Review an AI pull request

Module 2 : Work with AI on a real codebase

Plan du cours

Cette leçon n'est pas encore dans votre langue, la voici donc en anglais.

Petit test

Guess first, no penalty. An agent opens a pull request for "add CSV export" with 1,900 changed lines across 31 files. What's the best first move?

A pull request, or PR, is a request to merge a set of changes into the main code, so teammates can review it first. AI makes big PRs cheap to produce. They're not cheap to review, and a reviewer facing 1,900 lines tends to approve and hope.

Keep AI pull requests small, with one job each. Then reject anything in them that doesn't belong:

  • Changes nobody asked for: reformatting whole files, renames, "while I was here" fixes.
  • New dependencies (outside packages) that weren't agreed.
  • Tests that got weaker: deleted, skipped, or with looser checks.
  • Lines the author can't explain. If you opened the PR, that includes you.

A good PR description tells the reviewer where to look hardest. Saying what the AI got wrong isn't a confession. It's evidence that a person read the code.

La suite de cette leçon fait partie de Learn Pro

Le premier module de chaque cours est gratuit. Learn Pro ouvre toutes les leçons de tous les cours pour 10 $ par mois. Tous les outils JobRise restent gratuits.