Ruin Your Systems By Making Them Perfect
summary
Argument: perfection is counterproductive and destroys systems; aim just above the user‑visible “magic line” and treat acceptable failure as an error budget. Define reliability with concrete SLI/SLO metrics, e.g. a 10‑errors-per-30‑day budget triggers a 30‑day feature freeze and focused reliability work. Expect humans to err (about 20% in simple repetition), use chaos engineering and fault injection to find emergent failures, and remember extra “nines” cost ~10x each; blame the system, not people, and contingency‑plan to detect and roll back fast.
AI-written summary · about
chapters (10)
- 0:00Perfection harms systems
- 4:53Human imperfection and reliability
- 15:43Error budgets and magic line
- 22:43Blame to systems thinking
- 28:40Reliability and speed coexist
- 31:50Companies as human systems
- 33:18Two-sentence success secret
- 35:01Q&A on error budgets
- 41:04Zero-defect marketing critique
- 48:24Contingency vs over-engineering
related talks
06mLightning Talk: Reliability Engineering for Humans
29mThe Curious SRE: Cultivating a Mindset for System Resilience
32mPsychologically Safe Reliability Management
40mPsychologically Safe Reliability Management
01h 00mHow to be WRONG
39mConfessions of a Systems Engineer: Learning from My 20+ Years of Failure
41mFailure is not an option. It's a fact
58mStop Trying to Fix Everything