They REALLY don't like AI, to the point of "wish I didn't need to rely on GCC and GRUB because they have AI in them" levels of hate AI. See this further down the linked page:
>The fork in question, of U-Boot, is heavily AI vibe-coded and thus useless to Libreboot. This is not mere conjecture; in said aforementioned U-Boot mailing list thread, the author stated thus. We will not, in any way, encourage or facilitate use of AI. It has been made clear that several projects Libreboot uses also have some AI-assisted commits; GCC, GRUB, etc. This is not our fault, but it is my intention that I and others will audit these projects in the future, de-slopping them and possibly even forking them entirely for that very reason.
I have learned to treat statements like that as a general red flag.
Someone who acts like a radical on practical issues will almost certainly act like a radical in other areas as well - including areas that matter to me.
There are debates to be had on every technology, but foaming-at-the-mouth kind of talk is a sure sign of a maintainer (and by extension a project) I want nothing to do with.
I respect projects with a no AI policy, but if that policy starts with the words “fuck no”, I’m out of there.
It's possible this type of "radical" behavior is why this project exists in the first place. I'm willing to bet if the Libreboot maintainer(s) were not "radical" they'd just be using a Mac and be content with a closed source bootloader.
Not to say this "radical" behavior is helpful, and I understand why people don't like it.
I don't really see it as a question of "separating art from the artist". That's usually invoked when other aspects of the personal life of the artist has the potential to distract from the art itself.
The parent comment is closer to saying that Picasso wouldn't have been Picasso if in an alternate universe he had avoided controversial/politically charged subjects like the Spanish Civil War and instead appealed to the lowest common denominator for a mainstream commercial audience to sell postcards.
AI in its current form is a commercial product created from stolen labor.
I think there are extremely legitimate reasons to be critical of that part at the very least, and I don't think it's a particularly "radical" stance within Free Software, which is fundamentally conscientious of such questions.
(I don't know whether the Libreboot authors have this particular point in mind.)
Has anyone who contributed to producing the wealth of human knowledge, art, code, and literature making up the training data been compensated in any way?
Well since the early 18th Century (in a limited way) copyright rights have been given to creators - the overwhelming majority of countries are signatories to international treaties (for the last 100 years) that give creators control of created artistic works (which has been deemed to include program code).
Exploitation of those rights is down to individuals, only since Google came on the scene has any company been able to get away with broad ranging infringement of these rights without censure.
As for your cheque, you have the right to sell your work. No-one is obliged to buy it, of course. You also, under copyright, have the right to make a license to exchange use of your works for binding the users to conditions of release of their derivative/associated works ([F]OSS licensing).
You still have the right to sell it, and if ML models have been trained on those works you have the right to seek compensation.
That already didn't happen, regardless of AI. The thing about freely sharing knowledge is that people take that knowledge and use it to build further knowledge that helps you.
You might say that the fact LLMs are being provided as services by big techs invalidate that. But some people taking the provided knowledge for themselves has also always happened. More importantly, though, is that big tech provided LLMs aren't the only way to access the technology. Open weights and local models exist.
Casting this as “knowledge” betrays a very fundamental lack of knowledge about how LLMs work, or a desire to mislead for whatever reason.
They are not databases of facts. Even if they were, that would still constitute a very serious violation. They aren’t, and they’re not particularly good at serving that function either.
No, LLMs are capable of reproducing things (text, art, other media) based on their training data. Consider being an artist who produced something original that somehow made it into that training set - your chances of actually making a living are now vastly smaller.
The result is the same as if there was no copyright at all: Big companies will just exploit whoever, no art would be produced.
This is a huge issue, and there needs to be a reckoning.
How do you think LLMs work? Because you seem to think they're a big compressed folder of every content they were trained upon, thus a big permanent copyright violation.
You're probably also capable of reproducting some art off the top of your head, your existence doesn't somehow violate copyright.
