Should we not treat LLMs more as a UX feature to interact with a domain specific model (highly contextual), rather than expecting LLMs to provide the intelligence needed for software to act as partner to Humans.