Doing Things the Hard Way
summary
Company grows from ~25 to ~215 people and engineers from 8 to 60, and three concrete failure modes emerge: splitting product and infra early breaks the feedback loop (fix: move to a mono repo, improve docs, coach devs, never give one person sole pager ownership), chasing shiny hard problems leads to underinvested observability, and an “everything” rewrite — a custom container/orchestration stack built then discarded — wastes effort versus focused work on reliable deployments; a built 10–15 second failover automator proves lower ROI than better metrics and monitoring.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 1:52Talk biases
- 2:36Early division of product and infra
- 7:53Distracted by hard problems
- 14:36The everything project
- 18:50Common themes and conclusion
