AI Models Don't Know Their Own Updates: The ChatGPT Disconnect and a Proposed Sync Solution

The author notes that models like ChatGPT are unaware of their own shell updates. When ChatGPT Work is added, GPT itself doesn't know and must search, but search results only confirm the feature exists, not its capabilities or usage. Users perceive the model as incompetent because it lacks self-knowledge of its functions. The proposed solution is a mechanism where each app update includes a synchronized file or document that updates the model's knowledge, enabling it to know its new capabilities and help users learn them.