Don't Follow Leaders or "All Models Are Wrong (and So Am I)"
summary
Five years on, the SRE book is credited with spreading ideas but also fossilizing practice: widespread copying produces stalled innovation. Claims: SLOs and error budgets work socially but fail to model customer impact (100‑minute outage vs many one‑minute outages) and break for non‑request/response services. There is no numeric model of reliability that trades off feature velocity, simplicity, revenue and risk. The field needs empirical studies, new priors for prioritization, and fresh orgs to fund and coordinate that research, not just on‑call debates.
AI-written summary · about
chapters (18)
- 0:00THE ADVANCED COMPUTING SYSTEMS ASSOCIATION
- 1:04Some background re: the SRE Book
- 2:08Doing the book required a new model for working
- 4:09Impact of SRE book
- 5:33Criticisms of the book at the time
- 10:46Legacy of the SRE Book
- 13:44The value of reliability
- 19:11The parable of the sticky users
- 22:01The sparse hints we have
- 29:15Questionable Assumptions
- 31:45Homogenous independent requests good, two legs bad
- 33:09A horse-sized duck, or 100 duck-sized horses
- 34:29Error budget bankruptcy
- 35:31Book originated patterns we could re-examine
- 36:22How important should on- call be to SRE?
- 39:03Integrating Safety Cultures
- 40:59Ethics in SRE
- 43:59Call to Action
related talks
05mSLOs: You’re missing the point!
43mReliability Equilibrium: The Hidden Playbook behind SRE Influence
43mThe SRE I Aspire to Be
29mThe Curious SRE: Cultivating a Mindset for System Resilience
30mThe 'Success' in SRE Is Silent
37mReal Talk: What We Think We Know — That Just Ain’t So
35mThe Best SREs Seem to Be the Ones without an SRE Title—And What We Can Do
47mPersistent SRE Antipatterns: Pitfalls on the Road to Creating a Successful