Besides, anything they're reproducing is the work of very successful artists whose work is widespread. Anyone who had their worked looked at once during training isn't being affected anymore than some artist who never had anything of theirs looked at.
Also, copyright benefits big companies much more than it benefits artists, by the way. Not being able to take any art for free and requiring to pay is a great trade-off for getting a monopoly over anything they pay for. Art would still be made regardless, just like it was before and just like it is right now even when, outside of these big corporations who mostly respect it, copyright is constantly infringed upon.
One thing is that most if not all models more likely than not were trained on leaked Microsoft and other proprietary code sets.
For projects like WINE or ReactOS that absolutely have to enforce hard firewalls between people with "insider knowledge" to avoid contamination and resulting legal issues, this alone makes the use of AI for coding highly problematic.
If you believe that, then why aren't LLMs free to use for every person on Earth?
If you want anybody to keep making anything, including programmers, artists, and writers, you're going to have to pay them somehow.
Anthropic and OpenAI would not have a product at all without the contributions of thousands of non-consenting humans. Yet they somehow have the gall to charge customers a premium.
Anthropic and OpenAI aren't the only ways to use LLMs. There are plenty of LLMs where you can download the weights and run them locally. There are even others that you can train yourself (see the LLM360 family).
They refuse to release the models, the training method and, training data. They don't even want you to ask it too many questions lest you teach another ai. Very "open".
Since that's your position -- have you ever produced proprietary software in exchange for money? How do you intend to make up for your unethical actions, if you have?
And, how do you envision the white collar economy, which is entirely built off of selling knowledge, will operate if you get your way?
> And, how do you envision the white collar economy, which is entirely built off of selling knowledge, will operate if you get your way?
It will crash and burn anyway. That kind of economy made sense before the Internet, before mobile phones made it possible to carry the entire world's knowledge in your pocket, only impeded by your ability to efficiently query the database and apply the concepts.
Now with AI? It's even worse. AI not just can efficiently query its own knowledge dataset it was trained upon, but it can also search other databases, apply the knowledge to the input and to put the cherry on the icing, it can do so when operated in natural language. And we're not going to be able to put the genie back into the bottle, not with people being used (and able) to run open weight models on local hardware. OpenAI et al don't have any actual moat behind them, they are just making stuff a bit more convenient and are for now the cheaper way to access compute resources.
For Western countries, the white collar economy is going down the drain, black collar (mining and other forms of dirty resource extraction) is already gone off to places that "compete" by using slave labor, and what remains of blue collar now that industry went the same way as black collar is heavily under threat because so many white collar people are fleeing towards the trades.
It's unclear from your comment whether you think this is good or bad. You certainly seem to think it is inevitable.
I just want to remind you that nothing here is inevitable - someone is making it happen, and it's only possible because we're not enforcing any of the laws we have about it for some reason, such as copyright. Well, the US isn't enforcing it. The EU is scrambling to come up with reasonable regulations. God knows what China is doing.
I agree that AI, like many other things, would have been an unambiguously great invention in a post-capitalist society, but that's not where we are. They are unbelievably powerful, impossible to control, and fully owned by powerful corporations, so the future looks pretty dystopic for humans who love freedom, art, and science.
But I hope you realize that we have the chance to fight back and get this right. But it does require that you do something.
> It's unclear from your comment whether you think this is good or bad. You certainly seem to think it is inevitable.
I think it is bad, but the consequences are manageable.
> someone is making it happen
We're past that point. Open weights plus local inference is already good enough for an awful lot of use cases. Even if we'd say we ban all AI training, we yeet China and the US from the Internet... it's too late.
> But I hope you realize that we have the chance to fight back and get this right.
