Summary
When a team uses AI-assisted coding, the review discipline must change. Reviewers cannot rely on 'the author must have thought this through' when the author is a language model. Explicit AI-review norms are the difference between productive teams and drifting ones.
Objectives
- 01State the changes a team should make to code review when AI is involved.
- 02Explain why author responsibility remains unchanged.
- 03Design a lightweight AI-attribution policy.
The Lesson
What changes
Higher scrutiny of API correctness — the model may have fabricated it. Higher scrutiny of edge cases — the model tests the happy path. Explicit review of security-sensitive code paths. Explicit review of tests, which are often as generated as the code.
Key Ideas
- AI use raises reviewer responsibility, not lowers it.
- The author owns the code, whoever typed it.