Won't anyone think of the environment or small communities? AI is a nice-to-have, not an essential. We need to stop developing technology without thinking long term about the world we live in. We can't use the "oh we didn't know" excuse anymore.

Reviewing AI code is harder than reviewing human code. I'll judge humans on what they do, and LLMs on what they are.

Global solar generation has increased by 50% in the last 2 years, meanwhile a token is something like 900 times less energy consuming than it was a year ago?

Something comes first, almost free energy and deeply embedded ai usage or total environmental collapse, capitalism deems that we will find out which in due course.

I generally try to avoid using AI for my hobby projects, but I think there is still a niche for it in working out issues that take a disproportionate amount of time to track down manually relative to the scale of the fix. Things like a subtle typo somewhere or missing some hyperspecific nuance of whatever stack I'm working with. In those particular cases I would say it actually enhances my own creative ability insofar as I don't need to spend hours of effort banging my head against a wall until I eventually figure out what went wrong, and I can put that time and mental energy into the stuff that's actually interesting to me.

But yes, in general I do agree it's kind of useless otherwise for creative endeavors or passion projects. It's great for helping with code being written solely for a paycheck, but for personal projects it largely defeats the purpose of working on them in the first place. I think the trouble is that in general AI does make a lot of things easier and I have a lot of anxiety about getting into the habit of overusing it in that way, so I try to keep it as a sort of last resort.

It's really weird when people with a kink try to push that kink on everyone else and they act like everyone else is wrong for building their entire persona around the kink. We get it, you really love hand crafting your code, line by line character by character, but I don't need that to enjoy coding. I like producing useful things and then working on ways to make them even more useful. I have hand crafted enough widgets and thingamabobs that I don't need to personally fondle another one into existence. I don't need to be an expert at yet another language that will be obsolete well before I retire. When I was a young lad I could get off a really good wank to the bra section of the Woolworths catalog, but not so much anymore. That doesn't mean I am burnt out or that I don't like a good wank anymore, but frankly I would prefer someone with a pretty face and soft hands to yank on my dick for me. Have you never worked with or trained junior programmers before? I guarantee you that the nonsense a junior can produce is much worse than the crazed hallucinations of some AI.

There are a lot of reasons to hate AI, like the destruction of the dev experience pipeline... but shitting on people who manage to find ways to use AI is not really helping your crusade.

NetBSD hasn't fallen into AI psychosis but it still has some slop in it.
Year of the GNU/Hurd desktop!

I love how people with a fraction of his development experience are shitting on him now because his opinion on an aspect of development doesn't align with theirs.

it's crack for burnt out programmers who don't enjoy the process of creating software anymore.

That explains perfectly my use of it for my job (where we're expected to use it anyway). I'd been burnt out at my company for a long time, and LLMs give me the ability to at least still fulfill my feature quotas long after my desire to give a shit about the codebase and software had gone. I am a techmeth addict now. It kinda sucks, but I feel hopeless about tech as a career nowadays and just the future in general, so I don't even care anymore.

No. For an easy example, does fascism deserve tolerance?

Generated code was trully bad. These days? Not so much anymore...

Better tools, more acceptance.

If someone demands a piece of software that completely erases all available data when a certain button is pressed, no one is told about it, someone presses it by accident and deletes a whole bunch of data, that's not a software development error. That's a management failure. The software did exactly what it was supposed to do.

And LLMs wouldn't make anything better in that regard. An LLM is built to do as the user wishes, with much, much less pushback than a human. So whereas a human getting the demand to write that software might say, "what in the actual fuck are you smoking and can I have some of it", an LLM will typically say, "that's a very bold decision!" In fact, in the case of MCAS, the issue of safety was raised by the engineers many times, and was flat out ignored by management. An LLM is much less likely to make a fuss and much more likely to just do it.

First, define fascism.

People will throw the fascism label around for all kinds of things, sometimes justifiably, but more often to over-dramaticize and get attention.

Overly broad over used labels can't deserve zero tolerance. You might say that murder deserves zero tolerance, but what about murder of a person who is in a position where they are about to murder a large number of other people? Do you enforce your "zero tolerance" by doing the thing you label as "zero tolerance"?

You might say that genocide should be zero tolerance, but again - we're getting into over-broad over-used poorly defined words. When does killing of members of a certain culture or group become genocide? Clearly, it's killing more than one - but if the group is a small society of maybe 100 members, is it genocide to kill 50 of them? When the 50 of them are about to murder 1000 others, even though the 1000 are members of a society of millions? What's the threshold of "about to" when it's acceptable to take away the "murder" label and kill them anyway?

