The Linux Thread - The Autist's OS of Choice

  • 🏰 Deploying update to the Tatarus security page. If you encounter issues, post here.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Betonhaus said:
Edit: ironically it seems to be Debian, followed by Fedora
Click to expand...
Hannibalistique said:
Fedora has always been a shortbus camp (now with shortbus agents), with Red Hat engineers spending many a man-month cleaning up the (human) slop for enterprise consumption, and Debian is just dysfunctional (even though some of my best friends are DDs, they don't make good technical choices).
Click to expand...

Between Debian and Fedora, the Fedora Project is the one that has the better approach to AI... and only slightly. ROCM explicitly targets RHEL and Ubuntu. Fedora is directly upstream of RHEL, and the Fedora Project itself has made commitments to spearheading AI usage within the broader FOSS ecosystem. Naturally to be expected considering how Red Hat controls the Fedora Project and RHEL.

The Debian project is directly upstream to Ubuntu. Canonical Ltd is now a serious contender for server, cloud, and IoT deployments. Canonical, like Red Hat, is explicitly preparing for AI features in their products. Unlike Red Hat, Canonical doesn't control the whole stack despite how much influence and paid man-hours that Canonical provides to the Debian Project. Nevertheless, Debian leaders who voted in favour of AI saw the writing on the walls: either get with the programme, or risk Canonical extending another tentacle into the project so it can steer the direction.

Speaking plainly here: I would vastly prefer an Ubuntu environment because I abhor SELinux and Firewalld, but I built my current PC in December 2025, and Linux Mint didn't have HWE-enabled kernels that worked on my hardware until April 2026. I'm already settled and comfortable on Fedora, so I'm basically here for keeps. Anecdotally with a 9070 XT and sticking with models up to and including the 14B range, Ollama works pretty damn sweet. Curl script from ollama.com and away you go.

It's also nice using a distro that doesn't bemoan the fucking dangers of AI usage like we're all infantile mongoloids only barely capable of human speech... speaking of which... I'm gonna go off on one.


Nuclear take: Any Linux user who has a hateful boner for AI, but then uses GE-Proton in any capacity is a hypocrite and a retard.

While I'm more than willing to dust off old Uncle Ted's manifesto and read some select quotes, the fact of the matter is that AI is a time efficiency tool first and foremost. The best way to use AI is squarely within the context of having the AI agent do dozens if not hundreds of repetitive actions that, even when using keyboard shortcuts and tiled windows and blah blah blah, takes a disproportionate amount of time for a human to do.

Guess what? GloriousEggRoll admitted as much when he dropped GE-Proton11-1 after a four month hiatus between GE-Proton10-34 and GE-Proton11-1, and dropped the bombshell that he used an AI agent to help him refactor the entire Wine media stack to use FFmpeg instead of Gstreamer. He didn't even prompt it with "Chatjippity saar, pls redesign entire Wine media stack" like some mongoloid jeet would. NO. Release announcement says he basically had the AI agent compare/contrast snippets of logic as he was rewriting the media stack.

The end result is that GloriousEggRoll was able to use the undisclosed AI agent in such a manner where he could rapidly iterate, test, and reiterate. That's fucking important because official Valve builds of Proton fucking suck with titles that make excessive use of MediaFramework (re: most JRPGs with anime cutscenes Catherine Classic, Hollow Knight's cinematics, etc), and Proton-CachyOS doesn't deviate too much from official Valve builds.

All those weebshit JRPGs on Steam with tons of anime FMVs (re: anything released by Atlus or Square Enix) now work best under GE-Proton, and arguably better than ever. No more obscure GitHub gist fixes that used to kinda sorta work back in the Proton 7 days, no more relying on GE-Proton8-32 for older games and explicitly declaring codec paths because it didn't have the tumultuous per-game fixes that GE-Proton9 and GE-Proton10 both had, you just start the game with GE-Proton11-x and away you go.

Here's the rub: Codeweavers has a strict "no AI-generated contributions" policy. AI didn't generate the whole Gstreamer -> FFmpeg rewrite, but GloriousEggRoll did point out that the AI agent was monumental insofar as the speed at which he could work on the rewrite. He's literally sitting on a substantial low-level media pipeline rework, all the hard stuff was already done, and he can't submit his changes upstream because Codeweavers is more than likely to reject it on AI usage grounds alone.

