a chore with extra steps.
Just because your AI is good at coding doesn't mean you get to shut your brain off and let it freelance in your codebase like an idiot. PRs that are clearly unreviewed slop will be closed. Repeat offenders will be blocked. This project builds AI tooling — we know exactly what these tools can and can't do. They are not a substitute for thinking.