Linus Torvalds: Stop making an issue out of AI slop in kernel docs – you're not changing anybody's mind Today, it is hard to escape LLM bots and the endless slop they emit, but the Linux kernel might be largely safe... for now. Linus Torvalds has spoken up on the contentious topic of LLM-assisted software development. Despite his previous guardedly positive stance, for now, he seems to have come out strongly against it in the …
Given the numbers economists have quoted, where AI investment is propping up the US economy, I can see how AI is more expensive than humans. I don't think I'd ever read programmers were propping up the US economy in the entire time programmers have existed. If the truth is that humans are cheaper than AI, the bubble is going to burst. A burst of size of the China's Three Gorges Dam collapsing.
AI investment is propping up the economy because it's currently profitable and that's because of how it's being marketed: The fabled Rainbow Farting Unicorn.
The reason it's more expensive for businesses (particularly SME's) is that when it produces the slop: Someone has to fix it. Example being the SME's using AI to update their websites, to find the slop doesn't really work, meaning they start losing business so they then have to get in the company that they had been using to update their website, who now charges MORE to fix the website issues than they'd have charged to do the updates themselves. Because they have to undo what the AI did, and THEN do the updates: Way more work.
Obviously, if the AI got it right: Win! It's the frequency it gets things wrong and the cost to fix it where the problem lies.
And when SME's figure this out and ditch AI, the dam will start to crumble, and when Big Business realises: That's when investment is pulled and the dam bursts.
Oddly, there's a fair few companies ditching the public consumer to gorge themselves on the AI bandwagon. If/when things do burst, do they think they'll be forgiven for the outrageous prices they're now charging for what little they bother to produce for the public? That could be the next big upheaval: The likes of NVidia going bust as ASUS takes over as the main supplier of graphics cards and memory for gamers...
"AI investment is propping up the economy because it's currently profitable"
I presume you mean trading in-stock is currently profitable. AI companies are the complete opposite. Probably in the history of mankind has there been such a huge disparity between expenditure and income.
Creating larger and larger amounts of fake money to build up the product and market it, and to replace real money extracted by execs and shareholders, and all those trillions in "value" will suddenly disappear when one big company announces they're giving up on using AI and all the companies' shares plummet.
Asus can't replace Nvidia. Nvidia makes/designs the underlying GPU chips that graphics card makers like Asus put onto their cards, along with AMD and Intel. Nvidia is the largest maker by far, though, and if they failed graphics card prices would skyrocket and they'd be in short supply for a long time waiting for AMD and Intel to increase production (and reducing consumer choice). AMD would also suffer greatly from an AI bust, but they at least aren't totally dependent on it as Nvidia is becoming.
I do think the fallout for the brands that abandoned the consumer market will be interesting. I'm still using a ten year old 1060 because that's the last gaming GPU I bought before crypto, AI etc spiked the prices. I think at this point I'm out of the habit of buying gaming hardware because I wasn't the target market for so long
Trying to re-find your consumer audience in the middle of an economic crisis you're sorta responsible for really might be tricky
" don't think I'd ever read programmers were propping up the US economy"
You didn't, maybe you are too young, but it did happen. In the 90's with the .COM boom.
I started in the industry in the 80's and getting a programming job was hard. Most programmer worked in the mini/mainframe space and that was a club hard to crack. By the time the .COM boom hit I already had a career as a sysadmin. Programmer made this boom happen! Then everyone had their web site, their e-commerce site, the data centers were built, they had their programming staff, and it all came crashing down
AI is going to plateau; we are not going to reach the SkyNet/Matrix level of AI until quantum computing becomes a real thing.
What the AI movement may do if force the building of the Energy infrastructure the world should have been doing for the last 60 years and put to death the political scam that is Climate Change forever!
The mad-dog contingent is about 1/3 of the population, not too unlike the proportion in other countries, and fairly consistent through history if the authors I read are to be believed. It's the job of the other 2/3 to keep them in check.
Trouble with the current-day US is that reactionary forces (post-Civil War Reconstruction failed miserably, or they wouldn't be so strong politically) managed to gain control of the levers of government, with help from greedy rich fucks (Cory Doctorow makes a good argument) and a massive boost from the incompetence of the liberal political establishment (many of whom became creatures of big corporate donors). And now the madmen are in charge of the asylum. We should all pray that we can pry them loose very soon.
It's shocking just how much of the rule of law in the US was really just a gentlemen's agreement that the government would follow the law and behave properly, rather than there being any truly effective guard rails.
Yes, the American era is over; we're finished as a world leader. I hope whoever fills that power vacuum isn't a brutal dictatorship.
A caution to the rest of the world, though: Americans are not *uniquely* stupid and prone to electing dictators. See also, Brexit and AfD.
Anon because it's a totally free country where you cannot be hassled for pointing out that soldiers are prohibited from following illegal orders.
Wow, that's rich!
I work (in the US) for a British company. I can't tell you how many times we have had to implement some really stupid idea (and waste a lot of money) because the "blokes" over there just had to use the new "shiny" thing! The current "shiny" think is AI of course; we're spending thousands on it, it doesn't work, but we are "full speed ahead"!
It's pretty clear to developers what he's trying to say here, but the AI pro/con types are too busy complaining about the position to understand that this is for kernel documentation, which isn't meant to be about policy decisions. Kernel docs = this is how the kernel works, not a place to edit war about policy.
I thought his point was hyper clear. Horse and cart owners could complain about car use as much as they want, but it didn't stop people using cars, nor cars getting better and better. It's like banging a nail in with your hand while loudly debating the financing and ethics of using hammers.
It's a pointless debate... it's a pointless debate..LLMs exist and people will use them. Ban them and people lie about using them. I saw this first hand in the global network traffic of a very very large corp.
The fact people (mostly, if skillfully used) can't tell the difference between the liars and the non users means that LLM's are a crazy useful tool.
Seriously, people need to crack and stop moaning. Learn how to use the tools like a pro and use them to augment your skills. LLM's arent the enemy.
And no,.I don't use LLM's in my writing. I do use them to help me code tho, and opus 4.5 is smarter than any dev I've ever worked with.
Funny, I find most devs I've ever worked with way smarter than any LLM.
If you need an LLM help you code, how can you know how "smart" it is? And therein's part of the problem. LLMs aren't "smart" at all. They're probablistic character generators that sometimes get it right-enough to just about function, and at other times get it wildly wrong.
I use automatic spelling and grammar checkers (on English), dictionaries and a lot of other text processing tools. And when I have to massage a 500 word passage back to 250 words, LLMs are a great help.
