Linus Torvalds used AI to fix a Linux bug, and now LLM critics on Linux face an uncomfortable choice

"People are going to murder anyway, so why even make murder illegal?"

You are correct, you should make rules even understanding that they will be broken by some.

I think one way is to make people realize that there is a concept in copyright called Threshold of Originality, meaning that because machine generated output in itself is not creative work, it's not copyrightable. You can do literally anything with it, the licenses attached don't matter. The GPL is not enforceable and no proprietary EULA is enforceable.

I hate this attitude that if a rule or standard can’t be enforced perfectly, we shouldn’t strive for it.

"Murder bans may be impossible to enforce in the first place"

I think of it like age verification. Slipperly slope to a proprietary kernel-level "security compliance module" on all devices, aka the end of general computation.

This isn't a case where something can't be enforced perfectly, it is a case of something that can't be enforced at all going into the future. Every tool that can reliably detect LLM code is at the same time the tool used for adversarial training, making the generated code look more and more human. We are already at the point where for plain english the false positives and false negatives go through the roof, making these tools very unreliable and when applied automatically a liability. Code is a lot more formalized, with a lot less personal variance (spelling, vocabulary and grammar are basically fixed - only the used logic and how it is implemented is variable), making detection harder by default than in natural languages.

If LLM code can't be detected anymore by automated means - and that state of things is approaching fast - then any policy about allowing or restricting LLM code is not worth the paper you would use to print it out. But that's not so much of a problem. The more important policy to set, that can also be enforced, is that everyone submitting code has to take personal responsibility regarding the quality of the submission. Delivering bad code - when not happening while training to become a better coder and looking for feedback - has to lead to consequences based on the seriousness of the case and if it's a repeat offender. Anyone using an LLM to spit out bullshit LQ code will run into that kind of rule very fast.

I'm very critical of AI, but complete bans on AI use are, at the moment, pretty much unenforcable. Not hard to enforce, not impossible to enforce 100%, but completely impossible to enforce in general. There currently is no reliable way to verify if and how much someone used AI, except maybe if you monitor people's systems (and even with a totalitarian surveilence system, coders will probably be able to bypass it).

Obviously, if someone who has no clue about coding uses it to write the entire code, you notice. But that's simply not how most coders use AI..

Edit: And additionally, mandating disclosure of AI can make sense, but it can also create the false assumption that codeers/code without any disclosure are verified to not use AI when coding when it simply cannot be verified. It depends entirely on the coder being honest and transperant about their use of AI. And given the passionate pushback, coders are definitely incentiviced to not disclose AI use.

So how would one enforce this?

It is not that it is very difficult, it is impossible.

Just want to let you know from my canoebooted laptop that I run a 100% "free as in freedom" OS (trisquel GNU/Linux). Also you have convinced me about striving for a goal instead of doing it perfectly, which is not possible. Well said.

I hate this attitude that if a rule or standard can’t be enforced perfectly, we shouldn’t strive for it.

This is the entire basis for the War on (some) Drugs. Guess what? Drugs won.

Power is nothing without enforceability, and purposely implementing rules or laws that can't be enforceable is a form of malice and discrimination through selective enforcement.

AI code ban makes as much sense as code written in IDE ban. That's why it is impossible to enforce. It makes no sense.

Hackers use LLMs to spot odd exploits so it makes sense developers use them too. The usage of LLMs in development likely won‘t make Linux better or safer overall, because bugs and exploits wouldn‘t have been found by either side without it in the first place. It‘s like pandora‘s box where Linus is kind of forced to use it. If that makes sense.

Whatever's the outcome, the anthropomorphization in that message stinks 🤮

Regardless of how he got there, this is a pretty obnoxious and useless commit message...

I think Linus yelled at that LLM

Add two spaces after a line ends and add the /> symbol at the start of every line.

Example with:

I
wrote
this

Example without:

I wrote

this

Pretty wild seeing a programmer anthropomorphize their applications and mislabel them as "AI".

The Big Lai runs deep.

LLMs are a great tool for create bugs *

Fuck Linus.

ethical questions aside

And his here is the problem. Oh yeah lets just waive off the biggest critisicm on and issue of AI.

Hitler was a pretty neat guy, ethnic cleansing aside.

And before you say this, no, i am not comparing you to hitler, I'm only using an over the top example to hopefully make my point more clear.

Yeah, ethical, environmental, legal and fuck the consumer memory questions aside.

Any engineer fighting the use of LLM’s will ultimately find that their peers are achieving more in a shorter time then they are.

How do you measure that?

