Safety Properties in Distributed Systems
This day establishes core safety definitions such as linearizability and fault tolerance that underpin reliable AI infrastructure at scale. It matters in the arc because phase-4 focuses on preventing silent failures before any production rollout of tools like StartupTribunal. Learners must internalize these invariants now to evaluate later design choices against real failure modes.
Resources
- 25 min
Deliverable
Commit a 300-word journal entry to docs/day-138-safety-journal.md that names two safety properties and maps each to one file in app/maku/
Quiz · 3 questions
1. Which property is violated when two replicas return different answers for the same key at the same logical time?
2. State the difference between safety and liveness in one sentence.
3. Name one safety property relevant to rate-limiter.ts and explain a concrete failure scenario if it is broken.