- cross-posted to:
- linux@lemmy.world
- techtakes@awful.systems
- cross-posted to:
- linux@lemmy.world
- techtakes@awful.systems
Direct link to the LLM Policy Discussion: https://invent.kde.org/plasma/plasma-workspace/-/work_items/187
Example of KDE dev rejecting environmental consideration:
Example of KDE devs not wanting to require LLM use be disclosed.
And all of this is coming right after the debacle with an AI slop project being allowed to give a presentation at Akademy, and the person asking how this was allowed being blocked by KDE:
https://lemmy.nz/post/42029517
as well as the discussion thread bringing up legitimate concerns about AI usage being closed:
https://discuss.kde.org/t/sorry-to-bring-up-a-contentious-topic-kde-ai-llm-policy/46333/164
As someone who genuinely loves KDE, and has even donated to the project multiple times, I sadly have to accept that the majority of KDE have hopped onto the Pro-AI train, and they will not be dissuaded otherwise.
You must log in or register to comment.
I am locking the thread because I consider this cherry picking comments from KDE maintainers misinformation. Let’s wait and see where they go… The discussion as to include gen AI code, in which ways, how much, etc. is something that many projects (if not all) are undergoing. Let’s be respectful to their open decision processes! This respectfulness needs to be regardless of whether we agree or not with what they do. Once they make a decision, sure, go ahead with flaming commentaries and endorsing alternatives or forks (if they appeared)
If slop coding is so great why is everyone so adamant about hiding it from users?
If you’re not willing to disclose something, maybe take that as a hint.
As a user, I think this is the biggest issue I have with this policy. Disclosure is a big part of open source and hiding whatever tools are used is fundamentally against that philosophy.
Would you be willing to expand on what you mean by disclosure being a big part of open source?
I ask because it has not been my experience that OSS developers have ever had a habit of disclosing what tools they use. If you are lucky then a repo has instructions for contributors, that may describe tools being used, or CI configuration for running tests and linters, but most projects have neither. Mostly you just get instructions for building the project, but sometimes not even that
In some projects you’ll see a kind of a colophon section listing all the tools used in making some software. Not just obviously things like libraries, but OS, text editor, IDE, art tools, or even inspirations.
It’s not everywhere, but it’s frequent enough, and different from how proprietary software credits (or doesn’t) to make it a distinction imo.
I feel like using ai is like using drugs, they’re bad, but regulating them is muuuuuuuch better than just prohibiting, people will use it anyways.
It’s not a one to one comparison but I feel encouraging witch hunts will definitely do more harm than normalizing it.
To be clear, I don’t want it either, I’m acknowledging it’s the lesser of two evils.
I’m 12 and this is deep level of take right here
I feel like using ai is like using drugs, they’re bad, but regulating them is muuuuuuuch better than just prohibiting, people will use it anyways.
crime won’t happen because it is illegal?
Having an public policy that permits the use of LLMs feels like the opposite of “hiding it”. If a project has such a policy, then you can be pretty sure that LLMs are being used, even if you cannot tell exactly what commits involved the use of LLMs. It’s probably also a lot more discoverable than disclosures in commits or PRs, which most users wouldn’t know how to look for
How would you even enforce it. Its possible to make extensive use of local LLMs and have them write no code at all that you submit. Is that use of AI? Should it be disclosed? If you did you would get puritans calling it slop regardless of the result too.
Realistically, I think everything actively maintained is going to have some LLM code in it by now. You can either accept it or go mad, tbh.
Art is one thing, and I can see why we wouldn’t want the computer doing that if you have the resources to employ people, but code is rarely art. Most of it was already slop. It’s had things shoehorned into it by shouty managers decades ago. Claude is far from perfect, but so are most of the people doing the gruntwork.
Most of it was already slop. It’s had things shoehorned into it by shouty managers decades ago.
Manager: You have to get this done by end of week…
Dev: itll be really shitty and kind of work. Can we clean it up after?
Manager: of course!
Dev: its done, we can ship it. Ill start cleaning it up.
