AI on a team

You are strictly accountable for the code the AI writes.

The idea

When you use AI to write a large Pull Request, you shift the burden of understanding from yourself onto the Reviewer. This ruins team velocity.

To collaborate effectively with AI on a team, you must Own the Diff. Your PR description must explicitly state what the AI generated, how you verified it, and which architectural boundaries it touches. "The AI wrote this" is never an excuse for a bug.

PR #102: Migrate auth to JWT (+450 / -120)

Description:

"Used Copilot to rewrite the auth flow to use JWT tokens. Looks good locally."

Reviewer's Brain:

Wait, +450 lines? Did you read all this?
Did it handle the refresh token race condition?
Did it leak any secrets in the logging?
Are the tests actually testing the new logic?

REVIEW BLOCKED. A sloppy PR description for an AI-generated diff makes the reviewer deeply suspicious. They have to read every line.

How it works (The AI PR Description)

# When submitting AI-assisted code, always include:

1. What was generated?
   "The JWT encoding logic and the middleware were generated by Claude."

2. How did you verify it?
   "I manually verified the refresh flow. I wrote the tests myself
   to ensure the AI didn't write tautological assertions."

3. What are you unsure of?
   "Can someone double check the error handling on line 45? The AI
   used a new pattern here I'm not 100% familiar with."