3 data-backed ideas to deter AI code quality problems
summary
Data from 250m+ lines across repos at Google Microsoft Meta and 1,000+ others shows three concrete risks of AI‑authored code: duplication surges (copy/paste overtakes refactor), churn rises ~50% and added lines exceed pre‑AI rates by >100%, even as median developer productivity inches ~+9%. The measurements come from Git commit counts plus provider suggestion/accept APIs (Copilot Cursor etc.). Each risk pairs with a mitigation: enforce stricter review gates, quota new code per feature, and track accepted suggestion rates as a KPI.
AI-written summary · about
chapters (10)
- 0:00AI code quality trade-offs
- 3:26Narrative vs data-backed AI
- 7:21Data antidote to AI hype
- 9:50AI code productivity trends
- 12:13AI productivity gains and risks
- 15:48Cost of AI productivity fire hose
- 18:17AI fuels code duplication
- 22:35Code review burden and AI sabotage
- 25:00AI adoption heuristics
- 26:04Summary and call to action
related talks
34mAI and Code Quality: Building a Synergy with Human Intelligence
21mAI Code Quality Debt: What Strong Engineering Teams Do Differently
19mWith great power comes great responsibility in code quality
19mThe AI Code Review Bottleneck: Escaping the Velocity Trap
46mCode Quality in the AI era, needed more than ever
43mWriting for Machines: How To Capture Your Project's "Vibe" and Survive AI Slop
01h 01m"Looks Good to Me": A Practical Guide to Handling AI-Generated Code
19mIf AI Writes the Code, Who Runs Production?