Manager: No, you need to do this other thing by tomorrow!
Ive had some good ones that did honor the agreement when pushed to rush things, but its not common.
“accept it or go mad” always just comes across as extremely desperate, lol
nooooo you don’t understand you HAVE to accept it! There’s no other choice!! Nooo stop rejecting it you’re not allowed to do that!!!
You can reject it in your own projects. That’s about it. If you’re open source and accepting any submissions, you’ve no idea if they’re LLM generated or not.
Except for all the open source libraries you probably use that also use LLMs, or you find a blog post someone wrote explaining how to do something and they used LLMs
You can avoid direct knowledgeable usage, but you’ll still have LLM code in your codebase.
You have no idea if they copy pasted proprietary code either, but you try your best to keep it out.
I can accept some trace amount of rat shit is going to get into my granola, but I am not willing to say that means I’m allowing rat shit in my cereal.
Please at least try to keep it out, as I do when I’m making my own granola. Yes I know it’s hard, but let’s just try.
because yall gonna go on a witch hunt at a whiff of llm, regardless of the final code or product quality?
Somehow you consider even a little pushback crazy but won’t listen to people when they ask you not to jump into AI madness headfirst wearing blinders?
when they ask you not to jump into AI madness headfirst wearing blinders?
That’s not what you’re asking and they are not doing that. The policy is quite sane. Ye of little faith (in the developers) were perfectly fine using this free software these same people wrote beforehand, but now according to the general sentiment in the comments they have all suddenly turned into idiots who don’t know what they’re doing and we should all abandon KDE etc etc.
Because users are dumdums who chimp out at the smallest of changes and mindlessly rebel just to feel important?
I feel like you’re describing the people who are crazy about AI right now.
the policy in question:
The golden rule for LLM usage in KDE is Don’t be lazy:
-
Don’t try to use a tool to replace your own judgment, interpersonal communication, or learning process.
-
Don’t take unsustainable shortcuts.
-
Don’t avoid growing as a person.
The result will be poor-quality work that eventually becomes someone else’s problem.
Nobody in KDE should know if you use an LLM — not because you’re concealing it, but because your usage of it produces output indistinguishable from what you could produce yourself. Contributions with obvious and lazy LLM usage may be ignored or closed.
Using an LLM to generate contributions
KDE follows a “human in the loop” principle: you must be making decisions and adjustments beyond mere prompting. The output must express your unique humanity in some way.
- Don’t be a “meat proxy”. For example:
- Don’t submit throwaway LLM-created changes as a “first draft or proof-of-concept for maintainers to improve”.
- Don’t submit “vibe-coded” changes you don’t understand and couldn’t make yourself.
- Don’t disclose LLM usage as a way of trying to excuse the potential errors or poor quality of a contribution that you haven’t put much effort into.
- Don’t add “Assisted-by: [some LLM]” tags to your commits; it’s just free advertising for the LLM’s provider.
(emphasis original, except for the fourth bullet point)
“You are responsible for the code you submit, regardless of tools used.”
Literally the same stance Linus had on Linux. I’m okay with it as long as it’s enforced.
Yes, that’s fine. And it’s also easy to enforce - just treat any code as if it’s written by a human (except obviously machine generated code like hardware descriptions etc).
Being against AI and LLMs is all fine and dandy, but it kinda takes ridiculous shape now.
However, I would prefer if open source tools would only allow the usage of open weight models and self-hosted systems. But then again, that’s my idealism talking.
I hate vibe coding, detest it, i’m tired of it. it wastes time and resources. That being said I use AI for debugging and as a glorified rubber duck. I don’t let it write ANYTHING but I will allow it to review code and make suggestions. Generally, I’d say 8 times out of 10 the suggestions are garbage BUT it will lead me down the right path. it’s a rubber duck that can talk back. So in that regards it has improved my build times, it has sped me up without actually touching the code itself.
THAT’s what AI/LLMs SHOULD be used for. an assistant that doesn’t directly interact with the project. a consultant to your code if you will.
unfortunately many people can’t make that separation and they sit on both sides of the fence.