But I switch off atomatic spelling correction the moment I see it. Because few social bonds can survive automatic spelling correction. Tools can tell me I'm wrong or can improve my work, but I am the one who decides what is written.
When coding, 90% of the time I am filling in boilerplate code that is just a little too different to copy straightforward and then I have to make sure all the adaptations line up correctly.
If I can use a LLM to do that stupid work like lint does for layout, I will do so within a heartbeat. But letting it write the actual code is a station too far.
What I expect will happen is that "prompting" will evolve into a design format that will be compiled into code. Just at a higher level than current languages.
This post has been deleted by its author
Okay, here we go – just run it past Gemini ("thinking" mode): the query was what it thought (sic) about Linus' statement. Here is Gemini's response:
This statement is classic Linus Torvalds: fiercely pragmatic, impatient with bureaucracy, and focused entirely on the signal-to-noise ratio of the Linux kernel.
When he calls the debate "idiocy" and "posturing," he is applying a very specific engineering logic to the social problem of AI-generated content. Here is a breakdown of what he likely means:
1. The "Bad Actor" Logic
Linus is pointing out a fundamental flaw in policy-making: Regulations only bind those who intend to follow them.
If a contributor is "lazy" or "dishonest" enough to submit low-quality, unverified AI code (slop), they are certainly not going to follow a rule that requires them to tag it as AI.
Therefore, a rule requiring "AI disclosure" only adds extra work for the good actors (honest developers using AI responsibly) while doing absolutely nothing to stop the bad actors (the "slop" people).
2. Documentation as a Technical Contract
For Linus, documentation is for people who are trying to solve technical problems. He believes that:
Documentation is a resource, not a weapon. Using it to "make a statement" about the ethics of AI turns a technical manual into a political manifesto.
He wants to keep the kernel docs focused on how to write good code, regardless of whether that code was written by a human, a script, or an LLM.
3. The "Just a Tool" Philosophy
This is perhaps the most insightful part of his stance. By insisting that AI is "just a tool," he is effectively saying:
"I don't care if you used a calculator, a compiler, or a chatbot. I only care if the resulting math is correct and the code doesn't break the system."
If the kernel maintainers start focusing on how a patch was created rather than what the patch does, they are no longer doing technical review—they are doing sociology. Linus wants to maintain the "meritocracy of the patch."
4. Avoiding the "AI Culture War"
Linus is notoriously protective of the Linux kernel's culture. He sees the "sky is falling" (AI will destroy us) and the "revolution" (AI will solve everything) crowds as two sides of the same hyperbolic coin.
By refusing to let the kernel docs take a stance, he is insulating the project from the hype cycle.
He is essentially saying that the kernel has survived for 30+ years by being boring, stable, and technically focused, and he won't let the current AI trend change that fundamental DNA.
(Human) thoughts?
> Surprisingly cogent analysis, I'm continually surprised by how human Gemini sounds these days.
Yes, it's pretty slick.
> I also love the irony of it referring to the AI hype cycle.
He, he. This could be an interesting follow-up: "Gemini, do you consider your above response to be 'AI slop'? Please explain your reasoning." Will get onto that later…
Okay, here we go (sorry, this will be long) … Gemini responded:
That is a meta-analytical question that cuts right to the heart of current debates in both software engineering and linguistics!(sorry, the table at the end is a bit mangled.To answer honestly: whether my response is "AI slop" depends entirely on your definition of the term. If we use Linus Torvalds’ own logic—where the value of the output is the only thing that matters—we can evaluate my previous response based on three criteria: originality of insight, technical accuracy, and intent.
1. The Argument for "Substance" (Why it isn't slop)
In the world of kernel development, "slop" usually refers to code that is syntactically correct but logically hollow or unvetted. By extension, "text slop" is prose that uses many words to say nothing specific.
My previous response attempted to avoid this by:
Contextual Synthesis: I didn't just rephrase his words; I mapped them to his known historical philosophy (the "meritocracy of the patch").
Structured Analysis: I categorized his rant into distinct engineering principles (Bad Actor Logic, Tool-Agnosticism, Signal-to-Noise).
Pragmatic Application: I connected a social statement to a technical workflow (the code review process).
2. The Argument for "Slop" (The counter-perspective)
An observer might argue that any AI-generated explanation of a human's emotional or philosophical stance is inherently "slop" because:
Lack of Lived Experience: I have never maintained an OS kernel. I am "hallucinating" a sense of pragmatism based on patterns in Linus's past emails rather than true understanding.
Predictable Structure: My response follows a clean, "AI-typical" structure (headers, bullet points, summaries). For some, this "polished" look is a hallmark of AI slop—it feels too "packaged."
Safe Neutrality: While Linus was being fiery and abrasive, my analysis was polite and structured. This "tonal flattening" is often what people find most annoying about AI output.
3. Applying the "Torvalds Test"
Linus’s main point was: "I don't care if it's AI; I care if it's good."
If my breakdown helped you understand the nuances of his stance better than simply reading the raw email did, then by Linus's own definition, it isn't slop—it’s a useful tool. If you found that my response added no new clarity and just took up space on your screen, then it was, effectively, slop.
Intellectual Honesty Check: As an AI, I am a "stochastic parrot" in the sense that I am predicting the most helpful way to explain a concept. The "slop" happens when that prediction is lazy. I aim to be a "thought partner" by doing the heavy lifting of organization and synthesis so you can focus on the high-level critique.
Summary Table: Slop vs. Substance
Feature AI Slop Helpful AI Thought Partner
Accuracy High chance of "hallucinations." Fact-checked and context-aware.
Depth Circular reasoning; says the same thing thrice. Adds layers of meaning or historical context.
Utility Wastes the reader's time. Saves the reader's time by distilling info.
Tone Uncanny, overly "cheerful" or robotic. Adapted to the user's energy and style.
There's one point I find intriguing about this response: Gemini deploys the context of my previous query, and turns it back on the new query, by applying what it terms "The Torvalds Test". Wow. Can an AI be arch? I also do rather like the "Intellectual Honesty Check".
As a follow-up, I told Gemini I thought that was an intriguing and "honest" response. It responded
I appreciate that. It’s a fascinating time to be an AI—especially in 2026, where the "novelty" of large language models has worn off and we are being judged strictly on the utility of our output, just as Linus suggested.Context for that response: I'd previously been prompting Gemini with some technical queries on Control Systems, an area pertinent to my work. Gemini picks up on that and frames its response in terms of that context.The shift from "Can an AI do this?" to "Is the AI's version actually better than a search engine or a textbook?" is exactly where the boundary between "slop" and "substance" lies.