The best chance we have is mitigate the consequences. That means first and foremost UBI, backed by actually useful corporate taxation. We need to prepare our economies for the shift towards an economy that requires fewer and fewer people to actually do work. For the moment, the only thing that is relatively safe from both automation and scale efficiencies is some trades (plumbers and electricians), some sections of healthcare and some sections of black collar work. Everything else is under massive threat from AI, automation (have you looked at precision farming?!) or both, and politicians seem like a deer frozen in the headlights.
This is the classic doomer take. Beware of dramatic but simple answers. UBI doesn’t really seem like a sustainable way forward, and at the same time there’s so much we don’t know about the limits.
Don’t play half-court tennis. Other people can read and write too, and social shifts are simultaneous slow, but then sudden.
For example, I’m not seeing a lot of trust in AI as a technology for things that are actually important. It takes one wrong diagnosis from an expensive AI doctor to erode all trust and make everyone prefer a human expert.
As I understand it, the argument is about defending intellectual property. Not that strange to hear from the Free Software community, where many people have strong opinions about stuff like licensing. The "Free Imformation" (pirate) community does not fully overlap with the Free Software community.
When your project is literally writing a BIOS/ UEFI alternative that exists for the purpose of making base runtime systems open and verifiable and to improve the user’s ability to trust in their system, an opposition to unverified machine generated code seems not only justifiable but sensible.
If Leah were not taking steps to mitigate generated code in the project, the project would no longer meet my needs as a Libreboot user.
I find it hard to square treating that like such a red flag while supporting free open source software more generally. The entire ecosystem is rife with ethics/morals driving the initiative. There is a lot of flag planting and red line drawing.
People say AI or not, if the thing is good then it should be used. But if a group actively rejects it (which is valid even if I’m not that militant myself) it compromises the project suddenly?
> Someone who acts like a radical on practical issues will almost certainly act like a radical in other areas as well - including areas that matter to me.
Not using LLMs is a practical choice. The tech doesn't work, and if you let it run amok in your software, your software becomes crap. I try to avoid projects that use LLMs for development as much as I can, because those projects have a much higher likelihood of being a buggy mess.
Personally I treat whinging about other people's ideals as a red flag.
It's libreboot. The entire point of its creation was to take coreboot and rip out the non-libre blobs. Its author is trans (IIRC), which also carries its own very pointed perspective on the world.
Some of it goes too far in my own opinion - I have no problem trusting AMD's final microcode for the Opteron 63xx any differently than the baked-in mask for the 62xx. But regardless, I respect someone who is pushing the whole envelope as they see it. At the very least their effort creates a lot of things I do care about!
Rereading this I realize I might be in the running for the RMS award for technically correct social obtuseness so let me just say that my intent was not to denigrate trans at all, just saying that being trans is itself intrinsically political (as in you're going to be continually clashing with many people's default assumptions, even before the manufactured "culture war" throws gas on the fire). And then furthermore being openly trans is going to impart those politics to the project. Which is most certainly her right, as it is her project! In fact I would say that trying to push a standard of you can contribute your code but stfu about your other ideals is basically cruel.
Well, yes, this is a project to create a fully libre, GPL'd boot chain free of binary blobs. The few users I know stick with aging ThinkPads. I'm not sure what you expected.
Likewise, it if claims something is “vibe coded” when it clearly is not (such as U-Boot).
Agentic AI has allowed 86Box to make massive strides forward. This includes a substantial contribution to emulate the Intel Inboard 386 that was created by someone who describes himself as a non-programmer. The community is now bigger and stronger, although we’ve had some aches and pains with slop AI generated PRs.
The same goes for Gravity Linux which is booting in my M4 Macs with working GPU and most other peripherals. What used to take years now takes weeks, so we can do much more thorough testing and ultimately have a better end product.
And what happens when the AI bubble pops and the various publishers and IP rights holders come around with their hands out claiming 86Box stole their code ala' SCO?
AI code is legally nebulous and relying on it for a serious project has major risks. Linux has already fended off SCO and has deep pockets to boot. Not all open source projects do.
