Linus Torvalds told AI holdouts to fork off or leave [2026]

The most influential living programmer just drew a line, and it’s not the one the anti-AI crowd wanted. If you refuse to work on the Linux kernel because contributors use AI tools, Linus Torvalds has an answer for you: build your own fork, or walk.

He said it in his usual register, which is to say without a single soft edge. And it landed hard on social media.

The claim that stings

Torvalds’ real provocation isn’t that AI is fine. It’s his read on the people who reject it outright. Anyone still asking whether these tools are useful, he argues, hasn’t actually used them.

That’s a sharper cut than it looks. For a lot of developers, blanket rejection is a moral position. Torvalds reframes it as inexperience.

He treats “is it useful?” as a settled question and moves on to the ones he thinks still matter, like how the economics of AI eventually shake out. The usefulness debate, in his telling, is over.

Why this voice carries

You can dismiss a lot of AI takes. This one is harder to wave off.

Torvalds built the Linux kernel in 1991 as a student, a personal project that grew into one of the most consequential pieces of software ever written. He’s still the top maintainer, with the final say on what lands in the core code.

Then in 2005 he built Git, the version control system that developers and companies everywhere now run on. Two technologies, both foundational, both his. That’s the resume behind the blunt talk.

What AI actually does in a codebase

The reason the programming conversation looks different from the one raging in creative fields is what these tools do day to day. They don’t just spit out a few lines of code.

They read large codebases. They surface bugs. They generate tests, explain dense code and help review changes before they ship.

They’re not infallible, and they don’t replace the judgment and experience of a strong developer. But the progress over the past several months is steep enough that pretending they’re worthless doesn’t hold up.

Torvalds is candid about the cost, too. He calls AI “a somewhat painful tool,” both for the maintainers’ workload and because, in his words, it “keeps finding annoying and embarrassing bugs.” That’s not a pitch. That’s someone who’s felt the friction.

The context the takes usually skip

None of this erases the objections. In creative work especially, the concerns are real and earned: copyright, the data these models train on, the threat to people’s jobs.

Where it curdles is the absolutism. For a lot of critics, any trace of AI in a project, a video or a game means the whole thing gets tossed, no questions asked about how or how much it was used.

Torvalds is drawing a boundary against exactly that reflex inside his own project. Here’s his full statement, translated, the one that set off the argument online.

“I understand that some people really dislike AI, but this is an issue on which, as the top maintainer of the project, I’m willing to absolutely put my foot down.

Linux is not one of those projects that take an anti-AI stance, and if someone has a problem with that, they can do what’s done in open source and create their own fork.

Or just leave.

AI is a tool, just like the other tools we use. And it’s clearly useful.

This may not have been so ‘clear’ even just a year ago, but today it’s no longer in question.

There are other questions around AI, like how its economic model will ultimately take shape, but ‘is it useful?’ is no longer one of them. Anyone who doubts it obviously hasn’t really used it.

Yes, it can also be a somewhat painful tool, both in terms of the maintainers’ workload and in the sense that it ‘keeps finding annoying and embarrassing bugs.’

But the solution isn’t to bury your head in the sand and sing as loud as you can ‘la la la, I can’t hear you,’ as some seem to be doing.

The solution is to make sure these LLM tools help maintainers, rather than just causing them problems. There’s no doubt about that.

We’re not forcing anyone to use them, but I will very demonstratively ignore those who try to prevent other people from using them.

And no, AI isn’t perfect. But, Christ, anyone who points out AI’s problems would do well to look in the mirror at the same time and point at themselves too.

Because natural intelligence isn’t always so great either.

The Linux kernel project has always been about technology and will continue to be about technology.

Of course, the social dimension of working on open source is important and is often a very strong motivation for participating in the project, but ultimately that’s a side benefit and not the purpose of the project.

This is NOT some kind of ‘social activism’ project, it never was and it never will be.

In the kernel community we work on open source because it leads to better technology, not for religious or ideological reasons.

That’s why we make our decisions primarily based on technical merit and not based on fear of new tools.”

Where this leaves you

Notice what Torvalds isn’t doing. He’s not mandating AI, and he’s explicit about it: no one is forced to touch these tools.

What he won’t tolerate is one contributor blocking another from using them. That’s the group he says he’ll ignore, and pointedly.

Zoom out and the timeline is worth sitting with. Late 2022 is when ChatGPT pushed generative AI into the hands of millions, letting people write text, make images, dig up information, edit content and generate code just by describing what they wanted in plain language. Barely three years later, the person who steers the software most of the internet runs on is telling the skeptics their objection is out of date.

If you still think AI has no place near serious code, Torvalds has already picked your exit. Fork the kernel and prove him wrong, or find something else to work on.