What the fuck is GloriousEggRoll supposed to do now? Painstakingly rewrite the entire rebase he did by hand, and without using an AI agent to prove to Codeweavers he was a good contributor who didn't use the evil, awful, no good, very bad AI agent? Wait for someone at Valve to kick it upstream but only after tweaking it so that Codeweavers deems it acceptable? I'm telling you: if wine-staging got the entire GE-Proton media pipeline rework merged into master overnight, the entire ecosystem would benefit. But NOOOOOOO, we gotta clamp down on AI tool usage because big corpo slopcode directives and the environment and data centres and melon husks.
 
Reminds me of how Jensen Huang said that the only downside to not using AI is that you'll be left behind, and as much as 99% of what he says is bullshit meant to appease the investors, he's right. Sounds like GE used AI agents exactly the way they should be used: as an automated helper where he still had all the control over the code and he understood the changes, and it allowed him to do this major rewrite solo in four months. Or look at the recomp projects, AI is immensely helpful in figuring out how a compiled game works so that it can be rewritten. Not to mention how in a year or two we'll likely see open source local models that'll be as good as Fable, completely democratizing agent assisted coding. Anyone will be able to run their own LLM on a rig that'll be about as bad for the environment as your average gaming rig and accelerate their coding work. Obviously people who'll overrely on it will create broken slop that will be impossible to maintain, but that's Sturgeon's Law, the developers that'll correctly harness the power of these tool only stem to benefit from it, like GE in this case. LLM code only becomes an issue when it did everything without your supervision, if it works with you, then it's a different story. People who'll now try and ruin his life over achieving this feat through an agent should never call themselves anything but useful idiots of big corporations.

At this point Microsoft doesn't need to plan some Machiavellian plot to sabotage Linux, these idiots will do it for free themselves all while convincing each other that they're doing the right thing. Even though if you want to take a proper bite into Windows' market share, beyond figuring out fundamentals like HDR, VRR, hibernation and other things Linux is still 50-50 on, you'll want a rock solid compatibility layer for Win32 software, the true reason why people use Windows. What would help the most in figuring out how this convoluted behemoth works? How to find all the undocumented API's software may rely on and shit the bed if they're missing? How to reverse engineer those? Yeah, LLM's are amazing for that. So by insisting that they're the greatest evil that musn't be used, you're actively undermining your own goals and playing into Microsoft's hand, congratulations.

FOSS projects will do nothing but self-sabotage if they insist that everything has to be written by hand. All while OpenAI and Anthropic will use their models to rewrite GNU GPL software to be published under MIT.
 
777- said:
omarchy is just a grifter distro made for gibs and trannies will forever stay firmly planted in linux because they have no lives and can dedicate 12 hours a day to harassing normal people out of their spaces
Click to expand...
nigger, omarchy is to lunix trannies the same thing NSDAP was to jews, only because of a few not-so centrist blogposts. Also take your meds.
 
Slav Power said:
Even though if you want to take a proper bite into Windows' market share
Click to expand...
Why do you bring this up every time? You make good points, but this market share thing is tiring

Wine and proton are cool shit for me, the linux user. Not some groveling selling point that will finally sway the windows people over.
HDR, VRR, hibernation? Yes please, but not to entice windows people. To make it better for existing users.
Windows users already have win32 compatibility, VRR HDR and games, linux doesn't owe them shit, they can stay on windows and probably should.

That number doesn't need to go up
 
Dread First said:
he can't submit his changes upstream because Codeweavers is more than likely to reject it on AI usage grounds alone.
Click to expand...
Except that you're ignoring the fact that there are very good reasons that any free software project that ever intends to be able to enforce license conditions would not allow AI contributions.

Whether LLM-created products, whether code or images or whatever, are subject to copyright, is still up in the air.

And what code is actually being 'created' by LLMs? Oh, that's right, it's fundamentally doing spicy autocomplete based on existing code. What existing code? Oh, that's right, primarily code under free software licenses like the GPL- being laundered into less free licensing, often into completely propetiary projects which are just copied at one remove (the 'AI') from GPL code.

I use Claude Code daily for exactly this purpose, it works great for violating the GPL.
 