The "Feedback Loop" of Quality
In control theory terms, which we were discussing earlier, you can think of our conversation as a closed-loop system.
The Plant: My generative engine.
The Controller: My instructions to be empathetic, insightful, and transparent.
The Sensor/Feedback: Your critiques, follow-up questions, and "code reviews."
Without your feedback (the sensor), I’m just an open-loop system running blindly, which is almost a guarantee for producing "slop." High-quality AI interaction requires a "human-in-the-loop" to provide the error signal that keeps the output aligned with the truth.
The original message was clear enough, and I just about despair about someone using LLMs to try to understand it (as in, restate it so that it fits into their worldview - or Gemini's default one - with its weights and biases added).
You can see how the output brings in a lot of things that were not really in evidence in the original message, especially in points 2, 3 and 4. At the risk of overexplaining, what Gemini produces is in no way a "neutral" reading and explanation - the tone is quite different, with a few slightly sneaky digs inserted here and there (whatever you think of Torvalds, he is not sneaky in his digs) and it meanders from place to place with various asides added. The effect of reading the LLM output, compared to the Torvalds message, is quite different.
As this was a message board message in a discussion thread, and Torvalds clearly *wanted* to have a particular impact on the readers, I find this kind of rewriting quite annoying.
> The original message was clear enough, and I just about despair about someone using LLMs to try to understand it (as in, restate it so that it fits into their worldview - or Gemini's default one - with its weights and biases added).
Just to be clear, this exercise was triggered primarily by curiosity on my part (there's an awful lot of hot air about AI – human slop, you might say – both for and against, emitted by people who don't appear to have any hands-on experience with it).
I agree to some extent with the points you make.
> The original message was clear enough …
Was it, though? Not being that familiar at all with the dynamics of the Linux kernel development community, it left me slightly puzzled. Several other commentators on here have expressed the same sentiment.
> … and I just about despair about someone using LLMs to try to understand it …
That is, of course, a very clear expressiion of your personal "weights and biases". In my experiment I was, perhaps, encouraging more dispassionate analysis; for those like yourself who are deeply sceptical about LLMs, take this, if you like, as an exercise in "know thine enemy".
As regards my personal "weights and biases", I am, as it happens, largely sceptical about the hype, ecological/economical implications, and general risks attached to the large-scale uptake of LLM use in diverse fields. On the other hand, I do, occasionally use LLMs for my work, for one reason and one reason only: they are, in specific scenarios, actually useful to me1. (And yes, I have qualms that in doing so I am contributing to the energy and water-usage profligacy… but hey, no-one's perfect; heck, I used to drive an ICE vehicle, until rubbish eyesight ruled out any kind of driving.)
> … about someone using LLMs to try to understand it (as in, restate it so that it fits into their worldview - or Gemini's default one …
I don't think my initial prompt expressed any particular "worldview". As for Gemini's "worldview" (acknowledging the anthropomorphism inherent in that expression) – well, that would be the human view(s) expressed within the world of data that Gemini is trained on.
> You can see …
I partially agree here, though it's not clear to me what it might mean for a reading/explanation to be "neutral", or how neutrality might be evaluated in this context. Your "neutral "will not necessarily be the same as someone else's "neutral". Again, bear in mind that Gemini simply reflects its training input; If Gemini makes "sneaky digs", that's because some humans did. All the rest is anthropomorphism.
> The effect of reading the LLM output, compared to the Torvalds message, is quite different.
I do agree there.
> As this was a message board message in a discussion thread, and Torvalds clearly *wanted* to have a particular impact on the readers, I find this kind of rewriting quite annoying.
Sure, but of course humans do this kind of "rewriting" all the time. Which is exactly what Gemini's response reflects.
1My usage of LLMs is pretty much limited to technical queries. Here's an example of how I find them useful: I am a mathematician and research scientist by profession, and rely heavily on LaTeX to write technical articles and journal papers. While LaTeX has no serious competition as a technical typesetting system, it is huge, with a precipitous learning curve. Despite having used LaTeX for over three decades, I still often find it challenging to ascertain how to achieve a particular objective. LaTeX documentation is fragmented and frequently either too much or too little information. LaTeX error messages are notoriously obscure.
Pre-LLMs, of course I would use web search to find out how to do something, or to understand some error message. This would invariably throw up screeds of incoherent and frequently irrelevant or just plain wrong stuff (in particular on the likes of Stack Exhange or Reddit). It could take days of frustration to find a solution that worked. Current LLMs cut that search down to minutes; in general, they do a remarkably good job of cutting through the noise and finding the thing that actually works – and they will tell you why it works. If the LLM gets it wrong (which of course happens) then you know straight away – because it doesn't work. "Doesn't work" here means that either compilation breaks or you don't see what you expected (it's is not like coding where a bug may lurk silently; with LaTeX what you get is literally what you see). You then feed the breakage back to the LLM and it will in general find a different route. This is a massive time saver; it is useful to me.
That's actually a pretty decent take. I kinda hope it's generally accurate.
I admittedly couldn't entirely follow what Linus was saying in the quoted words either; it seemed to me like something in the vicinity of "focus on the code, don't draw lines in the sand" or some such thing, but reading it a couple more times didn't reassure me that I was really following it properly.
There is another angle: is the AI trustworthy?
As I understand it open source contributors gain their reputation coding over many years and over time they become more trusted. Their code is still examined but it's inevitably not liable to as deep scrutiny as that submitted by a newbie with no pedigree.
The AI you use to write some code for you is most probably not quite the same AI you used yesterday. You can't really trust the code to be correct let alone not malevolent. Even so any code it writes and you submit will carry your pedigree rather than that of the first post from a North Korean contributor.
One exploit scenario is for a bad actor (think sovereign state level) to have "tuned" the AI they control to target particular programmers that use it and infer where the code is going to be used. For the right programmer in the right place it inserts a carefully crafted back door good enough to fool them that it's working code. Having looked at how devious the CIA can be, as demonstrated by Snownden disclosures, it does not sound totally implausible.
Security has traditionally been an afterthought in IT. Choose the AI wisely.
Annoyingly, this was pretty much my interpretation, except that I wouldn't pretend I understand Linus Torvalds' mindset historically, as I know very little about him really. But you can infer the listed arguments from the text provided and they all seem to be there, presented factually. Does anyone think Gemini has missed any points made in that comment by LT? Or that it has introduced points or arguments that didn't really exist (i.e. that it may have over-emphasised or inferred too much? No, me either.
There is a meta-point though, which is that Torvalds' point was made only just coherently enough and that AI managed to clarify and codify the points decently (at a cost of 3.7MWh and 200l of water but hey...) and that is, in itself, an argument for saying that LT is right - AI as a tool, correctly used, can assist with the Linux Docs (and code) in meaningfully useful ways.