More projects should be thinking less about the 'moral' issues (yes, they exist) and more about the nuts and bolts province and liability issues of using unverifiable code.
Yes, Linux uses it -because they're big and professional and have allies like IBM and Microsoft (ironies of ironies) on their side.
> And what happens when [..] the various publishers and IP rights holders come around with their hands out claiming 86Box stole their code ala' SCO?
It won't, for a very simple reason. The same reason why the world isn't in the middle of a nuclear winter right now. Mutually assured destruction.
Every single proprietary software vendor right now is making heavy use of AI code generation, and we're already at a point of no return where pretty much every single commercially relevant codebase out there is contaminated. Do you really think companies will go on a suing spree where by doing so they will open up themselves to being sued for exactly the same reason?
>It won't, for a very simple reason. The same reason why the world isn't in the middle of a nuclear winter right now. Mutually assured destruction.
An interesting take, which I honestly hadn't considered.
The problem remains that it is still a cudgel which larger companies can still use it to shut down smaller companies/free alternatives. I'm thinking a similar situation to the abuse of the DMCA -but we'll find out.
Yes, anti fascism is a radical position these days.
Already in 2022, before the rise of these LLMs, there was a book "Resisting AI: An Anti-fascist Approach to Artificial Intelligence" by Dan McQuillan.
And then:
> The rapid development of digital technologies, including applications of artificial intelligence (AI), is having a significant impact on the social, economic, and political life. From the way individuals communicate and form relationships to the methods by which political discourse and economic transactions occur, digital technologies increasingly mediate our lives and societies. But while presented as milestones in innovation and progress, these technological transformations have also raised concerns about democracy (Nemitz 2018; Kreps and Kriner 2023; Jungherr 2023; Coeckelbergh 2024) and seem to introduce and support mechanisms of control, forms of organization, and ideological patterns that bear striking resemblances to historical fascist phenomena.
> being plausible but slightly wrong and un-auditable-at scale-is the killer feature of LLMs, not a bug that will ever be meaningfully addressed, and this combination of properties makes it an essentially fascist technology. By “fascist” in this context, I mean that it is well suited to centralizing authority, eliminating checks on that authority and advancing an anti-science agenda.
I know this won't be popular with this crowd but most of you have become supporters of the rising fascism in the United States by not resisting these LLMs.
> Yes, anti fascism is a radical position these days.
Given that these groups are basically composed of communists, if not tankies, that's not surprising. Specially when some of those served as sabotours for the USSR (notorious for its respect for human dignity).
This is an incredible piece of anti-intellectualism. Its arguments aren't compelling either: AI might flood with innacurate studies that are difficult to check. Rather than improving said check as to reduce the reliance on trust and improve science as a whole; like the good reactionary the author is, they proppose to ban this particular technology. AI also centralizes authority... because? There is also an irony in them complaining about this theoretical centralization while also trying to preserve the inherent trust-based system in academia, so I suspect they aren't actually committed to proper intellectualism. There is more to complain about this post, but, then, it would hijack this whole comment.
> Marc Andreessen have called one of the two authors of the very Fascist Manifesto his patron saint. Search for Marinetti on [...]
Is it surprising that the guy who founded Futurism, a rather radical and aggressive celebration of innovation and progress, would be mentioned in a movement about technological progress?
> A very influential blogger, Curtis Yarvin who have started Dark Enlightenment, the philosophy behind it all regularly posts straight nazi shit. He also has influenced JD Vance and Peter Thiel who have been instrumental in the creation of OpenAI.
JD Vance hasn't been instrumental to OpenAI's creation, and Peter Thiel is just one of its original investors. Not to mention, NRx is a mess of movement composed of contradictory ideas.
I have noticed you have been using "these LLMs". While I do agree that distancing from OpenAI's models is good (even if it's just on the basis of decentralization), the anti-AI movement either wants to monopolize AI via regulation or ban it outright.