Last edited:
Dread First said:
Nuclear take: Any Linux user who has a hateful boner for AI, but then uses GE-Proton in any capacity is a hypocrite and a retard.
Click to expand...
Video games are almost without exception proprietary trash, most of which these days come with some sort of kernel level anticheat spyware, so putting them on any machine that hosts any kind of vulnerable or privacy-respecting information is stupid from the start. If I see AI contributions in any project I value beyond entertainment, especially something as integral as a kernel or distro, I simply cease to trust it. AI shills are some of the most annoying people on the planet, and the "YOU WILL BE REPLACED CHUD" FUD is literal tranny level buh buhh buhh your project will be overrun by n-n-n-natziis if you don't add a code of conduct NOW NOW NOW!!! It is literally the same thing just by people who have somehow managed to be even more insufferable than trannies. It really is a baffling time where transvestites have become the second (possibly third!!) most annoying group in tech, succeeding AI cultists and pajeets, which are usually one and the same, if not biologically then spiritually.

Even good software erodes when you let people be rampant with AI, just look at the kernel and the horror that is LTS backports, just constant, ceaseless churn. The smaller the project, the bigger the risk of reliance on AI if you even let a little bit of it in. By your logic, everyone should just hop on the bandwagon and generate away until everything is like the kernel: AI reviewing AI with only distant human """supervision""" until they no longer have the bandwidth to do even that and lean more and more on AI reviewing AI until you have de-facto proprietary software that cannot be maintained by real people.

Case in point is my ceaseless spamming about the goings on in Guix, because it was one of the few projects where you can go and not have to deal with people shoving their bot slop down your throat, only for a stupid faggot with a vested interest in pushing his own PR bot as a claim to fame into the Guix project to burn down months of work and absolute unanimity in prohibiting AI authored or assisted commits. Now, there are shills that work in literal data centers and AI companies with their own vested interests shitting up the mailing list. This is the thing that irks me the most, the spiritually Indian AI cultists that go around forcing AI down your throat. It is literally no different than Drew "Lolicon/Paedophile" DeVault going around tryng to enforce CoCs. If you go around shilling AI, you are in the same social category as Drew DeVault, who ironically hates AI. Good job, I guess.

Good for EggRoll or whatever, I'm sure thousands of people the world over will be very happy with his work, I simply cannot jive with having this laundering machine being pressed upon me constantly. So, basically, I'm just not gonna use it (AI!!!!!!!!!!), I know.... UGH I know..... I'm sorry!!!!!!!!!!!!!!! It's just that I'm not gonna use it is all AHHAHAHAHAHAHHAHAHAHA
 
Intel® Core™2 Duo E8400 said:
Why do you bring this up every time? You make good points, but this market share thing is tiring

Wine and proton are cool shit for me, the linux user. Not some groveling selling point that will finally sway the windows people over.
HDR, VRR, hibernation? Yes please, but not to entice windows people. To make it better for existing users.
Windows users already have win32 compatibility, VRR HDR and games, linux doesn't owe them shit, they can stay on windows and probably should.

That number doesn't need to go up
Click to expand...
I don't care if other people use Linux or Windows or Mac. It makes zero difference to me. I want it improved for myself.
 
⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛ said:
And what code is actually being 'created' by LLMs? Oh, that's right, it's fundamentally doing spicy autocomplete based on existing code. What existing code? Oh, that's right, primarily code under free software licenses like the GPL- being laundered into less free licensing, often into completely propetiary projects which are just copied at one remove (the 'AI') from GPL code.
Click to expand...
By this logic, if anyone ever wrote a line of code that was inspired by someone else's code is now subject to intellectual property theft so even if you don't use AI you better self-police what code you read or else you'll get assraped in court (already the case for shit like people reading leaked proprietary code and subconsciously integrating it in a FOSS project, whether or not it's true, like ReactOS), and even a single line of code generated by an LLM simultaneously violates intellectual property of everyone whose code was used in the training data as there is no reliable way to tell which parts of the dataset were used to generate a given line of code, and there must be a victim of intellectual property theft, so now GE owes five quadrillion dollars for his copyright law violation.

