I was wrong about AI assisted coding... (well, maybe not entirely)
Originally published on Impl — Jaroslaw Michalik's newsletter on engineering, AI, and building software that lasts.
The assumption was that better context — more docs, better prompts, richer tooling — would be what separated teams that use AI well from teams that don't. It's a factor, but not the deciding one. Teams with strong engineering culture already in place — real code review, agreed standards, clear ownership — get far more out of AI-assisted coding than teams that hoped the tooling would substitute for that discipline.