No, I stand beside the two sides rolling my eyes every single time another "Look! Linus is criticising AI use in the kernel! He's fighting the devil alongside us!!" / "Look, Linus is using AI! He got converted to the one true faith!!" article gets published. Which is every other day at the moment, because it's really important to keep the main sport event going so the idiots don't lose interest or –even worse– start to actually think for themselves.

Citations needed.

I have not heard of a 900x drop. If that was true, we wouldn't be expanding data centres like we are.

2024 I'd use "the tools" to generate snippets that I would copy-paste and massage into my larger projects. Those snippets tended to be 2 to 6 lines long, and they'd be usefully correct about 80% of the time. I laughed at people who were "vibe coding" with those tools, even in early 2025. Like Linux in 1994, I'd call it "interesting, not ready for prime-time."

By mid 2025, I saw the potential for improvement - but the better the code output got, the louder and more irrational the online reactions against it got - up until the past month or so it was like a guaranteed brigade of negativity.

You're forgetting that usage can still go up while token cost goes down.

Or rather, efficiency and usage can increase at the same time.

Are you also sayogn there's over a 900x increase in usage?

Again, citations needed.

The reason I dont believe you, is software efficenies are WAY cheaper than a data centre. If they could devlope existing infrastructure to hold the increase of usage like that, then they wouldn't expand hardware.

You can find token energy efficiency information studies online, I was smashed when I pulled that figure out of my arse but the numbers are similarly huge.

They're already orders of magnitude more energy efficient than we predicted just a few years ago and both the software and the hardware continue to accelerate that.

Very true. LLMs are not likely to perform any more admirably in the Milgram experiments than people do. Although, with a bit of appropriate pre-prompting, you might regulate LLM output to better comply with regulations. I think people are even more easily intimidated to "do what they are told" both directly and implied.

If you ask LLMs about cybersecurity implementation decisions, and keep the prompts straightforward like: "is this adequately secure for the target environments?" they are coming out on the highly conservative / secure side of things lately. Of course you can prompt and reprompt them with ever more detailed expectations until they agree with whatever you say, but out of the gate a clean context will generally bring you every "best practice" and then some and point out a big list of things your legacy system isn't doing.

Oh I know its getting more efficient, just no where close to 900x.
I'm not gonna research for you. You provided the claim, not me.

You're just repeating the same point I've just addressed, so I'll explain in more detail.

Data centres expanding is perfectly compatible with huge efficiency gains per token. If the cost of inference falls, companies can afford to run vastly more inference, serve more users, use larger models and embed AI into more products.

It’s the same reason more fuel-efficient cars didn’t eliminate demand for roads or petrol. Efficiency lowers the cost of using something, which can increase total consumption.

So “if software got more efficient, they wouldn’t need more data centres” is basically backwards. The relevant question is whether total AI demand is growing faster than efficiency is improving.

Yeah I can tell you dont want to look into it, that comes across very clear 😂

He's been like this for a while now (at least a few months), I admit it affected my choice of OS for a new machine.

I use FreeBSD btw.

So your really making the claim that there's over a 900x demand for AI in the last year?

I'm calling bullshit.

Still more than your trust me bro sources.

Comparing AI to traditional tools is a bad faith argument and it's a bummer to see Linus fall for it.

What if the kink that you enjoy is actively killing the planet and ruining people's lives?

To be clear, I'm saying that those are side effects. Not that it's the kink itself.

And the person you replied to is saying it is not just another tool because tools are deterministic.

No.

?

When you lock onto something really hard in a discussion like this it becomes very obvious you're faltering and dont really have much to contribute.

More what?

You're not the best conversationalist but its good that you're working on it.

Affirmanti incumbit probatio.

Ive already said i was wasted when I pulled that figure out of my ass.

And the other guy has already acknowledged that the underlying sentiment that is has become a lot more efficient is true.

If they need more than that, its there for them whenever they want.

I agree, tbh. If your submissions are bad, they're bad. If your submissions are good, they're good.

I don't know... tools made by companies who can't wait to get in bed with the Pentagon to kill brown people around the world would seem to be toxic af.

NetBSD has one slop commit, and they have apparently since removed the code it brought in. Besides, I need some level of mainstream software support for uni, and pkgsrc is half-decent for that.