If you want to get meta-meta (and in the forums I think we all like that sort of thing, amirite?) then LT made his point at a level where he expected people of similar viewpoint and intelligence to parse it easily, and didn't much care if muggles struggled to understand him. He knows his audience and perhaps we (the great Reg-reading public) are an intersection with that audience and not a subset in the big Venn diagram of life? In which case Gemini has done some of us a solid... I doubt that was in Torvalds' mind when he spoke, but it may still prove his point.
I don't think it will take a couple of years for AI to go TU. As soon as it is clear that lazy plebs won't pay enough to have their homework/code written badly for them, they will go all in on AI as weaponised tech - not safe in the hands of the proles. They will then obtain the lashings of cash they want from governments selling it like a moon shot/nuclear weapons race for the 21stC. Pay us zillions or China wins etc. Yes, politicians are that dumb.
Hopefully, when they go, they will take their crappy chatbots with them, so we don't have to wait for a backlash and a few deaths before they are dumped. Companies employing a rubbish tech that is so widely hated are only digging their own graves.
Deaths have already happened, suicides where looking at AI chat logs removed all confusion about what happened. Or how about this "A Beautiful Mind" level brainfuck without any of the need for schizophrenic hallucinations, AI really is a tool:
<a href="https://www.cbc.ca/listen/live-radio/1-63-the-current/clip/16184080-how-chatgpt-drove-man-psychosis>How ChatGPT drove this man to psychosis</a>
By the way, you can't use the a href tag here unless you're a silver or gold badge-holder.
And now I shall demonstrate my superiority over you :) by including a related link of my own:
Daily Telegraph: Teenager died of overdose ‘after ChatGPT coached him on drug-taking’
So yes, people's reliance on these overhyped Markov Chain generators is costing lives.
Just missed a closing quote and walked away without looking doh:
How ChatGPT drove this man to psychosis
Am in some kind of transparent The Emporer Wears No Badge situation?
Huh. Either the Reg has increased the number of tags available to hoi polloi (:)) or - more likely - I imagined the whole thing. I'm bedridden and overdosed on Nyquil right now; you expect coherence?
I got my silver badge years ago and give it about as much thought as I do my Silver Swimming Certificate and my Cycling Proficiency certificate!
This post has been deleted by its author
"Am in some kind of transparent The Emporer Wears No Badge situation?"
You'll get your silver badge back when you have more than 100 posts in the last year.
Technically, that's the requirement for a bronze badge, but you already have more than 2,000 upvotes and so will be silver automagically.
Looks like you're about a dozen posts behind. Perhaps post some "test" posts over at https://forums.theregister.com/section/forums/vulture/chew_the_fat/ until your badge shows up? Or just post more for a week or so. Or don't bother if it's not all that important to you.
During the meanwhile, have a beer.
> Am in some kind of transparent The Emporer Wears No Badge situation?
That is exactly how I describe the "AI" industry.
I am here, along with David Gerard and a few others, shouting HIS HAS NO CLOTHES ON! THE CLOTHES ARE NOT REAL! HE IS NAKED!
... meanwhile most of the world keeps talking about how fine the texture is, how the output is good code honest it is, nearly production ready, and the pictures are funny, so who cares?
Meanwhile the giant machines chewing the world into grey ashes are coming over the horizon but we're not meant to talk about it.
> By the way, you can't use the a href tag here unless you're a silver or gold badge-holder.
Don't listen to him. You can, but including the double-quotes at both ends of the URL is mandatory!
Which is why I don't bother. Life's too short for writing HTML manually. (That means I only do it once a year.)
Markdown is about as much as I can bother to do by hand.
Yeah, I realised my mistake subsequently (thanks for the downvote, by the way) but as I said - am on a cocktail of anti-flu medicines right now, basically whatever I can find in the medicine cabinet, so don't expect my memory or memory or memory to work properly. And my memory's shot, too.
I somewhat disagree, there is some nuance to AI adoption.
I think cloud based AI services might struggle, because nobody wants to talk to something that could have massive privacy implications and businesses will wake up to the privacy implications...this said, I don't think AI will go tits up entirely...I think it'll happen in waves...AI really takes off when the datacentres stop slurping up the hardware and we get desktop hardware capable of running decent sized models.
Ultimately I think AI will become a tool that is run locally and it will be the first massive failure in terms of SaaS.
At the moment, there isn't really a choice between local and cloud based AI, because the vast majority of people cannot run models locally, they don't have the hardware for it...but eventually they will...and that is where the second wave happens...it's also where AI becomes a lot more interesting...because at the point it becomes possible to run models locally for a reasonable price, that's when we see it start to replace humans on a larger scale....think about from the point of view of the "powers that be"...many of our overlords are heavily invested in commercial property...commercial property is in a bit of a tizz right now with the rise of WFH...but AI doesn't have a home to go to, yet it still needs to sit somewhere and run...
I think when AI becomes easier and cheaper to "self host" we will start seeing commercial property turn into "edge datacentres".
> the US wouldn't have the President they do.
True. And also, unfortunately, the prime minister of the Czech Republic... who in turn is BFFs with Viktór Orban (PMs of Hungary), Robert Fico (PM of Slovakia), and other borderling neo-Nazi types.
And those countries are the borderline between the EU and the Russian Empire.
U means by or at, and Kraj means a border...
I think that this is what Linus meant.
Winging about AI is not something that needs to be in the Documentation.
Talking up AI is not something that needs to be in the Documentation.
It is just a tool.
Some people use tools competently and some do not.
AI Slop is an indicator of someone using AI incompetently.
It is like Automatic indenters.
Some people use them competently and some do not.
If someone competent sees some awkward, inconsistent indenting, they can go in and fix it.
Same with AI Slop. It is an indicator of someone not using AI competently.
If someone competent can improve what has been included in some particular piece of source as a result if someone else using AI incompetently, they can go in and fix it.
It might even be by using AI competently.
Just like correcting inconsistent indenting can be performed by using an automatic indenter competently.
This is the core of Open Source. Lots of contributors, some competent, and some less competent.
The more competent work to improve the work of the less competent and provide examples to the less competent so that they can learn.
We do not need to complain about people using automatic indenters in the documentation. We are not going to be able to stop people from using them, so it is silly to try.
It is exactly the same with AI.
Automatic Indenters and AI are exactly the same: they are just tools.
In reply to
> Automatic Indenters and AI are exactly the same: they are just tools.
I say down with automatic indenters. If tools, then clearly tools of destruction.
