The software industry: annealing, but wrong
Summary
The piece critiques the default preference for breaking changes into small PRs, using simulated annealing as a metaphor for software development. It argues that, in some contexts, large high-energy changes are necessary or beneficial, but require stronger processes, CI/CD automation, and thorough review. It includes real-world examples (Aperture) and contrasts fast-moving development with mature, risk-averse practices like those at Tailscale, advocating a balance between big jumps and smaller, well-reviewed steps based on context.