Safe Systems in a Chaotic World
summary
Argues that safety is constraints not tests and demonstrates a practical method: use STAMP/STPA to find hazards, model controllers and unsafe control actions, then turn them into enforceable rules. Shows concrete tools — pokayoke/mistake‑proofing, property‑based tests (QuickCheck style), CQRS/event sourcing, Terraform/OpenTofu, WireGuard and a distributed Beam cluster — used to ship a clinical app in 100 days that stays live across clouds, runs millions of tiny Erlang processes, contains LLM risk, and is loved by clinicians.
AI-written summary · about
chapters (18)
- 0:00Introduction and welcome
- 1:59Chaotic times in software
- 8:50Courage in systems theory
- 9:16Building safe spaces with invariance
- 10:42Radiation therapy machine failure
- 12:49STAMP and STPA overview
- 16:21Systems theory fundamentals
- 23:48Poka-yoke mistake proofing
- 33:02AI and vibe coding critique
- 36:58STPA safety constraints
- 43:30Property testing and constraints
- 45:37Erlang VM and reliability
- 50:06Let it crash philosophy
- 51:31Gleam and compile-time safety
- 52:54Case study clinical system
- 57:48Constraints and team trust
- 59:15Tools and resources summary
- 1:01:00Q&A and closing
