Tracking AI generated code with Git
summary
An open source git extension called Git AI attaches git notes to commits and records which prompt generated which lines, preserving that metadata through merges, rebases, resets and cherry‑picks. It integrates with agents like Cursor to mark commits as AI written, reports mismatched metrics (agents claiming 45,000 lines vs 6,000 actually merged), captures prompts, and shows one project’s rewrite rate dropping from about 60% to ~20% as prompting gets more iterative.
AI-written summary · about
chapters (7)
- 0:00The challenge of AI coding
- 1:30Limitations in tracking data
- 3:02Introducing Git AI
- 5:44How Git AI works
- 8:01Value and insights gained
- 9:31Looking ahead
- 11:13Q&A and discussion
related talks
15mAI's Impact on Development, Post-Production & Support Journeys
26mAI, You complete me
25mFrom AI Chaos to Production Ready: Using AI Agents to Improve Your Code Quality
25mThis Lying Has To Stop: Keeping AI Honest with OpenTelemetry
06mDon’t get one-shotted: Use AI to test, review, merge, and deploy code
26mAI Agents as Your Co-Developer
23mAI-Generated Code: Who Is Responsible for the Bugs?
24mDevelopers and AI Enemy or Frenemy Leveraging AI and Atlassian to empower developers