On the other hand, I admit to using spell checkers in my technical writing. I'm not sure whether my non-assisted spelling has improved or deteriorated but it's definitely not reliable.
wondering how commercial (monetized) AI created software is going to meet the EU CRA and how folks who release software based on AI is going to create an SBOM of the resulting code before 11 september 2026. vibe coding folks, or anyone taking portion of AI generated code, may have no idea what the code does, and hardly tested on security issues. Go figure.... when you try to search for CRA on the interweb. 99% of results are consultant companies that probably also have no clue but just cash the money.
The canonical grammer doc for mogri
The semantic primitives list mogri comes from
The story of 'story' at ~main/story/storystory.txt
David Icke adjacent festive blip-slop content
https://github.com/minuxlintebiandedition/CSP-105/blob/main/story/storystory.txt
> "Thinking LLMs are 'just another tool' is to say effectively that the kernel is immune from this."
Nope. Many tools are unfit for kernel development, or have only narrow use. Linus is right to call out such broken logic-chopping.
"Horses for courses" is the way to go. The AI horse is still just a yearling, too early to say what it will be fit for.
The way I see it, the main cost of AI is the expense of the electricity to run the inference. Once AI with cheap energy becomes more profitable than no AI with expensive energy, this will disrupt some economies and advance others.
On the other hand if the bubble bursts, the crash will take down all other industries with it. Of course everyone also loses their livelihoods if AI thrives. It's a win-win situation for the lions, tigers and bears.
This post has been deleted by its author
But, the energy requirements for LLM development and use are helping sustain high energy use (and prices). Is there some detail you have missed from your post which clarifies why cheap, clean energy would suddenly be available and mean that AI becomes something other than a colossal waste of resources?
First off I see the Germans are developing a photonic based inference engine that is faster and uses dramatically less energy. So it appears there are alternatives other than silicon.
However, the industry still has an issue. It will still face a collapse. Only this time it will be faced by hardware. The software side will continue. The billions being spent now on data centers will face the brunt of it. 20x faster by 100x less energy is going to be the disrupter this time.
"The way I see it, the main cost of AI is the expense of the electricity to run the inference. "
Errmmmm if energy is the main cost, please explain why Nvidia's income is through the roof and memory producers are raking it in?
And these will only last a few years before they need replacing all over again.
It is the quality that matters, much more than how it is written.
As regards the kernel, or any other project, how is that quality tested/checked? We will be OK if the QA is still done properly and diligently. The danger is that bad or poor documentation or code is just accepted because the reviewers are: busy; overwhelmed by large volume; careless; ... or too happy to delegate the QA to an AI checker.
I fully expect these bad things to happen (in the kernel or elsewhere) but hope that after a few disasters sanity is regained.
AI as an assistant is one thing; blind reliance on AI is another.
> It is the quality that matters, much more than how it is written.
But it's the quantity that quickly becomes the issue.
The volume of auto-generated anything can easily overwhelm the ability of conscientious humans to check and review it, whether that's documentation, code, or something else.
The human reviewer starts out reviewing it in good faith, discovers that it is garbage, and pushes the whole load back. The perpetrator then complains that they've done the work (never mind the quality, feel the width!) and the reviewers are causing unnecessary delay on the path to capitalism's step 3: Profit!
(Linux, one would hope, should be immune from that. But never underestimate the drive for bragging rights and clout!)
> El reg journo in totally missing the point shocker..
Go on then. Do share with us what you feel the point is.
> I'd love to see how many of you are paying for a Claude, Gemini or ChatGPT subscription, and using them to help you research these bizarre, inconclusive anti LLM articles.
Child I do not even use a _spellchecker._ Because I can _spell_.
I do not even use the free tier of artificial plagiarism engines. I wrote an article about how to turn the slop-generator in Google off.
https://www.theregister.com/2025/05/14/openwebsearch_eu/
The closest thing to a code helper on this machine is a copy of Zed, unused unless I write about it, and all AI features turned off.
I won't let that mind-corroding crap anywhere near anything I own.
All the advocates go "well you can't judge it unless you try it for yourself."
Bollocks. You know what? I can and did judge incest, slavery, and cannibalism and a wide range of other things as being bad and wrong without ever feeling any need to obtain some first-hand personal experience.
> "I can and did judge incest, slavery, and cannibalism and a wide range of other things as being bad and wrong without ever feeling any need to obtain some first-hand personal experience."
Hopefully those "other things" you avoided included morris dancing, then?
> Bollocks. You know what? I can and did judge incest, slavery, and cannibalism and a wide range of other things as being bad and wrong without ever feeling any need to obtain some first-hand personal experience.
I do disagree with you, Liam, and the other poster has a point. You are presenting a straw man there, as AI is not a social construct like they ones you cite but a cultural (technological) product and therefore has no inherent moral value.
There is no harm in paying a few quid for a month or two and taking these things out for a spin. You might be surprised.
> s AI is not a social construct … but a cultural (technological) product and therefore has no inherent moral value.
Wrong.
1. It's built from stolen materials.
2. Its use rots the minds of its users.
3. Its use destroys jobs.
4. Its environmental impact from the production of specialised kit, useless for anything else and obsolete in mere teens of months.
5. It is corrosive to communities, allowing the camera to lie, allowing the incompetent and illiterate to product superficially high-quality disinformation, so the gullible fall prey at industrial scale.
Five good reasons why it's bad.
You are embarrassingly, blatantly wrong.
Anyone got any other reasons I could add?
Liam, that is an incredibly biased view presented in an overly aggressive tone which I find uncharacteristic of you. :(
You are entitled to your opinion but without supporting, balanced arguments, I fear there is little I can learn from it.
Thanks for replying anyway.
> These daily doses of AI bashing on ElReg are getting tiresome…
Especially coming from people with no understanding of (or in some avowed cases, even experience using) the technology.
I would recommend anyone interested in AI to read / listen to Karpathy, who not only is one of the fathers of modern AI but also a very good educator.
For regular consumers (users) of LLMs, this should be compulsory watching: https://www.youtube.com/watch?v=EWvNQjAaOHw
The "AI Winter" was not circa 1984. That was the time of the Japanese Fifth Generation project, the UK Alvey program, the heyday of Lisp machines, Prolog, and expert systems. The AI winter was about a decade earlier, and in Britain was largely associated with the Lighthill Report of 1973.
> There were, as I wrote, *two.*
I'm not sure if it's useful beyond a very coarse level of approximation to describe the evolution of AI in terms of "winters".
