How Work Works
summary
Ten years of trends condense into a claim: simulate organisations with agent-based models plus LLMs to predict flow and de-risk changes. Examples: Thoughtworks radar shows ~300 new items per meeting; microservices (2012), Docker (2014), Kubernetes (2015). Queueing math (Little’s law) and Don Reinertsen’s principles show that reduce batch size (ideally single-piece flow) cuts cycle time dramatically and doubles throughput when halving batch size. Buildable tools: NetLogo/Netlogo-scale sims, a Rust epidemiology framework used in India, and Google’s generative agents paper integrating agents with LLMs. Measure queue size not cycle time.
AI-written summary · about
chapters (7)
- 0:00Intro
- 0:45Programming in the past 10 years
- 8:51What does good look like?
- 18:05A brief interlude involving turtles
- 39:44Another interlude
- 46:52How work works
- 51:55Outro
