Vibe-Coded Releash chat screen collapses in 5 months, rebuilt as throwaway

The Releash case is a central piece of evidence, and its timeline is striking: the chat screen was built over five months, then took five more weeks just to diagnose and replace — a period during which no other changes could be made. The author attributes the collapse not to AI's coding ability but to the absence of an expressed domain: the screen dealt with the SDK provider's internal concepts, which the developer did not understand and therefore never defined.

The no9desk example extends the argument to a business setting. What began as a simple copy of a core-system list gained a provisional-data workaround for advance distribution, then a merge function for duplicated data, and finally a manual morning reconciliation task. The application types grew from three to nine and releases from 185 to 788 in four months, yet the core goal — reducing manual effort — was not achieved.

The conclusion is that Vibe Coding is acceptable only where the system can be thrown away entirely. For tools that become the sole entry point for a department's work, the author argues the eventual cost of rebuilding — the principal — is far larger than paying it upfront, because features from conflicting requirements, bug fixes, and unnecessary additions all coexist and must be untangled.