For years, AI progress has been measured by output quality. If a model sounds intelligent, we assume the system behind it is intelligent. LLMs exposed the flaw in that assumption: Fluency is not continuity. Output is not identity. A conversat...