A guy at a diner in Toledo showed me his phone running a local model, and it made me rethink the whole cloud thing
Three weeks ago I was eating breakfast at a diner off I-75 in Toledo and the guy in the next booth saw me squinting at a map on my laptop and struck up a chat. When I said I hated how slow the signal was out there, he flipped his phone around and showed me a small model running right on the device, no bars needed, answering his questions in about two seconds flat. He said he loaded the whole thing at home over wifi and it just lives there now, no monthly fee, no sending his questions off to some server farm. I sat there kind of stunned because I always figured you needed a big connection and big hardware for anything useful like that. He even showed me it worked in airplane mode, which is wild to me. So my question for you all is, has anyone here messed with running models locally like that, and what do you use for it?