You can version control a txt file or svg file, and they won’t fail to compile. I mean, Idk what your process is maybe you’re using IDEs that make it less painful than what I experienced, but the extensive control you supposedly have requires a lot more knowledge than LibreWriter.

I do most things in txt and svg then move them into libre writer and export the PDF. LaTeX is good to generate math images, but otherwise it’s too much work for things that really only need to be good enough.

LaTeX is not worth learning. Learn enough to insert the math, then use any other program.

It’s not that hard to use llama.cpp directly anyway. Why would I use a wrapper when I can just run a python script?

Integrate all into one app, and it’s close enough functionally that it could compete, and that’s all you really need.

I mean, I’m not in networking so maybe you’re right and I’m totally out of touch with the basics. Downloads don’t even need to be instant. People scroll and save videos to watch later all the time. Maybe I’m ignorant, but it seems like the technology (torrent networks) already exists and this just requires slapping a pretty interface on top of it.

Honestly, it sounds more like a marketing problem than a networking problem.

Creating an infrastructure that could potentially ingest terrabytes of data per second, and then processing it into multiple resolutions, is a massive ask.

Yeah, but that’s a bad approach.

A website that just hosts links, a torrent to video script, and a like-button that’s a magnetic link is all you need to set up a youtube competitor. The creators can host their own videos, and their fans can help them. Sure there’s a lot of quality of life stuff that integrating a torrent client and browser would help with like automatically using file selection to host only the resolutions you watch up to or automatically deleting videos after a certain time unless you’ve extra likes them, but the underlying system is functional. For steaming, do whatever peertube is doing.