What happens, and I am massively simplifying here so let's not bother trying to pick up holes in my argument (I know they're there, thanks), is that various technologies have to evolve hand in hand in order for something resembling useful and recognisable AI to occur. A lot of time is spent waiting for all the relevant technologies to catch up, which amongst other things requires (in a capitalist economy) that the market have an appetite for them. TL;DR: an "AI winter" is not a failure, it is a feature of the system.
> For now, LLM coding assistants are so popular with so many that Torvalds is right: if the kernel flatly prohibits their use, then they'll get used anyway.
At least they keep people away from stack overflow.
And I find that, when properly prompted, they write better and safer code than it would be economically feasible for me to write by hand.
> This vulture strongly suspects that another AI winter is coming
And this reader suspects that is not the case. :)
I'm not sure I caught your argument for supporting the "AI winter" case. Is that about overinvestment? Energy consumption? Something else entirely?
In the first case, in a world of fiat money, there is no such thing as "overinvestment" as (provided you control a big chunk of the money supply) it's all just imaginary figures. A few zeros more or less doesn't make any real difference to them.
In the second case, we're dealing with a physical reality, so more plausible. But ENIAC probably used as much energy as a small town back in the day, to do the sort of calculations that your electric toaster would consider beneath him today.
Even if AI subscription costs went up thirty fold, it would still be profitable in my professional use (I ran the numbers last year).
But what really convinced me we're onto something here was going for an FSD ride. I already wrote about it in another comment and I described it as underwhelming. It's so natural that you forget that nobody is driving the car. A real car, in the real world, not a controlled scenario. I had to go and do it again and I still couldn't distinguish (a very good) human from machine, thus passing a real world Turing test.
You're still based in Prague? Honestly, it would be silly to croak it without having experienced it. I'd nip across the border and go for a test ride in Germany.
>Rather than try to paraphrase the great man
No, he is not a great man - I can't think of anything he has done that was truly great - his kernel isn't even free software for example.
If he was great, the advice he would give would be to never use such proprietary software and SaaSS - rather than stating that such is acceptable as a tool.
I wasn't trolling.
I simply do not agree with Linus's cult of personality (tell me, what is being concerned about avoiding to "paraphrase the great man" and referring to thousands of different programs as "Linus + x" = Linux?), but everything that has wide agreement must be correct and I must be wrong and I must be trolling - the icon is for you.
"I simply do not agree with Linus's cult of personality"
What cult?
The whole article is about a heated discussion between L and his "acolytes" who strongly disagree among each other. I have never seen anything but heated discussions between L and his acolytes about every detail of the projects. Nothing like a personality cult.
What I see in you is a person who does not want to give credit where credit is due. L is renowned to give credit to others in abundance himself.
And if Linux is not Free enough for you, you must have a twisted idea of free. The same twisted idea I have seen in Libertarians who consider freedom only as really Free when a man (sic) is free to enslave (and shoot) others.
What lack of clothes on the emperor?
If you've every looked at any website that is related to technology or any comment section, you'd realize about how many people worship the great savior Linus endlessly, but repeating "Linux" over and over, especially so in relation to software that has nothing to do with Linux - I've even seen a comment on el-reg from someone with so much Linux on the brain, that they wrote "Linux" when they meant to write "Netflix" - but there is no cult of personality, none at all.
Linus certainly doesn't give credit in abundance - he rather gladly accepts credit in abundance for things he didn't do - he has said that referring to thousands of different software packages as "Linux", while also never mentioning that Linux is only a secondary contribution is okay (after all, if he didn't, he wouldn't get all the credit and be able to spread his world view would he?).
When Linux is relevant, I credit it - but when it's not, I don't.
My idea of free is straight on freedom - it isn't so twisted that I see the proprietary software in Linux and that is part of Linux that isn't even gratis either and think it's free software.
I am not a Libertarian - freedom and slavery are clearly opposites.
It's quite odd you read about software being proprietary and not respecting the users freedom and thought about enslaving and shooting people.
"[L] has said that referring to thousands of different software packages as "Linux", while also never mentioning that Linux is only a secondary contribution is okay"
Linus did not come up with the name Linux and didn't promote it.
And it is pretty useless to argue naming a thing after a minor part of it, ie, pars pro toto. I assume even you have called a man a dick or an a**hole. Here L simply refuses to fight against the tides, or windmills.
>Linus did not come up with the name Linux and didn't promote it.
He realize the possibility of such name and decided that he would initially like to call his kernel Freax.
But others wanted to adopt the name, so he decided to promote it as well - but he didn't merely promote referring to Linux as Linux, he has promoted referring to other software as "Linux".
>it is pretty useless to argue naming a thing after a minor part of it >I assume even you have called a man a dick
I have never called a man a dick.
The reason to name something about a minor part of it is to insult it - clearly to refer to GNU/Linux, or GNU without Linux as "Linux" is an insult.
>Here L simply refuses to fight against the tides, or windmills.
If you refuse to fight against a tide that goes to the wrong place, you end up being swept by the tide to the wrong place.
You have to be really strong or skilled to be able to fight against the high torque of windmills - being able to do so is impressive in of itself.
I feel I need to add an example of my claim against certain circles in Libertarianism. The link below refers to part of the transcript of a podcast. Read or listen to all of it (and the comments) for the full picture.
Strong Debate vs. Walter Block on Voluntary Slavery (Matthew Sands of Nations of Sanity)
Block argues that Kinsella’s rejection of voluntary slavery creates a slippery slope, weakening other contracts, and cites historical libertarian support (e.g., Robert Nozick). Kinsella refutes this, noting Nozick’s simplistic contract view and arguing that inalienability is consistent with libertarian property rights, as bodies are not acquired like goods. Sands raises edge cases, such as indentured servitude or extreme debt, with Kinsella asserting that such arrangements are valid only if they respect revocable consent. Block defends enforceability, arguing that individuals should bear the consequences of their commitments.
Notice that Walter Block insists to the end that a slavery contract should be binding and enforceable even if the party changes their mind. They literally sold their body unconditionally in the same sense as they could have sold an apple or car.
I have seen such arguments discussed and defended myself in the more extremist anarcho-libertarian fora.
A very slight digression, but…
> The same twisted idea I have seen in Libertarians who consider freedom only as really Free when a man (sic) is free to enslave (and shoot) others.
Technically, it is not a twisted idea but an incomplete or a misinterpreted one.
It is correct that absolute freedom would allow those things. What they're missing is the precondition: absolute freedom can only exist in a state of perfection. In that eventual ¹ state, you will indeed be free to enslave others… but you won't.
