That is true, and to me the test of any ai hype has been “ok, let’s do some work with this”, usually leading to readjusted expectations. But ChatGPT was the fir

That is true, and to me the test of any ai hype has been “ok, let’s do some work with this”, usually leading to readjusted expectations. But ChatGPT was the first that gave a glimpse of passing this test, and I could imagine a successor that ran entirely locally on device and had native access to your data (assume privacy and anonymity are solved). You could essentially build applets by taking to this ai, and it would be centered around your data. If you don’t like what it’s doing then you talk to the ai to modify it. Each applet has a forked instance of the ai and is snapshotted, so it keeps context permanently. E.g “a run tracker: I just went for a run, show the route I just took. Graph incline vs my hear rate. Stitch together a street view version of my run, and animate any frames missing with your best guess. Make sure the weather matches today’s, here’s a picture of what it was like outside (snaps photo). Now show this run side by side with yesterday’s. (Etc…)” Idk, I could imagine it being a big deal. Although I have no idea how portability between applets could work. Like, if I make an applet it would be the only one in existence like it, so any networked communication between applets with interfaces would be weird/dangerous. I guess if you just stick to portable formats (text, tables, photos, etc…) and just throw it on the other side of network, you could let the recipient deal with parsing it. Anyways, something like this gets my imagination turning, and I think this + AR is going to be a new paradigm for sure. Hopefully we get the economics right this time around.