For comparison, I wonder how vulnerable Flathub (flatpak’s primary repo) is to these kinds of manipulations… Seems like every app manifest there is publicly available and is compiled on their servers, presumably making it easier to spot shady apps and updates, and the submission process requires manual approval.

Okay, the responses here are kinda disappointing because folks here seem to be unaware that (1) Mozilla has already added “AI” info Firefox a few versions ago (to provide machine translations of pages), and (2) the way they did it is very responsible (the whole thing is 100% local, no info is sent to other servers).

I understand that we’re all tired of this whole trend of language models being put where they don’t belong, but from what I see, Mozilla is actually the company I’d trust the most to do it right. (AFAIK, one area where the FOSS world is severely lacking and where Mozilla works to solve it is speech recognition with the Common Voice project, and if they start working on an LLM-based program to do that, I’d welcome it.)

Sounds cool, though I’m a bit confused as to why that is such a big priority given that ReactOS currently aims to replicate Windows NT 5.2 (XP x64 / Server 2003), which did not provide graphical set-up*…

* Technically all Windows versions up until, IIRC, Vista had their install process in two stages: a text-based stage where you’d input the most basic info (what filesystem to install onto, what Windows directory to use, etc.) and a graphical stage once the basic files are installed (where you’d be asked what devices the computer has, whether it’s networked, date/time, etc.). From Vista to the present day, the first stage is graphical as well. ReactOS’ latest release uses the pre-Vista model, but the latest blog posts indicate a move to the more modern one.

You can use notification settings to “Minimize” any unwanted permanent notifications – in that way they’ll not show an icon in the tray area. (You can also just disable any notification type, but Android is more likely to stop any background task that doesn’t display a notification.)

If you’re learning Japanese, then “10ten” is very good. It adds a little “puck” you can use to hover over words and phrases to see their dictionary definitions, readings, etc.

(On desktop, it instead works whenever yoy hover your mouse cursor over a word, but on mobile, that’s not a thing. Either way, it’s easy to turn on/off based on your need.)

It seems like on Invidious, the default setting is to still have the end user load the video directly from YouTube, whereas Piped defaults to proxying the video through its server as well. I would imagine this makes Piped servers a lot more noticable to YouTube.