![](https://slrpnk.net/pictrs/image/07212002-1220-41ab-bcc9-315e542e64f5.webp)
![](https://fry.gs/pictrs/image/c6832070-8625-4688-b9e5-5d519541e092.png)
I mean the biggest innovation here isn’t the “AI” (partially “on-device” or otherwise). It’s exposing the apps action hooks to the model.
I mean the biggest innovation here isn’t the “AI” (partially “on-device” or otherwise). It’s exposing the apps action hooks to the model.
This isn’t actually innovative it is just not OpenAIs business model. This recent trend is honestly a much smaller blip than most people recognize.
That’s how mastodon works now and i think it’s worthwhile as a basic feature.
Let’s be honest. They certainly plan to, but first they’re gonna see if saying “Apple Intelligence” a bunch is going to convince people they actually did something innovative.
It’s not that they don’t want an algirithm. They just don’t want one that’s profit motivated.
Are the apps not compatible?
They can always claim the second even when it’s the first.
Love to see unique features for piefed. So many different ways to tackle “threadiverse” as a project and I think we’re all better off with some experimentation
If anybody’s interested in signing up for the rolling beta, here’s the link
If anybody’s interested in signing up for the rolling beta, here’s the link
Seems like it’s less a problem with pixelfed and more an issue with pixelfed.social?
The major difference is that it’s modular. So you can pick and choose which traditional social network features you want to use.
Unmanic to optimize your library in the background. Encoding things to x265 can buy you a huge amount of space.
Edit: Reading again i see that you’re on a pi. Not at all sure what the video encoding performance is on those.
I was having similar issues with the zigbee ones, and I believe it was a known issue with the way transitions were implemented. The solution ended up being to set short transition times across the board. Not sure if that’s applicable here, but it may be worth a try.
Google’s trash now
You need this guy: https://github.com/music-assistant/hass-music-assistant
I honestly don’t remember but I do recall it’s way more of a process than it used to be
Pretty utilitarian on the ol thinkpad
More granular subcategories are absolutely not an answer to this problem that I’m satisfied with. I don’t mind having a handful of timelines to check but if I’m getting into double digits there’s a problem.
I dunno if ‘Siri but functional’ is good enough to get points from me. That said this is definitely good for the industry