LLMs are great for rubberducking and diagnosis, forensics, and the like.
We all know the enforcement won’t match the usage, but I’m still ok with the decision.
This isn’t a bad policy, but outright ignoring ethical and environmental concerns doesn’t seem very wise either. That should still be discussed, even if this becomes the final version of the policy.
i like how people exaggerated this into “oh no, KDE is s|op now”. I totally get the hate for greedy and overhyping ai corpos, but if this policy is a sign of slop, you a) most probably never coded anything in your life b) unaware that before ai-slop there was natural, human-produced™ slop and dumb stackoverflow copypasta that barely even worked.
It’s that humans cannot be trusted to not take shortcuts and AI has proven time and time again to be a literal gambling machine that uses persuasive writing techniques to convince you of things you wouldn’t otherwise think.
KDE is also fighting transparency which is terrible.
It’s not going to be shocking when KDE starts performing worse. If more than one person works on a project and you’re just hoping people use things responsibly, you’re dreaming.
Humans optimize everything and AI pretends to be the greatest optimization tool of all time while it spits spaghetti code it’s authors understand less of over time while it tells them "great idea! I’ll get to work right away!
you couldn’t trust humans to not take shortcuts long before the advent of ai. That’s the point of my previous comment.
KDE’s stance is literally “as long as the code is of sufficient quality and you’re able to take the responsibility over it, we don’t care which tools you were using along the way”. Policing people over what they’re (not) allowed to use never yielded positive results.
If anything, such policy allows less slop to slip through, than in any project with anti-ai policy, simply because the maintainers would be aware of the possibility, and thus would have to pay closer attention to the PR quality.
Finally, it’s an open source project. Even if anything sloppy slips past the eyes of maintainers, rather sooner than later it would be noticed by the community and fixed.
[…] AI pretends to be the greatest optimization tool of all time […]
Which is obvious to you, me, and trust me when i say that it’s obvious to anybody who ever tried using ai for even the most minor coding tasks harder than webdev.
This seems fine
This seems fine to me, it’s all they can really do since banning AI entirely just isn’t possible; someone could use it and submit very high quality code and no one would know AI was involved.
And if its very high quality, does it matter?
There are more concerns beyond the quality of the code itself, such as the ones I listed here
Sounds pretty reasonable to me. Don’t go saying “well chatgpt said it”, no, you are responsible for it.
One small problem: People are now also using AI to generate summaries of their vibe coded garbage, to push them through checks.
Using an LLM to generate text
Generally, don’t. What you are willing to write is generally the upper limit of what others are willing to read. LLMs, on the other hand, often produce long, rambling, soulless, corporate-sounding nonsense that nobody wants to read; don’t do it. This includes:
- Don’t use an LLM to “organize your thoughts” and copy-paste the output somewhere. Organize them yourself.
- Don’t use an LLM to generate commit messages and merge request descriptions. Write them yourself.
- Don’t use an LLM to generate a reply to someone’s question or comment, then copy-paste that as if it were your response. Understand and reply yourself.
emphasis mine
deleted by creator
-
This is EXTREMELY disappointing. KDE is my favourite DE.
This. Dunno where I’m going to go once KDE starts collapsing under the weight of AI slop. XFCE maybe? Do they have an AI policy? What’s Cosmic looking like these days?
Xfce have embraced AI with their new Wayland compositor.
Cosmic accepts AI code, but according to them, they take a preemptively hostile view toward AI code submissions and will rewrite some parts by hand.
There’s some lists linked in the comments here of DE’s that reject all AI.
Maybe you should reconsider your outdated views instead. Why are you even using linux to begin with, considering the kernel also allows “ai slop”?
Outdated views, huh? Tell me you’re deluded without telling me.
Not sure how this affects what I said. Having an anti-ai policy simply doesn’t work, you’d know that if you worked on any open source projects at all. There’s a reason all sane projects, like linux and now kde, end up accepting AI code. You can keep retreating to the next DE/WM until there’s none left but the ultra niche whose entire point is it doesn’t use AI, but it also has no contributor.