Or, we acknowledge that intellectual property should be abolished and the next best thing that we can do is assume that anything generated through machine learning algorithms falls under public license, as to nullify the copyright law clusterfuck that has arised from people desperate to worsen the already draconic IP and copyright laws just because they suffer from AIDS. It's fine if FSF goes "no AI" as a means of avoiding dealing with the legal gray zone that is ML-generated content for now, but let's not pretend that you could apply copyright/IP laws to it in a way that would make sense and be consistent with existing law without making an entire exception in the law for ML-generated content, which would be a wet dream for legislators, especially those bribed by Anthropic/OpenAI to put a stranglehold on the market.
 
Slav Power said:
Or, we acknowledge that intellectual property should be abolished and the next best thing that we can do is assume that anything generated through machine learning algorithms falls under public license, as to nullify the copyright law clusterfuck that has arised from people desperate to worsen the already draconic IP and copyright laws just because they suffer from AIDS.
Click to expand...
'we should do anarchism rather than work in the world'.

A 'public license'? Anything generated by AI is inherently based on code that includes AGPL licensing, and it should be forced to be disclosed under AGPL conditions.
Slav Power said:
It's fine if FSF goes "no AI" as a means of avoiding dealing with the legal gray zone that is ML-generated content for now, but let's not pretend that you could apply copyright/IP laws to it in a way that would make sense and be consistent with existing law without making an entire exception in the law for ML-generated content
Click to expand...
Untrue. It's all just laundered code from freedom-maximizing licenses, and anything autocompleted by AI can be reasonably held to be the property of all those who have ever publically released code under the AGPL, with any of them entitled to sue to enforce licensing conditions.
 
⠠⠠⠅⠑⠋⠋⠁⠇⠎ ⠠⠠⠊⠎ ⠠⠠⠁ ⠠⠠⠋⠁⠛ said:
Except that you're ignoring the fact that there are very good reasons that any free software project that ever intends to be able to enforce license conditions would not allow AI contributions.

Whether LLM-created products, whether code or images or whatever, are subject to copyright, is still up in the air.

And what code is actually being 'created' by LLMs? Oh, that's right, it's fundamentally doing spicy autocomplete based on existing code. What existing code? Oh, that's right, primarily code under free software licenses like the GPL- being laundered into less free licensing, often into completely propetiary projects which are just copied at one remove (the 'AI') from GPL code.

I use Claude Code daily for exactly this purpose, it works great for violating the GPL.
Click to expand...

Except GloriousEggRoll himself confirmed that the AI agent's contributions specifically amounted to comparing logic both before and after he was running a test, seeing what worked well, what didn't, and then going from there. The AI agent, to the best of my knowledge and according to GloriousEggRoll's own statements on the GitHub release page, did not actually write the substantial body of GE-Proton11-1's media framework refactoring. That was still GloriousEggRoll's own work. The real time save with LLM usage, in this specific case, was speeding up the sheer amount of manual testing that needed to be done across dozens of games.

You're telling me that type of testing, invisible to most users who download the release tarballs, is subject to the AGPL even when the public-facing source repository only shows the changes between each version, and not the sheer volume of shit going on behind the scenes to go from release hash asdfg123 to zxcvb456?

Ferryman said:
Video games are almost without exception proprietary trash, most of which these days come with some sort of kernel level anticheat spyware, so putting them on any machine that hosts any kind of vulnerable or privacy-respecting information is stupid from the start.
Click to expand...

1) Kernel level anticheat is a deplorable practice on Windows. Wine/Proton, by their very nature, are entirely isolated within userspace. Easy AntiCheat still exists within your WINEPREFIX, but it cannot touch the actual Linux kernel on your system. You're more than able to play Space Marine 2 and Elden Ring on Linux without ever fearing EAC having ring-0 access to your shit.

2) The day that we have a proper FOSS equivalent to Dark Souls or Elden Ring complete with asynchronous multiplayer, jank physics, ragdolling enemy corpses, and "Try tongue but hole" messages scattered everywhere will be the day I'll actually start caring about video games being proprietary rubbish. Until then, I refuse to content myself with SuperTuxCart.
 
Slav Power said:
Even though if you want to take a proper bite into Windows' market share, beyond figuring out fundamentals like HDR, VRR, hibernation and other things Linux is still 50-50 on, you'll want a rock solid compatibility layer for Win32 software, the true reason why people use Windows.
Click to expand...
Do you think we give a shit whether Linux is popular or not? Why do you keep on constantly bringing this up like anyone gives a fuck?

