Move Slow and Mend Things
summary
Argues that speed without direction wastes effort: velocity is displacement with direction not just speed, so use OODA/PDCA loops to treat architecture and features as hypotheses to test. Treat the codebase as codified knowledge to be updated, not a production line; collective knowledge and communication, not headcount, change throughput. Gives concrete failures: Facebook’s emotional‑contagion experiment (N ≈ 700,000) and the ExoMars Schiaparelli lander crash caused by an IMU angle error, showing simple checks and testing prevent catastrophic faults.
AI-written summary · about
chapters (21)
- 0:00Introduction to speed
- 1:32The myth of fast and break
- 3:20Speed vs. velocity
- 5:32Measuring software work
- 8:47Feedback loops
- 11:22Planning vs. doing
- 13:21The history of software
- 16:17Code as knowledge
- 19:06Knowledge acquisition
- 22:30Architecture and teams
- 24:28Wisdom of crowds
- 26:14The enterprise trap
- 28:25Communication overhead
- 30:53The myth of typing
- 33:41Contextual speed
- 36:00Lessons from failure
- 37:44Debugging the system
- 40:46Collecting failures
- 42:23High-stakes anomalies
- 45:02Social consequences
- 46:59Final thoughts on learning
