From Vibes to Outages: Riding the AI Code Wave
summary
AI-generated code already makes up a large share of commits (Cursor claims one billion lines accepted a day; Stack Overflow reports 84% adoption), and teams see higher churn and more builds (+40% builds, +15% commits, +25% completed tasks). Real examples: hallucinated or slopsquatting dependencies, AI-written tests that encode broken logic, emojis breaking builds, an agent that deleted files. Tools that help: mutation testing with LLMs (9,000 mutants, 73% tests accepted), MCP-style context plumbing, and postmortem automation cutting troubleshooting time ~20–30%. Strategies focus on lowering failure probability via tests, deployment guardrails, and observability.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 2:55AI adoption levels
- 7:31AI impact on incidents
- 12:13AI failure scenarios
- 18:32AI reliability culture
- 22:10Reducing failure with AI
- 28:01AI for SRE productivity
- 30:30Conclusion
related talks
50mPanel: Beyond The Vibe: Code Quality First
43mWriting for Machines: How To Capture Your Project's "Vibe" and Survive AI Slop
33mShipping Fast, Breaking Silently: Why AI-Generated Code Needs a Second Opinion
30mThe AI Revolution in Coding
23mVibes Not Vulns: Securing the Era of AI-Written Software
32mEscape the Try-Again Loop: Verifying AI-Generated Code
19mIf AI Writes the Code, Who Runs Production?
26mAI Agents as Your Co-Developer