Taming the Unpredictable: Reliability in Chaos
summary
Claim: AI-generated code will explode system size and unpredictability, forcing a shift from point-in-time reviews to continuous experiments and risk-first design. Empirical notes: models benchmark at 1.3% hallucination rate (others near 10%), agents already found and fixed a deadlock and bulk-updated outdated playbooks, and teams implement continuous cluster create register tear down checks as architectural fitness functions. Practical mitigations: codified rollback, live migration/drain, randomized UI “hopper” testing, and automated anti-pattern searches for retries, backoff, and concurrency bugs.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 3:25Jevons paradox in software
- 7:16Automation ironies and skills decay
- 11:04Curse of knowledge in engineering
- 15:20Complexity and chaos in systems
- 20:55From code to systems thinking
- 25:31Generic mitigations and remediation
- 30:01Rethinking prioritization and experiments
- 35:57Feedback loops for complexity
- 39:53Architectural fitness functions
- 41:19Future of SRE in AI world
related talks
09mAI and Chaos Engr: Smarter Failure Testing for Resilient Systems
32mEscape the Try-Again Loop: Verifying AI-Generated Code
27mChaos Engineering: The Path to Reliability
25mWhen Your "New Teammate" Hallucinates: Observability as the Safety Net for AI Code
26mBreaking Systems for Resilience
18mThe Failures That Don't Crash: MLOps for AI Agents
30mArchitecting Integrity: Responsible Software Evolution in the AI Era
35mManaging the AI Surge: Governing & Scaling in the AI Era