Linus Torvalds Backs AI in Linux: 'Fork It or Walk Away'
The man who created the Linux operating system kernel has drawn a clear line in the sand. AI coding tools are welcome. Critics who disagree can start their own project.
Key points
- Linus Torvalds, creator of the Linux kernel, publicly backed AI coding tools in a post to the Linux kernel mailing list this week.
- The debate centres on Sashiko, an AI-powered code-review tool that its creators say catches 53.6 percent of bugs later fixed by human developers.
- Sashiko also produces false positives, meaning it flags problems that do not exist, at a rate its own maintainers put "well within the 20% range".
- Torvalds told critics they are free to fork the project, meaning copy the code and run their own separate version under a different name.
Linus Torvalds did not mince words. Writing on the Linux kernel mailing list, the public forum where developers debate changes to the operating system that powers most of the world's servers, Torvalds said this week that "Linux is not one of those anti-AI projects." Anyone unhappy about that, he added, is free to "do the open-source thing and fork it. Or just walk away."
The spark was Sashiko. It is an "agentic" code-review system, meaning software that works through multi-step tasks on its own without a human steering every move, built specifically to scrutinise Linux kernel code. Its creators say that in tests it independently found 53.6 percent of the bugs that human coders would later fix.
That sounds impressive. There is a catch.
Sashiko also sends false-positive reports, alerts about bugs that turn out not to exist. Those land in the inboxes of the volunteer maintainers who keep Linux running, costing them time they cannot recover. Sashiko's own team estimates that false-positive rate sits "well within the 20% range," which means roughly one in five alerts could be noise.
For ordinary computer users, none of this changes anything today. Linux runs quietly inside Android phones, cloud services, routers, and smart televisions. The debate is about how the people who write its code work, not about what the software does for end users.
Should ordinary people care about this argument?
Yes, a little, because Linux underpins so much of the internet. If AI tools help maintainers catch bugs faster, the software billions of people rely on could become more reliable over time. The trade-off is maintainer burnout: volunteer engineers already stretched thin cannot afford to chase phantom bugs all day. How well tools like Sashiko are tuned will matter.
Torvalds has form here. He built Linux in 1991 as a student project and has steered it ever since with a famously direct style. His word carries weight, as first reported by Ars Technica AI. When he says AI tools are staying, they are staying.
The real question now is calibration. A tool that catches half the bugs but wastes a fifth of a maintainer's attention is not yet a net win. Sashiko's developers will need to push that false-positive rate down before the broader Linux community warms to it fully. Torvalds has opened the door. The tool still has to earn its place.
