It's a great tool for the right tasks. What's annoying is that it's marketed as being a great tool for tasks it can barely do.

It has really sped up the process of writing things in languages I'm unfamiliar with. All the stupid little mistakes it will find much faster than trying to google them. As long as you're critical of the answers I also found it pretty good at explaining how to do things. It will often get some details wrong but as long as you have general programming ability and access to documentation you can usually figure those out somewhat easily.

I doubt anyone here hates AI other than for the big companies pushing it constantly. ML and language models have been a thing since the last decade but we only hate them now cause of how desperate the corpos are about it and oh the data scraping too but that's expected.

For me it helped with:

• Fixing electron apps not launching in wayland
• Fixed my ignis shell code(ricing, these shells keep changing and it's a burden maintaining them)
• automating my system with a few systemd timers.

Albeit I could do these myself by looking at docs but it's not worth the time. Now it just works instead of "maintaining my arch setup".

I felt the same way regarding AI. I always felt like it was too unreliable, and while you obviously shouldn't trust it blindly, it has been doing a pretty good job explaining things to me. I've been using a local model with ollama, i think it's called qwen2.5 coder or something, and it has been helpfull explaining some NixOS stuff to me, where i couldn't find it myself (cause NixOS documentation is all over the place). I also use it for explaining some bash stuff to me sometimes when i'm writing a script. Usually i don't just care about the answer but i also want to learn why something works if it isn't clear to me, and so far it also has done a good job explaining how it works when i ask it a follow up question.