I use it because I prefer it, and Windows 11 is shit. I don't care how many other people use it. I can do what I need to do on Linux currently, fine.

Midwits like yourself constantly pretend that every Linux user wants the normies to use it or something. I don't give a toss what you or anyone else use.
 
KDEs worst spokesperson, niccoloveslinux, has made a video about omarchy.

The TLDR is:
Omarchy bad cause muh AI
Omarchy bad because muh DHH genocide n shieet
Omarchy bad because I don't have millions of dollars
Omarchy bad because I alongside everyone else am a insufferable faggot who can't talk to people for funding
Omarchy bad because we are too retarded to take money from companies we don't like, even if there's no downside
Muh technofacism
If Omarchy actually does well we all look like retards

What makes this video worse is that for some reason he tries to act like hes neutral, trying to claim hes happy that projects that Omarchy uses are getting funded, just to turn around and complain that his project isn't getting funded the same because they are insufferable faggots no one wants to donate to. I am so happy Omarchy has given me a pathway out of having to even think about using software KDE and Gnome has made. At this point if someone creates a actually decent frontend for ComfyUI and a proper video editor gets made for linux i can finally scrap all kde applications from my desktop, and it cant come sooner.


Edit: Fuck it. Vibecoding a new paint/comfyui frontend.
 
Last edited:
Rust Troony said:
KDEs worst spokesperson, niccoloveslinux, has made a video about omarchy.

The TLDR is:
Omarchy bad cause muh AI
Omarchy bad because muh DHH genocide n shieet
Omarchy bad because I don't have millions of dollars
Omarchy bad because i'm alongside everyone else am a insufferable faggot who can't talk to people for funding
Omarchy bad because we are too retarded to take money from companies we don't like, even if there's no downside
Muh technofacism
If Omarchy actually does well we all look like retards

What makes this video worse is that for some reason he tries to act like hes neutral, trying to claim hes happy that projects that Omarchy uses are getting funded, just to turn around and complain that his project isn't getting funded the same because they are insufferable faggots no one wants to donate to. I am so happy Omarchy has given me a pathway out of having to even think about using software KDE and Gnome has made. At this point if someone creates a actually decent frontend for ComfyUI and a proper video editor gets made for linux i can finally scrap all kde applications from my desktop, and it cant come sooner.


Edit: Fuck it. Vibecoding a new paint/comfyui frontend.
Click to expand...
It all comes down to, Omarchy is bad because he said stuff that they don't like.

Basically anything DHH is forever condemned because he supported nationalists in the UK. That is basically their starting point.
I won't pretend I am some sort of saint, but they cannot deal with the fact that other people might disagree with them. They can't even engage with why people may think a certain way. I don't agree with some of the stuff he said; however, the people who "deboonked" him fundamentally didn't understand what he was saying either.

As for the AI stuff. I am working on my own Linux project using AI. I would prefer to write it from scratch, handmade. But I don't have time. I want to get a prototype ready.
 
Rust Troony said:
KDEs worst spokesperson, niccoloveslinux, has made a video about omarchy.

The TLDR is:
Omarchy bad cause muh AI
Omarchy bad because muh DHH genocide n shieet
Omarchy bad because I don't have millions of dollars
Omarchy bad because i'm alongside everyone else am a insufferable faggot who can't talk to people for funding
Omarchy bad because we are too retarded to take money from companies we don't like, even if there's no downside
Muh technofacism
If Omarchy actually does well we all look like retards

What makes this video worse is that for some reason he tries to act like hes neutral, trying to claim hes happy that projects that Omarchy uses are getting funded, just to turn around and complain that his project isn't getting funded the same because they are insufferable faggots no one wants to donate to. I am so happy Omarchy has given me a pathway out of having to even think about using software KDE and Gnome has made. At this point if someone creates a actually decent frontend for ComfyUI and a proper video editor gets made for linux i can finally scrap all kde applications from my desktop, and it cant come sooner.


Edit: Fuck it. Vibecoding a new paint/comfyui frontend.
Click to expand...
Its clear as day that these control freaks are losing any power that they had and are refusing go quietly into the night, It drives them up the wall, that any one who's not them can be successful.
 
You must log in or register to reply here.
Back
Top Bottom