“Torvalds said that the time has come to make it harder for new people to contribute code to the kernel. People like Kroah-Hartman have spent years trying to teach people how to contribute; it is time to stop doing that, he said.”
https://lwn.net/Articles/1099594/
It is difficult to summarize exactly how Linus got to this point. Most of the discussion at this years Summit appears to have been related to AI assisted code and how the maintainers are dealing with the increase in code submissions to LKML. Many of the maintainers expressed varying levels of frustration with AI assisted drive-by code submissions. Linus’ train of thought is that they make it intentionally more difficult for humans to contribute. A later suggestion of his was to create an AGENTS. md that makes generating kernel code expensive in compute and token time.
Is this the Linus who leaked the whole Linux Kernel source code on GitHub? Wow!
Why is he not in jail yet?
he is, i mean obviously no one would willingly move to the US.
Linus, buddy, you’re the one that opened the floodgates by welcoming LLM code with open hands…
But he didn’t “welcome LLM code with open hands” really - he took a relatively neutral stance and said it needed to pass the same standards that human code does.
he didn’t. it’s more like him attempting to evaluate if it’s any useful, and news websites suddenly sending out misleading headlines that linus is now ‘vibecoding’ or smth
That’s not remotely what happened. Does it make you feel better to turn the truth into a cartoon with no nuance and obvious villains? I don’t suspect you’ll be argued into being reasonable, but people reading the comments should know that you are spreading lies.
Linus used an LLM to experiment in a branch. There is absolutely no evidence that he actually submitted any kernel code written by an LLM. He shouldn’t have to hide his experiments with technology because some people are mindless drones who will become enraged and lie about what he’s actually done.
There is absolutely no evidence that he actually submitted any kernel code written by an LLM.
The person you replied to never made such a claim.
Does it make you feel better to turn the truth into a cartoon with no nuance and obvious villains?
Okay, true. But is there a better interpretation of “opened the floodgates by welcoming LLM code with open hands” than “look, he said he used it recently, so he demonstrated that he is fine with slop in the kernel”?
Trying to understand what the hell it would mean to say he “welcomed it with open hands” would mean and I assumed they incorrectly took him recently saying he experimented with llms to mean he pushed code written by it. Apparently like half the linux people here thought/assumed that.
deleted by creator
This is something OSS projects everywhere are dealing with, no matter what their previous policies were nor what their personal beliefs about LLMs are.
I can’t believe people here are acting like Linus and Linux “brought this on themselves.” Vibecoding spammers do not care about a project’s guidelines, nor their maintainers’ workload; they’re coming either way. And friction + stricter contribution standards are how other projects are dealing with this, too.
I get it, quality vs quantity
suggestion of his was to create an AGENTS. md that makes generating kernel code expensive in compute and token time.
Why would one use the agents file if it makes it expensive?
I guess it states testing, simulating, double checking the submission. As should be done and would be done by humans. Instead of one shot guess in the wilds.
The quote sounds like a penalty. And your guess sounds like the minimum work that has to be done anyway. Completely different framing.
“Penalty”? Fucking hell, that’s exactly how my AI development is done. My personal tools repo that isn’t shared with anyone has almost 2k tests that run every time there is a change. Also the documentation is updated, and multiple adversarial agents are run to challenge every decision from a security standpoint and a simplicity standpoint. And there’s another guard I’m forgetting.
If you’re going to code with AI, that’s the least you can do.
“My solution is to throw so many digital monkeys at the problem that they will write the works of Shakespeare.”
Under capitalism, this token burn makes the shareholders climax at the expense of all else - datacenters are oligarch crack.
Shakespeare isn’t the point — it works. It has saved countless hours of production support investigations, and has made it possible to answer questions and catch problems no one has been able to do in years because it would be prohibitively time consuming. I’m answering questions in a couple of minutes that would otherwise have just been a finger pointed at another team.
Why can’t this guest log in? WAF problem because another team has a piss-poor page design? I would never have even thought of that. Forwards log entries to networking team for resolution.
I assume drive-by coders wouldn’t know any better.
Yeah, this is it. It’d be more of a spam prefilter.
I’ve been sitting on a few patches from several years ago. I never submitted them because I don’t want to go jump through the hoops again.
I get it though. Even before LLMs, the internet was already going down a path of slopification, with github repos being spammed with CV padder pull requests.
have not rtfa
maybe a simple referral system?
its going to make devs with connections a mega hot commodity $$$
edit: a future where sys admins get $$$ for know how to patch a kernel? more likely a future where we are stuck with shitty vendor vibe coded forked kernels with -1 chance of up streaming 🫠
after decades on the internet, i ultimately agree with enforcing a skill floor when necessary. things were better when you had to kinda know a little bit about what you were doing to manage to get online.
Well well well, if it isn’t the effect everyone with a working brain predicted would happen by loosening up to AI.
It’s also somewhat ironic that the “solution” to the bazaar is to become more like a cathedral - https://en.wikipedia.org/wiki/The_Cathedral_and_the_Bazaar
The kernel is already shit for humans to contribute to🤷
Let’s be honest, mailing lists are garbage and just increase the barrier to contribution. They lack nearly everything useful for reviewing code and giving feedback.
They are wonderful for LLMs though. Somebody unwilling to learn the arcane crap needed to contribute just pushes it out to an LLM. It’s easier for them to contribute than humans.
deleted by creator
linux is what it is in spite of linus not because of him.
Yes, the man by far most responsible for a semblance of OS choice today is the real problem here because he tries to think objectively and avoid premature overreactions. Maybe you’d prefer an OS kernel that runs on outrage. Like every PR could be voted on in social media and the most upvoted gets merged and the least upvoted get posted on a wall of shame where the contributors are mercilessly mocked? Who needs decades of experience when you’ve skimmed some headlines that either confirm or deny your own worldview.
