It feels like this is the required level of speed-up needed re. time-to-first-token to make continuous vision useful for on-device applications like an assistan

It feels like this is the required level of speed-up needed re. time-to-first-token to make continuous vision useful for on-device applications like an assistant that can see and take action on your screen, ala the original Apple Intelligence demos. It’s very impressive seeing the app in the repo and I’m excited to build it tonight and play around.