How Work Works
summary
Argues that self‑service clouds (AWS/Google) behave like an M/M/C queue where C→∞, explaining why they speed delivery, and that many platform teams are just a “half‑assed private cloud” that reintroduces human queues. Uses real numbers: a value‑stream example shows 47 days lead time with 5 days value lost to waiting. Claims halving batch size doubles throughput—batch size halving doubles throughput—and demonstrates a working agent‑based prototype (and cites Google’s generative agents paper) to simulate congestion collapse and team designs.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:19Technology trends and ThoughtWorks Radar
- 7:12Self-service cloud platforms
- 9:39Microservices, containers, JavaScript
- 12:30Queuing theory pitfalls
- 18:25Sensible defaults in engineering
- 22:18Canonical books and discipline
- 28:34Work flow and value mapping
- 37:26Generative science and agents
- 49:23Complex adaptive systems
- 53:12Simulating organizations
- 59:35Summary
related talks
50mHow Flow Works & Other Curiosities
42mHow Flow Works and other curiosities
50mHow Flow Works
53mHow Work Works
50mHow Flow Works and other Curiosities
01h 01mSoftware Architecture, Team Topologies & Complexity Science
56mDefending against CDD: Chaos-Driven Delivery
30mArchitecture & Teams at Less-than-Google Scale