That’s like telling a vegan to update their views and eat meat.
There are many legitimate reasons to want to avoid AI code and avoid encouraging corporate AI use. Not all of us have to adopt your views on AI.
The differerence is being a vegan affects you, and the world is pretty accomodating to it. Vegan options everywhere these days. Non-AI repos are getting ultrarare and will only get more so.
Using a project with AI code simply doesn’t affect you. KDE is open source, using it doesn’t contribute to anything, doesn’t encourage and AI corporation, doesn’t contribute to water/power usage during training, etc… If the maintainers check the PRs propely you won’t be affected whatsoever.
X2
Do you want to say Kde was your fav DE.
It starts with “oh, just clean up any code generated” and “you must understand what your code does” and then in a decade that person won’t remember how to code at all because they self-lobotomized themselves for the convenience.
I don’t think becoming a worse coder is necessarily a big problem as long as code review really happens, as long as the human in the picture retains their algorithmic thinking there is no lobotomy. Of course we all know how human laziness works, so thinking that people would actually keep reviewing code after they find no major issues N times in a row is a pipe dream.
And I don’t know about you guys, but I for one dislike reading someone else’s code. LLM generated blocks can be especially bad for testing my sanity.
Yeah when I’m reading code and it becomes obvious there was zero thought behind a bit of logic, I lose all will to care about understanding the code or even what it does.
time to stop donating then. Ai crap results in death by a thousand cuts.
Actual proposed policy:
Nobody in KDE should know if you use an LLM — not because you’re concealing it, but because your usage of it produces output indistinguishable from what you could produce yourself. Contributions with obvious and lazy LLM usage may be ignored or closed.
Generally, don’t. What you are willing to write is generally the upper limit of what others are willing to read. LLMs, on the other hand, often produce long, rambling, soulless, corporate-sounding nonsense that nobody wants to read; don’t do it. This includes:
Don’t use an LLM to “organize your thoughts” and copy-paste the output somewhere. Organize them yourself. Don’t use an LLM to generate commit messages and merge request descriptions. Write them yourself. Don’t use an LLM to generate a reply to someone’s question or comment, then copy-paste that as if it were your response. Understand and reply yourself.
The one acceptable usage is to write text in your native language and then machine-translate it into English, with no stylistic or tonal changes.
But of course reading is for the LLM, it’s far better to be outraged about ‘hopping on the Pro-AI train’.
That section only relates to descriptions of pull requests, not the AI code itself. They don’t want long rambly AI generated explanations of how the code works, but they are fine if the code itself is AI generated.
Yeah, I forgot to paste part about the code.
I switched to Niri a few months ago as an experiment but now I guess I will not go back to KDE anytime soon
You cannot tell from the code necessarily anyway so anti-AI policies are hard to enforce. In the end the developer (or AI wielding human) is responsible for the code quality and its license which is what matters. All this discussion feels just like the usual culture war dividing the community. I’m more concerned about the intentions of people adding fuel to the fire.
In the end the developer (or AI wielding human) is responsible for the code quality and its license which is what matters.
They literally may not have any control over the license, that is one of the many big problems people have. Please read the article linked by OP about the copyrightabilty of LLM generated code - https://fsfe.org/news/2026/news-20260825-01.html. It depends heavily on how they used the LLM, so if there is zero disclosure required the entire project could get into licensing hell down the road.
Yes obviously the code you commit must be yours and you need to make sure it is. That is already a rule, what’s the point of telling people how to work above that?
Due to the nature of LLMs being trained on existing code there is a chance that generated code is lifted verbatim from another piece of software whose license doesn’t permit the copy without attribution. Its a potential legal minefield that I think FOSS project should just steer clear of until the legal landscape surrounding generative AI cools off some.
deleted by creator
Yeah, the people getting upset at this are dumb as rocks.
It’s a actually completely reasonable proposal
It all comes down to one thing - understand the LLM generated code you want to submit, and use the LLM for nothing else than write code, like making other people having to read LLM text. This is a whole lot like the kernel’s LLM policy.