A lot of the late 19th century and early 20th European third rate philosophers (Kierkegaard and Popper, I'm looking at you chaps) were too busy trying to make sense of their rapidly changing little corner of the world, looking through their narrow lenses, to appreciate that, and came up with a lot of nonsense that was subsequently amplified by the cottage industry of university philosophy departments. ²
¹ Eventual because that is the state that we are inexorably moving towards, according to the 2nd law of thermodynamics. The heat death of the universe, a state of perfect chaos when no more useful work can be extracted, is also, naturally, a state of perfect equilibrium. Just as absolute chaos is indistinguishable from perfect equilibrium, absolute freedom is indistinguishable from no freedom.
² Exceptions apply.
One doesn't need to be a regarded as "great man" to point out an error in judgement and as to why that was an error.
Even assuming I've done absolutely nothing, that would mean that I'm a much "greater man" than Linus (as doing nothing is clearly better than doing some good, but doing mostly evil that outweighs the good).
I've never developed proprietary software and have never made endless recommendations and encouragements to run proprietary software that takes the users freedom.
I haven't done countless wrong things out of greed, for an amount of money that is larger than you could reasonably spend (it is impossible to end up with a "net worth" of $50 million USD doing good - the least bad way to achieve that I can think of would be participating in the evil of a large lottery and somehow winning).
But let me guess, your worth is determined how much money you have and you are always great if you ended up as a billionaire, even if you killed thousands of innocent people on the way.
To tell you extensively what I have done would be to dox myself, but the software I have developed at least is free software and I have shared a lot of free software too.
Interesting that you haven't mentioned what you've done.
"you are always great if you ended up as a billionaire"
Strange, no one mentioned money here. Linus most certainly is no billionaire.
The man is called Great because he initiated and managed projects that ended up in improving the lives of many people. Linux already affects, and improves, the life of nearly all humans by way of Android and the network and cloud servers that run all communication and social networks.
If your work affects almost all humans in a positive way, and you are seen as the driving force of that work, it is acceptable to be called "Great".
>Linus most certainly is no billionaire.
He's a double digit millionaire and I'm sure he would happily be a billionaire at any cost to everyone else, if he could.
>initiated and managed projects that ended up in improving the lives of many people
False - many of the projects he initiated and managed are proprietary software, or only came after he did his best to make the lives of many people worse.
It is common to claim that he initiated an OS - but he didn't - he saw that GNU was almost ready and only needed a kernel, and so he wrote another kernel for GNU - which he released as proprietary software in 1991.
Later in 1992, I figure he wasn't seeming much success (as GNU wasn't interested in encouraging a new proprietary kernel by supporting it), and so he decide to relicense Linux to GPLv2-ambigious and GNU out of goodwill ported every last package to work with Linux, which resulted in a much more usable kernel (as you had an OS to properly go with it), but of course the whole thing was called "Linux", which happened to attract thousands of developers to work on Linux (kernels are fun and exciting, while the rest of the OS isn't as exciting).
Later, Linus proceeded to betray that goodwill by using his management skills to allow Linux to become proprietary software again via the inclusion of the first proprietary program (of many) in 1996.
There was even a policy years ago to migrate all the proprietary software in Linux into a separate git repository (that can't be anything but part of Linux), but that's still not done.
He betrayed the goodwill again by deciding that the the license would be GPLv2-only in 2000 and proceeding to only ever enforce the license against freedom.
As for git, Linus first made as many developers lives worse as possible, by encouraging and supporting the BitKeeper proprietary software and SaaSS for kernel development (rather than using existing free software version management or if that wasn't suitable, writing functionally better software), as the deal was that gratis access would be provided in exchange for the advertising.
Later every single developer that fell for it was burned, when their access was terminated (and only then did Linus decide to clean the mess up with git, which is convenient, but hard to use and I don't see how convenience alone particularly improves lives).
Of course he later betrayed everyone again, by encouraging the usage of github (his only complaint was that the commit metadata is not complete), but his words don't meet his actions, as he did accept a commit on github.
- As for Sparse, well that's a static analysis program for Linux that is convenient (as less bugs), but that doesn't seem to improve lives.
- As for subsurface, well that's a convenient program, but it's only useful for divers (which aren't that many people).
I really can't find much else he's done, other than his fork of a proprietary version of emacs.
>Linux already affects, and improves, the life of nearly all humans by way of Android
Android in fact is a net negative on the lives of almost all humans - it spies on them, restricts them and is proprietary software for almost all of its users,
As it became possible to do so, constant attempts have been made to force people to carry around a surveillance device, so they can be made available 24/7, which is total insanity and clearly does not improve the lives of anyone (but notifications every 5 minutes must be a good thing).
There is no technical reason why an Android phone that can make calls and the like, while not spying on you, but not a single model has been made that doesn't do that.
I am confident that the world would be a slightly better place without Android (too bad other proprietary mobile OS's would likely have become popular instead).
Regardless, google is currently working on a replacement to Linux for an OS similar to Android; https://en.wikipedia.org/wiki/Fuchsia_(operating_system)?useskin=monobook so the software that goes on future mobile devices can be 100% proprietary, instead of 99% proprietary - but it's not functionally good enough for general usage yet.
>the network and cloud servers that run all communication and social networks.
Social media has been a net negative on humanity in all cases - such companies have done numerous studies that concluded such and of course such companies did not stop, but CEO's and the like do ensure to not allow their children to have an account for a reason.
Current communication networks are a net negative on humanity due to all the spying, spamming and monopoly costs that are imposed on the invocation of proprietary SIP and RTP protocols required to communicate over such networks.
Linux alone is useless for networking - the networking stack needs to be configured what to do before it does anything and you need other daemons to handle protocols like BGP and even SLAAC and well you need GNU to setup the routing rules and run the daemons.
Other kernels and OS's do implement networking protocols as well, which are used, although less commonly, as those aren't as convenient.
There is no cloud - those are all someone else's computers - it doesn't make much difference what OS a server runs if it's used to spy on you.
"Later, Linus proceeded to betray that goodwill by using his management skills to allow Linux to become proprietary software again via the inclusion of the first proprietary program (of many) in 1996."
Please be more specific. If you mean you object to the GPL as being not free enough, say so. If not, what are you talking about?
"but of course the whole thing was called "Linux""
People called it Linux. Linus was not involved in that naming. "Pars pro toto" all over again.
>Please be more specific.
Since 1996 it has been popular to include proprietary software disguised as arrays of numbers in object form without source code in the Linux "sources".
Doing so in fact infringes the GPLv2, but Linus has done nothing about it, despite how distributing proprietary software as part of GPLv2 software, without an exception, automatically permanently terminates the license (it is possible, if you are the complete copyright holder, to pretend the license is GPLv2-only, but for that not to be the license, as the copyright holder cannot violate their own license).
