A developer argues for writing code and tests instead of maintaining English-language specs when working with AI coding agents. The core idea: code is precise and self-documenting, while specs are imprecise and create a dual-maintenance burden. Tests serve as executable requirements, and by writing code to establish patterns, developers build a 'gradually expanding surface of trust' where agents can be given more autonomy in well-tested areas while humans stay hands-on at the frontier. The factory analogy frames the developer as a mechanic who must understand the system deeply to maintain quality, even as agents handle more of the work.
2.1K Impressions