That's awfully pessimistic. I don't think the 2nd option is doomed however (using only "sane" software). This is indeed elitist, but I feel like at this point, this is required to run quality software. If you can't use a computer, you use multi layered software that makes the UX easier. Same goes for cars, agriculture or coffee machines, both have "elitists" users, that just want to "do things right" rather than easily. We take more time to do things, but with higher quality, and fiability. If some people want high quality stuff, I can help them, so they don't have to bother.

As an example, I'm part of an association for local food delivery in my town, and I run some opensource software to help with organization. I run it on OpenBSD for security, and make backups everyday. I settled on a simple installation because I know how to run this. Otherwise, they would have needed to run the #1 on the market which is much more expensive and complex.

That reminds me of Jon Blow’s talk, which has a lot of similar themes around complexity growing too quickly for anyone to keep track. He’s also especially worried that we’ll lose the knowledge of the lower levels, as new programmers gravitate towards the top levels (e.g. everyone builds apps instead of going into OS or browser engines).

I’m not sure I subscribe to the “everything will collapse” scenario, but I think there might come a time where we will essentially get stuck. Maybe it already happened, Electron apps are cool because they can do all kinds of flashy things very easily that would take a lot of work to build otherwise. If you don’t like Electron, you’re often stuck with a UX of the past, and the common user just doesn’t make that trade off, so software houses don’t.