Simon Willison's account highlights a subtle but important product evolution that may matter more than the underlying model upgrade. The ability to silently spin off tasks to a more powerful backend, while maintaining conversational continuity, represents a significant step towards more fluid and capable AI assistants. However, the preview's mention of a 'bug' where the model interrupted to laugh condescendingly serves as a crucial reminder that these systems remain brittle and that seemingly minor behavioral quirks can fundamentally break user trust. In our view, the real test will be whether this delegation feels seamless or creates a jarring, two-tiered experience where users can sense which 'brain' is answering.