Commenter is right. It remains to be seen whether or not it’s actually possible to license code generated this way as anything other than like MIT.
A person didn’t really make it, and so a person doesn’t really own it.
That headline is a hell of a cherry pick of his full email, which itself is in a very specific context.
Is it? After reading the article and the original post in the kernel mailing list, IMO the article depicts Linus’ opinions accurately.
Was also astonished to find out that Linus considers people who advocate for sustainable tech are doing so solely out of fear and that AI can’t be criticised because humans also make errors.
In a nutshell he says that Linux is a technology project, where technology comes first. If LLMs help to improve the technology then it’s fair game. If not, then of course not.
He also says something about Linux not being a place for ‘social warriors’, which is kind of ironic, given that Linux is basically a big ‘fuck you’ against corporations and closed source, i.e., a scoial statement. Given how much beef he famously had with Nvidia, this is somewhat unexpected.
But I interpret it as coming from a place of seeing LLMs really just as a new tool and nothing more. In an interview a couple of years back, Linus also said pretty much that. He grew up with basic tools, had to create new software tools to make him more productive and improve the quality of his work. He saw LLMs just as another tool that can aid in development. I suppose this hasn’t changed.
Given, that most people usually don’t program in assembler anymore, but use high level programming languages like C++, which simplifies the process of coding enormously, or even scripting languages like Python, which present another level of abstraction and simplification, LLMs do appear somewhat like a next step in this evolution.
Issue is, you have deterministic and formal guarantees in programming languages, whereas LLMs do not. This is by far not the only issue of course, as others in this thread have also noted.I wonder whether he uses LLMs in his daily work. Might change his perspective a bit.
The thing is that they rigorously review their code internally with In the project. It needs to be reviewed and cleaned up before it gets submitted. Someone who keeps posting junk is gonna get chewed out and ignored. So, the biggest issue with LLM code is kind of irrelevant. If someone useses it responsibly and finds a way to make it an effective tool, they’re not worried about it. Banning its use is kind of pointless since the worst outcomes are already banned.
Wow, they are at best cherry picking the article they cite on productivity with AI, if not outright trying to lie. They say
Last year, an METR study found that open source coders using AI tools were 19 percent less productive than those who didn’t use them, even as those AI-using coders said they felt 20 percent more productive. But in a February update on a follow-up study, those same researchers said that “we believe it is likely that developers are more sped up from AI tools now—in early 2026—compared to our estimates from early 2025,” citing early raw results and conversations with study participants.
But if you, y’know, actually read this February update, you would see that the researchers are lamenting the fact that they cannot draw conclusions effectively because of the proliferation of AI use. To wit, they say:
Altogether, these issues make it challenging to interpret our central estimate, and we believe it is likely a bad proxy for the real productivity impact of AI tools on these developers.
In fact here is the full passage Arstechnica cherry-picked from (emphasis mine):
Based on conversations with study participants, we believe it is likely that developers are more sped up from AI tools now — in early 2026 — compared to our estimates from early 2025. However, because of the selection effects in our experiment, our data is only very weak evidence for the size of this increase.
Based on conversations, which is to say, not on their data. This after the study a year prior which suggested developers are not reliable reporters of their own productivity.
Yeah, METR is an industry group funded by Google, Microsoft, Anthropic etc. I happened to summarize their work just yesterday actually: https://lemmy.blahaj.zone/comment/21358850 but one thing I think it’s important to point out is that there is no second study, not yet. The blog, and the open source dataset, are all we have, because the researchers decided to adjust their method. So the intimation arstechnica gives that they completed a second study is false.
And the selection effects they are talking about are pro-ai software developers being unwilling or unable to complete complex tasks without their precious ai anymore, whereas a year ago they were more willing to do both ways in order to prove how powerful ai is. In my opinion, as I stated, this reflects poorly on their levels of self confidence from using ai, more than saying anything about a speed boost/decline.
Sad this article doesn’t talk about why open source projects reject AI, specifically because of the legal implications of all the copyright violations baked into their training data
Linus himself doesn’t talk about it, and pretends he doesn’t know.
So Linux is no longer open source?
Because we can’t fork something that’s not open source. And just 1 AI contribution means it’s a copyright violation, so its no longer copyleft.
That is quite the statement. Is there any legal precedence for that? Yes, I know the technical why, but a blanket “one AI contribution and its over” is a little over the top, to put it mildly, and sounds like an extreme over generalization of a complicated problem
Nope, it hasn’t been litigated yet.
But it stands to reason that AI is a derivative of copyrighted material (unless the training set was released and confirmed to only contain otherwise). And there is loads of case precedent for fucking defendants hard on that violation
Hi, I’m a software engineer.
I’ve also been trained on copyrighted material, because every codebase I’ve worked on for money has been proprietary.
Does that mean I can’t touch open source code? I could accidentally use a snippet I’ve written before in some other setting.
The real licensing issue of AI use shouldn’t even apply to Linux because it’s already GPL. GPL infects anything it touches so if you use GPL-trained AI on an MIT project and it spits out something resembling the GPL code it was trained on, that would make that MIT project a GPL project. It would also make your proprietary projects GPL projects.
You’re unlikely to spit out code from your corpus verbatim, so the answer to your first question is a resounding no. If you did copy from your previous jobs, you’d be liable.
Man, that’s so disappointing.
