[JumpCrisscross]

> Would it not make more sense to use OpenAI LLMs to be hosted away from these device

The lag is annoying. I don’t remember Ive’s position on privacy, but shipping everything off to the cloud isn’t great for that.

There might be satisfaction in having a childlike AI running on device who can call grandma—or whatever-when it needs help. One could tune it to different levels of spontaneity (“that’s a Magpie, in case you’re interested”) and proactiveness (“I noticed a new restaurant nearby and you haven’t eaten all day, so I grabbed a placeholder reservation”) that shouldn’t require a lot of phoning home to the mothership.