The cost YAGNI was never about

The pushback was not that speculative code is free. It was that Beck’s advice becomes brittle once the domain has real constraints that are knowable up front. People pointed to systems with table-stakes requirements, long-lived data models, and hardware-like delivery costs where "we’ll add it later" is often how teams paint themselves into a corner. In those environments, the failure mode is not over-abstraction. It is under-building around requirements that were predictable if engineers had listened to users or domain experts. Several comments framed YAGNI less as a law and more as a bet about how forecastable your domain is and how expensive change will be. AI sat underneath much of the conversation without producing a clean consensus. The optimistic view was that AI lowers the cost of restructuring, test shoring, and migration work, so the premium on keeping code safely changeable has gone up. The skeptical view was that AI mostly accelerates code generation, not the hard parts of shipping with confidence, and can just as easily multiply debt when weak teams let agents spray code into a fragile architecture. That left a practical middle ground. The old tradeoff has not disappeared. Fast edits matter less than preserving trust in running software and maintaining a codebase that can absorb change without turning every future request into a rewrite.