[twiceaday]

That's only one factor. I think the thing that is going to trigger a phase shift when it comes to interacting with LLMs is reducing response latency to conversational ranges by running enough of the models offline. You say something and it starts responding immediately, you interrupt it mid-sentence and it adjusts immediately. That is going to be revolutionary.

[galaxyLogic]

Also a needed feature is speech-understanding, which requires a microphone, which everybody has in their gadget.