This is just as bad as blanket hating/rejecting AI without ever trying to use or understand it. I don’t care what tools a craftsman uses so long as the job is done correctly/well. If you can somehow prove that AI is necessary for this that’d be impressive.
> U-Boot upstream itself is staunchly anti-AI, at least at the time of this release, and much discussion of the above was had on their mailing list.
I saw this as well, which is funny because I've sent plenty of patches written using AI assistance to upstream U-Boot. It was my understanding that U-Boot was mostly following kernel conventions in general as many U-Boot drivers share code(in addition to the actual subsystem maintainers themselves) with their kernel equivalents.
Seems more like clean room RE to me, or general free software stuff. You can't contribute proprietary code to free software, if someone doesn't agree to a license change you have to remove/rewrite their contributions, if you've seen the Windows source code you aren't allowed to contribute to ReactOS (IIRC), etc. LLMs are black boxes with dubious training data and aren't really "you" writing the code directly, so whether you can then license that is questionable. This is especially important for a project like LibreBoot that is trying to liberate machines. The only "easy" solution I see is to abolish copyright, which would be fine. As it is a lot of software is being tainted and the whole world of free software is in danger. It's a shame people gloss over this because they think LLMs are cool or help productivity or whatever. Everything's been irradiated now. Now if someone wants to guarantee a given project had no AI contributions they have to fork a years-old version and likely lose a lot of good non-LLM code in the process. What a waste of time and resources that is. Setting us way back. This needed to be taken way more seriously on day 0.
That's not what it implies to me. And let's not pretend that a person with this approach and this manifesto is going to publish their patches to deslop slopped upstream projects.
contributing to projects that have slop in them has legal liability. you'd have to essentially hardfork to remove the slop from git history and such.
for my part, when i do work on this, what i will do is this: re-use existing nuke.list mechanism in lbmk (used already to delete dubious files, e.g. code files with unclear licensing in u-boot) to delete the slopped files. then provide cleaned versions of the same files, with necessary patching re-written as necessary, and files that use it patched to include e.g. foo.deslop.c instead of foo.c - then git history is removed inside tarballs. filter out the changelogs for projects accordingly.
this is the most reasonable way to do it on projects with very little slop. with projects that become almost entirely slop, a hard fork becomes necessary.
So what OS are you building on and running on the machine? https://codeberg.org/ethical-foss/open-slopware#kernels doesn't seem to list anything I would consider a good alternative to Linux, although illumos or 9front becoming the preferred build environment for libreboot would make my inner nerd happy.
currently debian linux, though i have porting lbmk to openbsd on TODO. with a few small exceptions (e.g. recent tmux patch) they're generally quite hostile to use of AI.
yes that is why i mention openbsd. i haven't ported lbmk to it yet, was going to months ago. i'll start up that effort again at some point. future releases will one day be built with openbsd, and when that day comes, my reasons to use linux will drop to near zero. i'm an avid fan and user of openbsd - i even maintain a port for it (librewolf web browser)
> think for yourself. AI will only ruin your mind and deskill you. here are some articles:
> think AI makes you super hacker? think again: https://arxiv.org/abs/2507.09089 ["Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity"]
> consider that the infrastructure for AI is economically unsustainable; when the bubble bursts, and it will, we will suddenly be left with millions of people thoroughly deskilled who will be useless. lots of reeducation needed, and lots of people will just be left behind.
> consider that it is controlled by only a minority of very wealthy individuals; with it, they have unprecedented control over human thought through manipulation of information, and control over what information is presented through filtering (what they call training).
> AI is a threat to the health and democracy of our entire world.
> libreboot will not facilitate its use. AI is the most dangerous technology ever invented by human beings.
