Agent Systems
Loop Engineering
Bounded state machines, safe effects, recovery, and replay for long-running AI work.
8 chapters · 5-7 hours reading · 2-4 hours labs · 5-8 hours project
Chapters
- 1Loops turn observations into controlled progressChapter 1
- 2Specify states, transitions, safety, and livenessChapter 2
- 3Make history replayable and upgrades explainableChapter 3
- 4Make effects safe under duplicate and ambiguous deliveryChapter 4
- 5Budget retries and detect loops that cannot convergeChapter 5
- 6Coordinate durable work without hiding concurrencyChapter 6
- 7Test transitions, histories, faults, and propertiesChapter 7
- 8Operate loops with traceable authority and safe controlsChapter 8