GNU Mach (a component of Hurd), meanwhile, have been using Claude for code reviews (https://lists.gnu.org/archive/html/bug-hurd/2026-03/msg00100.html, https://lists.gnu.org/archive/html/bug-hurd/2026-02/pdffAOvuZusjj.pdf); and the GNU Project itself is still on the fence (GCC, for example, has 57 slop commits and counting).

It's your stance that's wrong. I usually refrain from putting random bullshit on the Internet when smashed, but that's just me.

My stance? What about stance?

Thats good for you, I usually dont preach to people about how to interact with the Internet, but thats just me.

Ypu claimed up above a 900x increase in token efficiency, and claim that demand is outpacing efficiencies.

So yes you are.

I'm calling bullshit on demand outpacing efficiencies in general.

Based on what rationale?

You made the claim it is.

Its based on every company I know pulling back on token spend because it keeps getting more expensive.

Based on the fact the the growth they anticipate is everyone will use it everywhere. And thats not going to be true any time soon.

Based on the fact queries cost more and more tokens on the new models.

Elon, I keep telling you, the answer is no.

The original point was that expanding data centres doesn’t disprove a big increase in efficiency per token. Usage, advsnced models/tokens per query and infrastructure can all grow at the same time as the energy required for an individual token falls. And thats an ongoing process.

Companies you know cutting token spend, you thinking ubiquitous AI adoption is unrealistic, and newer models using more tokens per query are all arguments about future/total demand. None of them actually tell us whether a token is dramatically cheaper or more energy efficient to produce than it was a year ago. Realistically thats a learning curve problem, if someone is throwing a basic question at a model far more complex than it needs they are creating the inefficiency and that needs to be made more obvious/streamlined.

I've not claimed demand had increased 900x. You inferred that from the existence of new data centres, which was the faulty bit of reasoning I was responding to in the first place.

You can't really control the use of tools though, that is the problem. AI detection doesn't work and leads to a lot of false positives. The only thing you can do is check if the product is good.

AI is horrible for the world but it is impossible to ban it for coding, as you can easily submit AI generated code that looks human.

He is not comparing the effects that tools have but that you can't control if someone uses them or not.

The only way I see for controlling the way AI is used, is if the AI companies put restrictions on it.

If you ban AI, the vibe coders will just lie about the origin of their code.

So this:

between idiots believing AI is the devil and idiots believing AI is their new lord and savior

wasn't mud-slinging, in your opinion? Calling someone an idiot isn't an attack?

Strongly agree with this, except for "everything is political", which I firmly disagree with.

Linus seems to be tunnel vision in focusing on how good it is as a tool. That, relatively speaking, I couldn't care less about. That tunnel vision, however, is existentially dangerous. He is, in some ways, acting as bad as the likes of Thiel.

AI is a tool

If you mean "only a tool", this is wrong. See for example https://www.youtube.com/watch?v=yZUuKzDQSsI (ignore the stupid title)

He's right about the tool aspect of it, but, and I cannot emphasise this enough, it is not just a tool. Though we all bow (or salivate) before our fake materialist gods, and cast aside the beauty and relationships in the world, things are more than just their function.

You can’t really control the use of tools though

You start on a false premise. Immense swathes of the legal system are designed to control the use of tools, are they not? Legality aside, most people aren't sociopaths, and can handle being responsible humans. You don't even mention that, in many cases, you can just ask.

He is not comparing the effects that tools have but that you can’t control if someone uses them or not.

I don't agree that he's talking about verifiability or honesty.

I don't see the argument that if some bad actors are going to lie in PRs, that we should just let the floodgates open.

You're correct but still completely missing the point.

The only way to implement anti-LLM rules is to detect AI-assisted code. That, in and of itself, is a problem because it's impossible to detect good code that was AI assisted. And you don't need rules against slop because it's bad code.

If I'm following it correctly, Linus is advocating for tightening submission requirements before it gets to human review, which directly targets the harms FOSS projects are having under the deluge of slop: it raises the bar so that the slop needs to be good enough to warrant a review, which filters out must of the noise, leaving more of the signal.

Anti-LLM policies are like having closed-book take-home exams: you're left wildly guessing what's cheating (or not). Instead, make the exam open book, but expect better answers.

Linus also says AI is a useful tool, and he is an expert, so we must accept that he feels that way. I don't know if he's commented on the environmental/social/political consequences of AI hyperscalers, but that's a separate argument.

The tech is here. It's impossible to detect. It's foolish to think you can ban it's use

And it can be run locally, so there's no stopping it, even if there was political will to do so. (And with a reasonable economic and ecological cost, no less—ignoring IP/political consequences here, since that's a whole other topic.)