I am asking because I tried all the AI coding tools (copilot, cursor, claude code, etc.), even the paid versions, to find the utopia that doubles my productivity where lots of people are talking about. But everytime I find they take more time to generate code than me typing it, they rarely get things right the first time, they produce way more code than needed (which is much more to reason about) and all in all it takes longer to achieve something using "AI" instead of doing it on my own.

The worst results are the ones that seem to be correct, but can only be found incorrect on closer inspection. For instances, a part of my job is to write data collection applications for sensors; those sensors come with a datasheet (usually 100 to 200 pages) that explain how to "talk" to the sensor. Sounds perfect for an LLM - throw the datasheet in, get a small C application out. But either I get something that does not even compile, or something that seems to collect data, but configures the sensor completely wrong. Debugging that takes more time than just working through the datasheet myself and program it myself.

I remember people being angry at IDEs for making programming too easy and belittling code colourization as for children.

There will always be a crowd that wants everything to stay the way they know it.

find that their peers are achieving more in a shorter time then they are

I've been reading this for 3 years and all I've seen so far is my peers making more work for me by going on wild goose chases and making me wade through slop. I do believe there are ways to use it responsibly and productively but that is definitely not what is happening on average.

I don't understand why so many people decided to take "LLMs increase productivity" as an axiom instead of using their fucking eyes and looking at what is happening around them.

Well it’s about time those slackers caught up. And all they needed was a slop machine.

No, pragmatically speaking, LLMs and in general AI will only make code worse each day.

I think most misplaced critic comes from people not understanding or knowing the different types of AI. Some even hate all types of intelligence it seems.

Didn't you love T9 back in the days?

I’m just a movement generator.

Saying “it’s a text generator” cannot be the full force of your criticism, if the text it generates is incredibly useful. What could AI be on a screen if not a text generator?

Your views have got to be formed by WHAT text is generated given that AI cannot express itself through any other channel than text.

If you can give AI a deeply complex code base and ask it to find the likely cause of an obscure bug, and the solution is right, then you’ll have to acknowledge that some form of reasoning is going on, no?

And given that no other system has previously enabled this form of reasoning, it seems a shame to judge AI on semantics, rather than usefulness.

I find that depends a LOT on what you're asking the LLM to write, how well you're specifying it, etc. As for the code reviews, if it's code that matters: remember to open a new instance and ask the exact same question again on the code that has been reviewed and "fixed". Back a year ago, that could get you into a waffle-loop where the engine would change its mind back and forth about what's optimal and just oscillate between the two. These days they seem to record (and read) enough context to prevent that behavior, but I definitely get behavior of: "Are there any bugs?" "Yes, here are seven bugs." "Fix those bugs." "The bugs are fixed." (and they ususally really are...) "Are there any more bugs?" "No, we have fixed ALL the bugs." "Are you sure, look again." "Yes, I am sure we have found and fixed ALL the bugs." --- new context window --- "Are there any bugs?" "Yes, here are seven bugs." different bugs.

I did that on a bigger project and literally repeated 20 times, finding 140 real bugs - granted, the later bugs were getting pretty trivial / far out edge cases, but they were still real, still fixed, still denied there were any more bugs until opening a fresh context and asking again. This was on Google's Gemini 3.7 Flash High... Claude Opus 4.8+ seems quite a bit better about being able to continue in a context without becoming blind to issues "it has already solved."

Line completion is pretty damn handy, it's when you start asking for whole functions that things go downhill

There are already several large projects that have hundreds of thousands of users and those projects were almost completely vibe coded. It's almost the entire retro "recomp" scene now.

IMO: Specifically, they're bad at architecture and refactoring a small project into a large project. You have to jump through some hoops to make it craft something that needs more than 8m of context ram. If you can manage orchestration and multi-agents that don't need to know each others context, you can start to pull off bigger stuff, but it's not a forgone conclusion that it'll be fine. The worst output comes from it getting stuck on something and trying less likely answers successively until it works. You really have to watch for it to struggle and at the very least stop and start to try some new randoms.

Also, anything other than Claude-code with some really well-done project definitions is a waste of time.

Opus-class and higher models are actually great at writing code, but maybe it's language dependent? I primarily work in TypeScript.

At my company we are currently moving to implement AI for exactly that. A cost effective review buddy.

Various developers have tried to use it for coding and I think apart for one time scripts or getting an initial structure generated, noone is convinced of AI.

If we ever find a way to deal with the whole "burning the planet" issue, their best use case is to find and match patterns, not to imitate them.

