Why Software Factories Fail: Harness Engineering Is Not Enough
Why Software Factories Fail (or: harness engineering is not enough)
I argue that the current rush to build AI-driven software factories is failing because companies rely too heavily on loop engineering while ignoring fundamental model training limitations. Data from Faros AI shows rising incidents and declining code quality, proving that simply generating more code without human oversight creates chaos. The solution isn't just better prompts or more tokens, but a shift away from the illusion that AI can fully replace human alignment in complex, long-term codebases.
A developer vibe-coding a side project a dozen people will ever run, and a team keeping a ten-year-old enterprise system alive for another quarter, share almost no constraints worth naming, and most of the advice in circulation is really one of those two people telling the other how to live.