Apple Intelligence + Social Media thought: Do you think that "the algorithm" will move from the server to the edges? 

Does an open network like Farcaster have an advantage in this direction? What would it take for a native iOS client to leverage it?

Dad, geek, conversation liquidity provider. vrypan.net (home), @l--o--l (fun). devcoin:$lemon3

I think onboard capable models and compute are going to make huge impacts. I've noticed more devices starting to incorporate "what's on the screen" to supply information for suggested typing, and having direct access to scrollable lists of information at a view level (text, image elements, not just composited screen info) even in traditionally closed products you don't have the total information might offer some improvements in that.

Models seem like they're a lot easier to access and do inference with these days as well, with tools and support for these kinds of devices being a bit more portable to the standard formats you might expect on them

What if your device downloaded your interactions and replies from a hub, re-ranked the users and types of data on device when you charged it at night, and then used recent time-frame posts as well as some other local view / interaction metrics throughout the day polling for new data since you last opened the app