The idea of using a language model to process search engine input isn't the dumbest part about Gemini and plenty of people report good results finding information easier and quicker with ChatGPT. In those cases, the deviation from rigid keywords is desirable because it can match results with related words rather than literal word-matching. Google Search already did a decent job at that (before the enshittification ran rampant), which more complex language models could improve even further.

The landmine is in their reproduction of those results, where the generated "summary" is the equivalent of a cunning bullshitter that convincingly sounds like he understood the topic but actually has no clue and just delivers a best guess. That's where the deviation becomes a risk of misinformation or introducing bugs.

They should narrow things down by finding the likely answers where that matters, not produce more stuff that humans will have to double-check.

LLMs suck at writing code,

So far.

I'm fortunate my distro allows me to install a different Linux Kernel, whether officially supported or otherwise.

I'd take a model whose training data is open source and legitimately obtained. The only ones I know about are Apertus and OLMo, and they aren't really competitive.

The problem with open weight models is that they are still hard to deploy. For DeepSeek for example you require NVIDIA HGX B200, these things are expensive.

Agreed, although simply using LLMs to do your thinking for you is only going to decrease your value in the workplace and make you more replaceable. The goal needs to be learning how to use AI to expand your thinking, counteract your biases, improve the quality of your work, and accelerate your learning.

This falls apart the second you look at it when you see that it doesn't take much learning to use that tool for most applications.

At the same time, people overly reliant on that tool take a big hit in developing and strengthening their own skilks and thinking capabilities.

Combine the two things if you still can.

People who are vehemently against any form of blockchain technology will be left behind. That may be difficult to hear, and even more difficult to accept, but that is the cold, hard, truth. Blockchain is a tool, and just like any tool it has uses. It can be used inappropriately, and it's valid to criticize those instances. But if you are plugging your ears and going "Lalala!" when people talk about actual valid use cases, or stubbornly pretending that blockchain has none, **you will be left behind**. You will be like office employees from the 80's and 90's who refused to use a computer and then suddenly found themselves without a job. It's your choice if you use blockchain or not. It's also every company's choice to not hire you because of your unwillingness to adapt. If you haven't learned it yet, this is your warning and your wake-up call: Learn how to use blockchain now, or find a different career. It's not going away, no matter how much you may *removed* and moan. There will soon be two classes of people, those who learned how to use this emerging technology, and those who refused. Things will be much harder for that second group. I don't say this to be snarky or smug. I hate blockchain myself, but I also recognize how fucking stupid I'd have to be to refuse to use it. It's a bitter pill to swallow, but it's also the truth.

Ah, there we go, I knew your rhetoric sounded familiar!

Seriously though, what "skills" are you building by learning how to use LLMs at your corporate job? How to tell a chatbot to do a task for you?

Sure, if you're trying to build a local LLM server from scratch, tune a minimalist harness with a RAG and custom skills, optimize context compaction and concurrent token throughput, and integrate all that into existing automation pipelines, that takes some technical skill. But that's a tiny fraction of people in the corporate world that need to do anything like that.

I could teach vast majority of corporate workers everything they will ever need to know about using "AI" in a half hour Lunch-N-Learn. There is no skill involved in asking a chatbot to move all your emails about "Project ABC" into a new Outlook folder called, "Project ABC" or asking a chatbot to, "draft a department memo about the new parking policy found in the HR SharePoint site."

It's not the next industrial revolution, it's not the beginning of the Singularity, it's not the dawn of a new age of our species. It is merely the latest hype-cycle of Capitalism that exists to siphon wealth from the population up to a handful of ultra wealthy billionaires and megacorps, and artificially prop up a dying economy just a little longer. And long term, we will discover the same thing about LLM use as we are finding out now about electronics in school and the mass proliferation of smart phones: That it stunts people's critical reasoning, reading comprehension, self-image, and social confidence.

agreed

the issue I have with how people use it is that it's easy to just create more shit that they spew out into the world

I've got no issue with people using tools to improve their work. I think it's fair to say that AI is enough like predictive text and spell check when it comes to tools you can use at work. and like any tool, the user is still responsible for the end result

my main problem with AI is that it is absolutely fucking our world (environmentally and economically) to create pure garbage that is thrown in my face every day. like, scale it back a bit everyone...

I am 100% fine with being left behind by a technology that I see as an attempted corporate surveillance scheme that has metastasized into a giant money pit. The difference from the 80's and 90's is people were actually using computers to make money back then. If any AI company is actually making more money than they're spending, I haven't heard about it, and you'd think they'd be advertising it. https://isaiprofitable.com/

I've been happy to fully sit this one out thus far. Don't feel like I'm missing a damn thing.

