I’ve had several occasions in my life where I’ve tried to switch to Linux only to run into hardware incompatibilities that would have taken me man-months to solve. Thanks to GPT my primary personal PC, home media and project server is now running Linux and fully functional. I’ve got Jellyfin, NextCloud, Navidrome and Joplin all happy. I’ve also been able to wipe an obsolete MacOS off my 2017-era Macbook Pro and it’s happily running Mint.
There’s a deep irony in Windows being enshittified by AI, while the same AI makes Linux about a billion times more usable. Now I have a leftover Windows micro-PC running a local LLM model, so I can be less dependent on GPT, as long as time isn’t a factor…
You must log in or # to comment.
I have used AI a little for Linux adaptation but I do want to caution overuse since it’ll sometimes give you really bad advice, like regarding important dependencies and such. Sometimes it cites info for a wrong distro or outdated info, so it’s not so much made up BS, just bad search results essentially.
Nothing ironic about that. Thanks to OpenClaw and open weights AI models I started self hosting and figuring stuff out that would have made me frustrated and quit years ago. I have home assistant running on bare metal on a mini PC, a NAS with custom bios, migrated everything away from windows, Kavita for ebooks, jellyfin for music and video, nightscout to ingest my wife’s insulin pump data, and some fun custom sidecars that tie together some nifty automations like automatically diarizing and transcribing voice notes of meetings from my phone without having to pay a subscription fee - latest projects have been a local sync point for logseq, kitchenowl, and nextcloud - super productivity android app and caldav and the to-do lists in home assistant. I never would have had the time to figure all this out with my full time job and parenting but I can cut the cord to all my subscription services and I look at things a new way now. I solve my own problems now when before I would just look for a way to buy a solution. I bought bunk beds for the kids but the bottom bunk didn’t come with slats and I couldn’t find the right size so I went to home Depot and built my own with instructions from my ops agent. I had a fun idea to make an e-ink display sign for my office door that shows what town I’m currently in and what my current MS outlook calendar block is and when it’s over. I never played with Arduino before and now I’m hooked. It gets me over that first execution hump. Sure I break stuff and have to figure out how to put it back together again but that’s part of the fun. I finally found a way to learn stuff that works for me, you know, comes naturally. It’s like I’ve been waiting my whole life for this ever since growing up seeing people using computers via natural language in Star Trek. It extends to my work. The data I need is gated behind ticketing systems of business intelligence professionals who don’t have the time or inclination to give me what I need so hey looks like copilot basic is great and m-code, guess I don’t need BI analysts after all. Sucks for them.
Hey man, welcome! Enjoy the freedom.
Got a buddy that has essentially the same story as you… he’s now running Linux as a primary and jellyfin + like 10 other containers in docker.
This time last year he was a Windows guy because he was scared of Linux.
Fast forward to ai being a thing and he’s all about it.
Ai has absolutely made it easier for folks to transition into Linux.
Hell I’ve been on it for like 20 years at this point and I still run into stuff I’ve not seen before. So I use it when ever I need to do so.
Handy.
AI is good at RAG tasks. When you’re asking it questions about text that is in the context window, the hallucination rate is significantly lower.
It can fill that gap between documentation and implementation that isn’t always obvious to someone who is new and answer question that they may not be able to form into a search string. The kind of simple questions that, in some Linux communities, would invite assholes (like ‘how do you set permissions’ or ‘I want to add my user to a group’) to tell them to RTFM. Now TFM can answer their questions directly.
Now I have a leftover Windows micro-PC running a local LLM model, so I can be less dependent on GPT, as long as time isn’t a factor…
Might as well go all in. LM Studio, Stability Matrix, Comfy UI, Ollama - everything’s native on Linux.
(i personally haven’t ever used llms personally but) i’ve heard many negative things about ollama from the localllama comm, may be worth searching before using it
Yes, well, my use for Ollama is solely as a plugin for Home Assistant. But you can wrap llama.cpp with whatever you want or use the half-dozen semi-popular competitors! My point is just that there are no tools that are exclusive to Windows, they’re all on Linux.
I’ve had zero issues with ollama, and every day I get closer to no longer using the frontier models.
found this from my bookmarks
That article is dated, and a lot of the new improvements specifically offset a lot of the claims being made there which just looking at the tone seem to be more of a hit piece on ollama then an objective look to compare. While llama.cpp does have some minor benefits, a lot of the testing I’ve seen and done shows that they are minimal.
For me at least olama is easy to use, simple to set up, and gets the job done. Other tools that’s fine, but as I said, I’ve been using it with no issue for a while now, and when I’ve looked at switching it just doesn’t make any sense.
Acronyms, initialisms, abbreviations, contractions, and other phrases which expand to something larger, that I’ve seen in this thread:
Fewer Letters More Letters DNS Domain Name Service/System IoT Internet of Things for device controllers LXC Linux Containers NAS Network-Attached Storage PiHole Network-wide ad-blocker (DNS sinkhole) VPN Virtual Private Network
[Thread #133 for this comm, first seen 17th Oct 2026, 00:00] [FAQ] [Full list] [Contact] [Source code]
Yeah. I’m not pro-AI by any stretch but internet searching powered by AI has made it MUCH easier to tinker and troubleshoot. The Linux community can be pretty awful when you’re new to Linux and trying to make things run.
Just RTFM!!!
*reads the manual and there’s nothing about this*
…ok. Now what?
Now search for a stack overflow comment, with one answer that’s not relevant and doesn’t work anymore and 30 “this is a duplicate it’s already been answered in a post 10 years ago”, and 1 niche forum with someone who had the exact issue, followed with “Thanks I figured it out”.
