Code Quality in the AI era, needed more than ever
summary
Companies lose $2 trillion yearly to poor code and developers spend about four days per month fixing bugs; Sonar/SonarQube telemetry finds millions of issues across 35+ languages and static analyzers drive 6,000+ rules. Generative AI often misleads—developers discard ~60% of suggestions, Copilot answers are only 29% correct and complex tasks introduce ~70% issues—so treat AI like an intern, run linters/quality gates (SonarLint, SonarCloud), prefer language features (Java records, virtual threads, collectors) and apply Clean As You Code by focusing on new code.
AI-written summary · about
chapters (10)
- 0:00Cost of poor code quality
- 3:38SonarQube and free tiers
- 8:53Code quality struggles
- 13:11AI assistants reliability
- 19:58AI as intern not replacement
- 23:57Common vulnerabilities
- 30:45Cognitive vs cyclomatic complexity
- 35:23Java best practices
- 40:45Ship of Theseus decay
- 44:01Key takeaways
related talks
49mClean Code in the AI era
46mClean Code in AI era
19mWith great power comes great responsibility in code quality
34mBuilding Guardrails for AI Coding Systems
34mAI and Code Quality: Building a Synergy with Human Intelligence
43mClean code, is it really worth it? A story about monsters and...
41mCode Security Reinvented Navigating the era of AI
21mAI Code Quality Debt: What Strong Engineering Teams Do Differently