It's a classic scammer tactic to create a false sense of urgency, so people sign up for something without thinking.

Slow down, and do some critical thinking. The promise of AI is that you can type in a simple prompt and the AI will figure out all of the complicated computer stuff for you. How would someone be "left behind" by this technology? Do you suppose that experienced developers will lose the ability to write simple sentences?

I don't think people will loose the ability to type in "make that button green". Even if AI fulfills it's promises (which is doubtful, read up on the halting problem) someone who knows how computers work will be more employable than someone that only knows how to write prompts. Because the person that knows how computers work can also write prompts and will be better at doing that than someone that doesn't know anything about computers.

More important in this day and age is learning how to spot a scam. Someone pressuring you to do something immediately is likely scamming you.

Facts. Whether you like it or not. It's good to have a healthy amount of criticism towards AI and to fact check it and not blindly trust. But you can't wish it away or pretend it has zero merit. I guess you can....

Oh no I'll be left behind when the Titanic leaves port?! How utterly tragic!

Tragic for everyone who WASN'T "left behind", anyway.

Talk me more about how in the future, game controllers will be made obsolete by motion-sensing cameras, and we will all be jumping around in front of a big flat-screen TV playing party games and finger-gunning down bad guys in rail shooters!

Why can I not upvote and downvote?

good thing that downvotes can only be used to suppress narratives on lemmy. lol

Amen.

Still looks like they allow it as long as it gets approval by the core team. Not saying I agree with that choice, but there's still a very real chance LLM-generated code makes it into NetBSD.

Code generated by a large language model or similar technology, such as GitHub/Microsoft's Copilot, OpenAI's ChatGPT, or Facebook/Meta's Code Llama, is presumed to be tainted code, and must not be committed without prior written approval by core.

I've been thinking about doing just this, but I haven't seen any evidence that BSD is actually usable in real life yet.

Linux is actually good today, it's a viable alternative to the big corpos for virtually anybody, and the community is big enough to be sustainable. Also Steam has made Linux gaming a real thing.

If I could build a little pc to try BSD out on I would, I'd love for the competition to be there... But even a raspberry pi is like $500 today. It's insane. Fucking AI.

because it's a perfect storm. The way it's being managed by businesses is causing :

• Layoffs
• RAM shortages
• Propaganda
• Environmental problems

So, the tool is getting a bad rep. never mind that for over a decade, machine learning was something that was tinkered away and thought of as "one day this is going to be amazing", now it's being wielded by assholes trying to be TRILLIONAIRES and people are losing their jobs, access to technology, political and environmental collapse.

To top it off, before you could contextualise ML in statistics and say "it's a work in progress, some things are not quite right" but now you have ads saying "Replace your humans with our AI/Robots" as if it's 100% reliable, a drop in replacement and not going to provide any social ills.

I think the people most extreme about it haven't done much development.

Or they have, professionally, and associate it with very real AI psychosis at work.

Or they're in school and see LLMs ruining the student body, as we've seen some in other Lemmy posts.

It all comes down to ethical tool sourcing, and using those tools in moderation, responsibly. Like all sorts of other automation/technology that's popped up in programming...

But I guess I'm saying I can sympathize with absolute "no-LLM-code" positions, too. The way true vibe-coding has been peddled to the public really has done tremendous damage, and honestly the space deserves the negative attention its getting, and then some.

Over the past 5 years specifically, nearly every single piece of software I use from OS, to apps, to even legacy programs I trusted for decades have started to produce more bugs, are less stable, and less reliable.

At the same time, AI has been forced on people and AI using devs are proliferating while we still lack an understanding of AIs cognitive affects on us.

What exactly am I supposed to think? That all the programmers pushing AI on me like Crypto Bros while their programs work worse are smart, well intentioned people?

You can say whatever you want, I use the software, shits gotten absolutely terrible lately, and it's not just the purposefully enshitified software either.

Oh. No. I'm sorry, unrelated to the ai debate, but I have to correct this. Software architecture is intensely not raw logic sequences like you learn in uni. Everything before compilation is cooperative abstract modeling that does not necessarily impose any change at all in the binary. You are essentially teaching laymen that all programming languages for the same plattform are a creative choice, that extensibility, robustness, scaling or maintenance are just completely imagined fields of research. I'm sorry but that statement is complete nonsense misinformation. The binary itself, while it's a stretch; sure, in a way it can be considered only logic. From a certain angle. But that viewpoint is actively unhelpful for understanding what code is. The product is always so much more than the machine code. Consider writing raw arm instructions vs the obfuscated js that our engines churns through that decides most of our executing logic. All that is architecture and affects output only through the lens of the machine code. You have to enter the "code“ any time anything breaks, or is updated, if even a single thing is not as expected, or if you wanted to iterate on the program, or collaborate, you have to move into abstract land and then the architecture matters. The term spaghetti code does not mean that the logic is messy. The compiler wil usually remove most such mistakes. Instead the whole architecture, the "blueprint" the "workbench" is the magic of code, even if it is often obscure for the user is much more important for their experience than the raw logic.