“Thanks I figured it out”.
That’s probably the worst part…
Sorry, that guide was written for systems running systemd and the install manual you followed was a from an opinionated Arch user for SysVinit.
Oh, this guide is 11 years old written by someone who hated Wayland, so no, modern KDE won’t work with it.
Just use Arch Wiki, it’s the best for beginners, but also don’t install Arch unless you know what you’re doing already because it’s the worst for beginners.
Yeah, any time someone tells me to just read the arch wiki for something unrelated to arch I automatically know they will never provide anything valuable to humanity
I have made the completely opposite experience. I tried so many distros and every single time some non-trivial issues came up my search ended in the arch wiki finally explaining stuff and writing out available options in config files instead of just telling my what option in what GUI config tool (that might or might not exist in my distro) I should click, then pray it works (felt very alike to windows actually where you get two dozen guides to solve a problem, each one editing another obscure registry key and trying to find the one actually working with the current version).
So I wasted years listening to the “No, Archlinux is a meme and totally unusable for inexperienced people” before installing archlinux, which is now running for more than 9 years without problems, and will happily point people asking questions for other distros to the relevant parts of the arch wiki that actually spell stuff out properly.
Archwiki is pretty good, it has a lot of breadth, but it’s not a manual and it’s not easy.
I suspect that LLMs do pretty well with troubleshooting Linux thanks to those manuals and old forum posts. After using them to troubleshoot Linux based devices, I tried the same thing on Windows and it struggled to figure out what commands to run.
Windows is a mess to configure and troubleshoot, even with the help of LLMs. Microsoft could have spent their investment trying to fix that, but I don’t think they have.
“Copilot can troubleshoot your computer” could have been a selling point, but they’d have to admit that windows 11 is a buggy mess
Yeah that’s the proper way to do it.
There’s a guy at work constantly asking AI everything on his second screen. He came to me after failing to figure out an issue using it. I said “oh yeah let’s look at the documentation, I bet there’s an example for this exact scenario” and sure enough there was. Made the change, tested and working.
On the other hand, any time I want to do something with Ansible, I start off reading the docs, can’t find what I need, scour a bunch of search results, and when that fails somehow AI can produce a working example of what I need in a few tries.
Agreed. The only advice people gave me with issues on Windows as “just use Linux!” OK, I switched…here’s my hundreds of issues, suddenly no help from them. ChatGPT made it possible for me to build a server and switch to Fedora. Not how I wanted to switch, but it worked
Using AI is generally a bad idea
It will bite you if your not careful
Great call out on the irony. So true. Welcome to the future.
Pretty much the same experience here. During my first week of full-time Linux I managed to break KDE Plasma with a widget IIRC and AI was able to tell me which file in .config I needed to delete to restore to default.
Welcome to the other side. I highly recommend checking man pages when looking up commands to start getting a feel for how to use them, also I recommend getting tealdeer (TLDR, get it?) It’s a TLDR of man pages that shows you how a package is usually used. I’ve often found that it bridges a syntax gap that is often missed in man pages.
Learning how to navigate and use man pages saved my ass when I was screwing around with my network.
thank you! I’ve been on linux for years but somehow hadn’t heard of these
Man, this makes me very proud of myself. I switched to Linux 10 years ago when it was really rough to maintain. Oooofff the sleepless nights I’d spent hunting for fixes that AI now can fix in mere seconds. I use ChatGPT a ton now to fix issues and my distro has been solid ever since.
I was never really against AI, and used the free things for most of the time. I moved to paying for AI once I started homelabbing, I have saved so much money with being able to move my things to my LAN. If I could afford a computer with a strong enough GPU it would be great.
I was able to setup a firewall running OpenWRT that I had tried searching the internet for help but it was always way over my head.
So, word of caution. Be careful. I tried using Copilot to open a connection for a single port between two LANs on my network in pfsense, as I didn’t know pfsense. It happily recommended opening 0.0.0.0/0 on the first LAN’s gateway address. It worked!.. By opening everything on the whole network up to anything at all. The problem with AI: you can only use it if you already know enough to catch its horrible screw ups.
The problem with AI is that it doesn’t push back. It is fantastic at manipulating your emotions.
I noticed that there are sometimes huge differences between models and providers. ChatGPT, which most people use, is a useless sycophant because it’s made to value user satisfaction higher than factual correctness (so it makes shit up und basically never pushes back if you challenge it). When using Mistral AI I had the AI tell me outright that I’m wrong (though it still generated very friendly text to do so, and it only properly does it when in “thinking” mode). No clue how Anthropic or Google do in this regard, just that ChatGPT is especially awful. In the end all models are risky.
Yeah I do know a little bit of networking and I would have definitely looked up opening 0.0.0.0/0. Not saying you did a bad or anything dergatory but I do double check things I am uncomfortable with.
Yeah, I know enough too, but I worry about the sneaky details I don’t know.
Good for you. The only two times I decided to glance at the AI overview of a search cause I wasn’t finding anything, it wasn’t useful at all. So yeah, I’m gonna save the planet a little more.
I still don’t understand where people are findign these useful AI answers. Every single time I try it seems AI is an acronym for artificial idiocy and I wouldn’t even trust it to tell me the time.
(In fact I just resently tried asking that stupid meta AI in sand-boxed Whatsapp (I still have to keep because of some stubborn examples of humanity) about the time in another time zone… It could tell my local time, in which time zone I was, what timezone I was asking about and what the difference between those zones was. Yet it could somehow not answer the question directly before being led there step by step. And then it still did the math wrong… 🤬)
Was about to say now you should look into self-hosting AI, but you’re already onto it. 😄