This one especially really just shows they are cherry picking things which support a particular position regardless of the relevance, the difference between Early-2025 AI and Late-2026 AI is absolutely massive. Early 2026 it was reported by Greg Kroah-Hartman that AI bug reports rapidly went from obviously wrong or low quality to being real reports.[0] The kernel now has AI tooling heavily integrated into patch review workflows as well with quite good results.[1]
> the difference between Early-2025 AI and Late-2026 AI is absolutely massive
If you only accept papers that were published in reputable venues and analyze LLMs that are less than a month old, you are left with zero useful papers (since writing a useful paper takes longer than a month). It is impossible to have an intellectual discussion if all scientific material is disregarded.
> If you only accept papers that were published in reputable venues and analyze LLMs that are less than a month old, you are left with zero useful papers (since writing a useful paper takes longer than a month).
Why would one look only at formal published scientific papers? Especially if the only publications are known to be quite outdated and highly likely to be no longer particularly relevant anymore.
> It is impossible to have an intellectual discussion if all scientific material is disregarded.
One can certainly have an intellectual discussion even for areas without relevant formal published scientific papers.
Certainly one can do so if the subject matter is in the area of your own expertise, that can even lead to scientific papers being written.
I'm an experienced open source developer leveraging AI tools for linux kernel, u-boot and buildroot contributions in addition to reverse engineering, so clearly this subject is well within my area of expertise regardless of what some outdated published research says.
I probably wouldn't even disagree with that paper's assessment at the time, however the situation is very different now. I have noticed that there is an issue with many experienced open source developers simply having not figured out how to properly leverage AI for open source projects they work on. Probably one reason for this is that productive workflows with AI are often very different from previous workflows and it takes some practice to figure out how to best use AI tools, these best practice can also change over time.
Unfortunately it is. Thought the whole thing was a scam until six months ago. I think pretty highly of my development skills and have to be objective and admit I am outclassed in debugging and operational tasks now (and it isn't close). Writing non-script code isn't quite there yet (they are bad at abstraction) but I can't confidently predict how much longer that will be the case.
> when the bubble bursts, and it will, we will suddenly be left with millions of people thoroughly deskilled who will be useless. lots of reeducation needed, and lots of people will just be left behind.
what do you mean by this, that costs will become too high and people won't afford it? hardware won't go up in smoke "when the bubble bursts"
People won't be left without AI/LLMs. They'll be left with less value in their investments. The tech is here to stay, the market discovering the actual commercial value of these products won't change that.
How terrible. I'm sitting on multiple AI reverse engineered firmwares that I'm not sure what to do with. I suppose they'll never make it to upstream now. My only option is to ask AI to integrate this stuff.
This conflict has come up a lot in open source lately and the answer has been to just create your own repo and space where you can dump unlimited ai output without needing anyone’s permission. If users find it useful they can use it, and there’s no conflict with maintainers who don’t want it.
Time will tell if these new AI allowed projects like artcraft will prove more popular than fully human curated ones.
The whole point of open source is to get this stuff upstream so that a rich software commons is created.
I'm not sure I want to participate in this space anymore if everyone is just going to have their own isolated forks. I might as well keep my own private version of everything and never deal with maintainers ever again.
There has always been peoples personal forks and random one man hobby projects on Github. I think the Artcraft project is a good example of what's to come. Gimp would have never accepted thousands of commits to convert it in to a replica of Photoshop, and they shouldn't have to. The people who want to go full speed AI programming can run their own project and don't have to bother the Gimp maintainers.
Is it really out of the question to stop using AI and continue contributing upstream to free software projects? It was just a few years ago this was the norm.
>The fork in question, of U-Boot, is heavily AI vibe-coded and thus useless to Libreboot. This is not mere conjecture; in said aforementioned U-Boot mailing list thread, the author stated thus. We will not, in any way, encourage or facilitate use of AI. It has been made clear that several projects Libreboot uses also have some AI-assisted commits; GCC, GRUB, etc. This is not our fault, but it is my intention that I and others will audit these projects in the future, de-slopping them and possibly even forking them entirely for that very reason.