Niko is right, the environmental corcern is not something this policy should go against - what can they do despite ban LLM usage altoguether (and per consequence get LLM generated code submitted anyway), they can’t have a blacklist for models, or force contributors to run them on their local hardware.
what can they do despite ban LLM usage altoguether (and per consequence get LLM generated code submitted anyway), they can’t have a blacklist for models, or force contributors to run them on their local hardware.
But they can? Like OP pointed out, wine forbids contributions from those that have seen Windows source code. How in the world are they supposed to enforce that? Turns out the open source community is pretty good at self enforcing.
I’m with you, but the sad reality is that, unlike Wine devs, AI devs can’t be assumed to act in good faith.
Potential Wine devs who have seen Windows source code will self enforce the rules, because they care about their integrity and the integrity of the project.
AI devs already don’t have an integrity, and they feel smug and superior for actively trying to fool any sort of measure against AI. They also seem to believe AI is the future and isn’t going away, so why bother regulating it if it won’t matter soon.
We need to come up with stricter and more effective ways to regulate and ban AI use, and that’s gonna be the difference between an unusable codebase and a great one in a few years.
Sure, but that doesn’t mean that it can’t go into the contribution guidelines. If they don’t listen then ok, you remove commits that have claude watermarking. Without that guideline you can’t stop the PRs at all.
I don’t think the average person who wants to contribute to wine has seen windows source code in the first place.
It is much easier to attempt on enforcing something like this that to enforce a full ban on AI. The problem is that AI has a much higher exposure rate than the windows source code.
Nobody said anything about enforcing it. Just put it in the contribution guidelines!! Like for fucks sake, this is like saying you shouldn’t have a law saying you can’t murder people out at sea because it’s impossible to police. The threat of action is half of the reason people don’t do bad things. There’s ZERO REASON to not have it in the guidelines.
World sees AI doing awful things like nudifying children, making nearly every software we use worse, and stealing the world’s knowledge to sell back to us, and generally dislikes AI.
Moronic software devs with massive egos: "WE SHOULD USE AI SEE HOW EASY THIS MAKES MY LIFE NOW I DONT HAVE TO LEARN. "
Also them: “I swear I just use it to augment my work! I’d never take a shortcut and eventually keep taking shortcuts and outsource my thinking! Humans notoriously NEVER take shortcuts!”
Guys wtf are we even doing anymore.
AI is not a thing, it doesn’t act, it doesn’t think. Humans do. Humans nudify children, humans enshitify software, humans steal content
Exactly.
All of the Anti-AI people are actually anti-capitalists, they’re just not finished diagnosing the problem.
AI isn’t building data centers, capitalists are. They’re doing it in such a destructive and rapid manner because they’re trying to leverage their infinite amount of cash into a future monopoly. If it wasn’t AI it would have been some other product.
Fighting the product and ignoring the actual human beings who are making these destructive decisions is only adding to the smoke screen that shifts blame away from names like Huang, Altman, and Musk.
I think you can be anticapitalist and also anti-AI. I’m certainly both
I think you can be anticapitalist and also anti-AI. I’m certainly both
Yes, the venn is basically a circle.
All of the ‘bad things AI is doing’ are bad things that capitalists are doing.
Nobody is actually angry at AI for solving age old mathematical problems, folding proteins or locating cancers. It’s not the technology that people care about.
People are angry at rapacious capitalism and are misdirecting it at the technology and not the rapacious capitalists. Anti-AI is just Anti-Capitalism for new leftists.
Same here! I don’t care who is responsible for it, no matter where on the political spectrum the creators are, I’m still not supporting it.
it doesn’t act,
I don’t disagree with anything else you are saying or the sentiment, but with the rise of “AI Agents” it is becoming more common that the AI actually does something. It seems like there is a new story every few days about an agent deleting an entire database or buying plane tickets or sending an e-mail to the FBI. Unfortunately, we have people giving it more and more permissions and ability to “just act” and it is only getting worse.
