Accelerate Code Quality in the AI Age
summary
Data come from 100k AI refactorings and a benchmark that shows LLMs almost always produce syntactically valid code but only objectively improve code about one third of the time and preserve behavior in roughly one third to one half of cases, so many suggestions break tests. A code Health metric shows green files are ~10x faster to change than red ones. Adding model selection plus validation yields 97–98% correctness, enabling automated refactorings that can safely pay down a large slice of technical debt.
AI-written summary · about
chapters (15)
- 0:00AI and code quality intro
- 2:3155% speed claim limits
- 5:43Faster code quality risks
- 7:51AI for code understanding
- 10:58Refactoring vs AI refactoring
- 13:27Measuring code quality
- 17:20Safe AI refactorings
- 19:49Live AI refactoring demo
- 23:41Factchecking AI refactorings
- 25:25Risks in automated refactoring
- 27:54AI learning importance
- 29:18Speed vs technical debt
- 30:41Guardrails for AI coding
- 31:45AI for code understanding
- 32:27Q&A
related talks
40mCode Rot at Machine Speed: Preventing AI-Induced Technical Debt
46mCode Quality in the AI era, needed more than ever
21mAI Code Quality Debt: What Strong Engineering Teams Do Differently
46mClean Code in AI era
49mClean Code in the AI era
34mAI and Code Quality: Building a Synergy with Human Intelligence
45mAgentic AI Coding Practices for Speed with Quality
36mThe Code Quality Advantage: How Empirical Data Shatters the Speed vs Quality Myth