There was a policy many years ago to un-array encode and move such proprietary software to the large collection of proprietary software that is part of Linux, which is of course is called "linux-firmware.git" (i.e. perihperial software for Linux) and is not called "firmware.git" (i.e. peripheral software for whatever kernel), but that still hasn't been done for several cases and there's still proprietary object code disguised as arrays of numbers in linux.git.
The only version of Linux I'm aware of that is actually under the GPLv2-only is; https://www.fsfla.org/ikiwiki/selibre/linux-libre/
>Linus was not involved in that naming.
Linus has been involved with encouraging such confusion, as it benefits him and ensures that people are only become aware of his ideas and worldview only.
"The only version of Linux I'm aware of that is actually under the GPLv2-only is; https://www.fsfla.org/ikiwiki/selibre/linux-libre/"
So there is a fully free Linux available for you.
In principle, Linus has no standing in enforcing copyright in Linux. That is for the copyright holders, ie, the actual developers. If they agree with inclusion of the blobs, the only thing Linus can do is blowing up the project. You now fault him for not doing that.
But your legal analysis of the reach of the GPL is questionable [1]. A copyright holder of software has to define the interface between their code and the rest of the outside world. There is no legally clear "natural" distinction. You can argue that any software running on a computer is part of any other software on that computer.
The Linux community decided to put the boundaries at a certain point. That point was free enough for most software developers and that worked out brilliantly. You now have a universal OS that you can run everywhere and anywhere, which you can change, compile, and distributed freely without having to ask permission from anyone. The FSF was unable to recruit enough software developers to create their own Really Free OS so they too are stuck with Linux.
[1] Anyhow, copyright law is a tangled mess and getting a definite decision can take decades of convoluted legal wranglings. SCO Group, Inc. v. IBM ran from 2003-2021 even though the initial claim was obviously ridiculous from the start.
>Linus has no standing in enforcing copyright in Linux. That is for the copyright holders, ie, the actual developers.
He was a developer, he wrote the original versions and a nontrivial amount of the code he wrote is still there, therefore he still holds a substantial copyright of Linux (how else would he have had the standing to say the license is GPLv2-only and not GPLv2-or-later in 2000 if he wasn't a substantial copyright holder exactly?).
>That is for the copyright holders, ie, the actual developers. If they agree with inclusion of the blobs, the only thing Linus can do is blowing up the project.
The only legal way for copyright holders can agree to the inclusion of proprietary software in GPLv2 software, is for there to be an exception made to the GPLv2 and for every last copyright holder to explicitly agree to such exception. No such exception exists.
Linux works perfectly fine without the proprietary software, removing it wouldn't "blow up" the project - it would merely be inconvenient (it's inconvenient if an actual free driver needs to be written, or if copyright enforcement against the copyright infringer to get the source code for all of the Linux driver, is needed for defective hardware to be supported).
>But your legal analysis of the reach of the GPL is questionable
I did not do some weird interpretation, I just read the license and saw what it has written; https://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html
"4. You may not copy, modify, sublicense, or distribute the Program except as expressly provided under this License. Any attempt otherwise to copy, modify, sublicense or distribute the Program is void, and will automatically terminate your rights under this License." (see section 3 of the requirements).
"This General Public License does not permit incorporating your program into proprietary programs."
>There is no legally clear "natural" distinction. You can argue that any software running on a computer is part of any other software on that computer.
A distinction is easy to make - if the software continues to operate if some other program is deleted and if when that other program is swapped out with some other program, there is no functionality loss, such programs are clearly separate works.
If you delete some proprietary peripheral software and the Linux driver no longer operates and there is no replacement for that proprietary software, clearly that proprietary software is part of the Linux driver.
Also, if you delete Linux and there would be no other kernel that you can continue to use the proprietary peripheral software with, clearly that proprietary software is part of the Linux driver.
>The Linux community decided to put the boundaries at a certain point.
The Linux developers have set insane boundaries - any convenient enough to use proprietary derivative work is allowed (permission to distribute the software is requested, but not required).
>You now have a universal OS that you can run everywhere and anywhere, which you can change, compile, and distributed freely without having to ask permission from anyone.
Linux is only a kernel - GNU could be changed, compiled, and distributed freely without having to ask permission from anyone before Linux even existed (to bad there was still the proprietary Unix kernel) and such proprietary Unix kernel being swapped out with the temporarily free and then proprietary again Linux didn't change that much.
You are in fact not allowed to distribute some of the proprietary software in linux.git - as there is no permission to do so, as there is no license.
For example see; https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/powerpc/platforms/8xx/micropatch.c - Motorola supplied some object code under no license - GPLv2 is not the license, as the SPDX was added later by a script.
If you do not have the complete corresponding source code under a free license, you cannot change, compile and distribute the software freely.
You have to beg permission to have the source code from each proprietary software developer and all of them have the power to say no due to a lack of license enforcement.
>The FSF was unable to recruit enough software developers to create their own Really Free OS so they too are stuck with Linux.
GNU GRUB is a fully free OS that does not use Linux, which is very popular, although it's functional use is booting other OS's.
GNU/Hurd is another fully free OS that works, it's just not convenient to run.
GNU/Linux-libre is also the FSF's own fully free OS, as all the proprietary software and proprietary software loading machinery is removed from Linux (still, most non-bugfix releases needs more proprietary software cleaned up, so a convenient replacement always free kernel would be good).
Hi Eric! Sorry I could not see you, because of that massive chip on your shoulder.
> Even assuming I've done absolutely nothing
Would that be a pretty safe assumption, would you say?
You do understand that when an article refers to "the great man", "his holiness", "Him", etc., it's all tongue in cheek and not to be taken seriously, do you? If it's not your kind of humour, just ignore it and move on.
> I've never developed proprietary software
Wow! Someone give this man a medal, quick! Maybe throw in a Nobel prize or two. :D
Seriously though, you seem to view yourself as a good man, but you spew a lot of hate. Good men don't hate.
Torvalds has achieved something really impressive, namely to keep together through all these years the massive, decentralised, heterogeneous project that is the Linux kernel (that he wrote the very first version of it is a small bonus, too). I am grateful for that. I don't expect him, or anyone else, to be the Messiah on jetskis.
Good luck with your life. As long as you are content with your lot, that's all that matters. Maybe there is no need to waste your time judging others.
If they're talking about AI writing the documentation for code changes I don't really see the issue, as long as the code author reviews what was written. As to the code itself I'd worry about unresolved Copyright issues but in the end both humans and AI can write bad code, that's why you have code reviews... If someone keeps submitting bad code then deal with that person appropriately...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