Most of those ‘news’ stories are sensationalized to the point of fiction.
You sound like “Guns don’t kill people! Robbers do!”
So I’m going to ask you, would you give a gun to a child? Would you? Just answer that
Except AI won’t kill you
What in the bullshit bad faith non sequitor is this.
And it wrecks the environment, and it stunts the user’s intelligence without actually being able to do anything truly original or novel, and it makes worse, buggier, less secure, less maintainable software. The list goes on and on
The stupidity of bandwagoners is on full display in this comment section.
So many people say they reject AI, but then how can you even use desktop Linux?
The kernel & the popular browsers have an increasing number of AI contributions, there is no escaping this.
As it is impossible to completely avoid AI use with Linux, I must content myself with only minimization instead of complete elimination, as I won’t let the pursuit of perfection mean I just give up entirely. Any reduction is still meaningful, just as it’s still worthwhile to minimize fossil fuel usage and meat consumption as much as is reasonably possible.
This is all completely out of your control. You cannot tell people how to work in opensource and projects are completely unable to control if someone uses AI (or which one) to perform tasks, since it is not visible in the result. The easiest thing for you to do, if you don’t want to take part in it, is to put down your phone and go outside.
…No it isn’t? I can choose what software I use, there are projects making efforts to avoid AI as much as is possible, and most contributors tend to follow guidelines, so having no-AI rules, while not 100% effective at avoiding AI code, will still drastically reduce what gets in, which is all I’m aiming for, and which is absolutely achievable.
I’ve noticed that virtually all Pro-AI people are constantly pushing this idea that resistance is futile, and that we all just have to accept AI into all aspects of our lives, almost like a cult. I even saw some if that rhetoric in the KDE proposal this thread is about.
You can drink the AI koolaid as much as you want, but you’re not going to stop me from trying to limit it in my life wherever possible, and I’ll continue to advocate for others to abstain for a multitude of reasoned arguments as well.
Modern society is built on destruction and cruelty. Doesn’t mean I can advice somebody to give up living in society and sound smart.
To suggest that “compromising on ideals for practicality” is the same as “roll-over and accept it” is dishonest and only done by people who want to win the argument by attrition so they get what they want.
I’m personally trying to switch to NetBSD. But, it’s not going well. My laptop can’t right-click in X under NetBSD, e.g.
A lot of the ecosystems I used to enjoy and support, I simply can’t anymore. Firefox, Linux, GHC, Debian, KDE, etc., etc. It’s a distressing time for me.
Can relate, everything I used to think of as a safe haven, a return to 90’s style computing, is seemingly being corrupted, and the few remaining options we have require a rather big sacrifice in usability.
deleted by creator
Well, fuck.
Any viable alternatives, or should I switch to BSD? (Gnome obviously doesn’t count, it’s designed for tablets, and unusable on desktops).
If you want to get away from LLM assisted code your best bet is NetBSD, with an older version of an X11 DE , using an older version of Firefox with JavaScript turned off.
If you do this you are basicly fully avoiding LLM assisted code. Be sure to check every thing you install though. Or use only software versions from before 2023.
BSD has a libertarian energy to its culture, so I suspect they will be vibing a lot too.
AFAIK NetBSD is explicitly rejecting AI contributions.
That and OpenBSD’s Theo de Raadt is rejecting LLM code based on copyright arguments (not “respect author” but rather “we get sued” kind of point), here’s an example (click “next in thread” a bunch of times if you want).
Though back in 2001 he claimed the songs on the CDs were “as free as everything else” (no), but like meh.
Check out the options listed here: https://slrpnk.net/comment/24475952
GNOME is great if you use Dash to Panel.
Damn… I guess I’ll have to start trying out more DEs again to find something to switch to.
Niri is solid. If you’re into it.
If you don’t mind me asking what OS do you run now? The LLM policy for KDE is essentially the same as the Linux kernel.
I’d wager that almost every large open source project has had AI code submitted and merged. There’s really no way to detect it.