In my experience, the code AI generates, can be VERY bad. Like, sure it might work, assuming it didn't randomly hallucinate and decide to randomly change a variable name half way through for no reason. But it'll be a code base only AI can read. It looks like the spaghetti code I wrote when I first started learning. I've watched people build up massive documents to help guide the AI and it still crap out a mess.

If you know how to code, and you know exactly what it needs to do, and you manage to contain it... it can produce code that's fine. But if you're not a SME in the language, libraries, frameworks, etc... and you just let AI do whatever, it's garbage-in garbage-out. But because the computer did it, because AI did it, people just implicitly trust it. Despite the warnings and general knowledge that AI will make mistakes.

AI has really shown how and why so many people fall for obvious lies on social media.

1. Simply using an AI that's run on a corporate data centers encourages the construction of yet more data centers, with all of the environmental/climate negatives they bring, as well as local harms they induce on the people living near them, such as increased electricity rates.

2. Using corporate AI directly helps the financial situations of those giant corporations (by boosting usage/user numbers, they are able to attract more investment capital), most of which are ran by right-wing CEO's who are more than willing to collaborate with and fund fascist governments to ensure that they are not regulated in search of both maximum profits. Some of these companies, such as Nvidia, Palantir and Oracle, genuinely appear to be seeking to use these tools for what would previously be considered crackpot conspiracy theory levels of public control and surveillance.

3. I'm wary of the potential effects of AI usage resulting in declining or stagnating critical thinking based on some preliminary studies (example 1, example 2).

4. I do not think LLM's can be "just a tool" as many claim.

5. 99% of LLM's (the ones virtually everyone use) are trained on copyrighted code that is incompatible with GPL licensed projects. LLM's have a 3 to 10% chance to unknowingly reproduce the copyrighted code perfectly, introducing legal plagiarism into an open-source project, which could easily open up the developers to being sued in the future if companies scanned open source repos for copyright violations.

6. Many courts are ruling that LLM generated code cannot be copyrighted at all, meaning that it also cannot be legally made GPL, losing the protections from corporate exploration the GPL normally grants. https://fsfe.org/news/2026/news-20260825-01.html

why people treat AI-generated code like it’s the Plague 2

It's just another part of the grift. The so-called "resistance" actually promotes the Big Lai that "AI" exists.

FreeBSD is also slopped, sadly. https://forums.freebsd.org/threads/will-freebsd-adopt-a-no-ai-policy-or-such.100101/

why not a genuine piece of pre-AI coding art like Windows Vista, Windows ME or the absolute pinacle of software: Slackware 9.1 before they switched to the awful X.org server from the beautiful XFree86?

We need to finish Hurd.

Maybe get an M2-series Mac and go all-in on Asahi Linux?

Or live the 90s dream with Haiku OS (they just released beta 6)?

The “never ai” list is getting pretty short

If you really think that FreeBSD whose developers often use Macs on desktop is any different, think again. Maybe OpenBSD but certainly not FreeBSD.

I agree that the corpos being linked to fascism doesn't make the technology fascist, but plenty of people are switching from MS and Google for that among other reasons. Acting like any interaction with these companies is just as bad as going all in is completely incorrect. All small acts of resistance add up and are important, even if you can't totally avoid a company that is causing harm. Taking a bus is not as harmful as driving a car just because they both use fossil fuels.

there are robust anti Microsoft and anti Google movements getting more and more mainstream over time. there are plenty of great alternatives to MS and Google software and services. Linux and FOSS as a whole is benefiting immensely from anti Microsoft sentiment right now. We can and should boycott these two terrible companies

"Avoiding fascist-supporting companies is inconvenient" is a poor excuse, especially when it's so easy to degoogle compared to 10 years ago.

A lot of stuff was called ai and wasn't intelligent. Stop using this argument, it weakens any discussion that wants to argue against ai/llm usage

You're a statistical organic blob with absolutely zero intelligence.

Unfortunately for you, the English language has shifted to include LLMs under the "AI" umbrella.