AI generates code at a pace that traditional git workflows cannot handle. I learned this when 3 AI sessions created 47 files and a merge conflict that took 2 hours to resolve.
How do you structure branches for AI coding?
One branch per AI session. Before starting an AI session, create a branch: git checkout -b ai/session-description. The AI generates all code in this branch. If the output is good, merge. If bad, delete the branch. This keeps your main branch clean and makes reversion instant.
How do you handle AI-generated merge conflicts?
Do not resolve merge conflicts manually for AI code. Delete the conflicted branch and regenerate the feature with a more specific spec. AI code is cheap to generate and expensive to manually merge. I regenerate instead of resolving and it saves 80% of my merge time.
Leave a Reply