antirez

InfoWorld

It's Foss

Linus Torvalds posted to the Linux kernel mailing list this week and made his position on AI tooling about as clear as he ever makes anything: Linux is not an anti-AI project, and if you want to argue otherwise, he will loudly ignore you. “Fork it. Or just walk away.”
The immediate trigger was a debate over Sashiko, an agentic patch review tool that uses LLMs to flag potential bugs for human maintainers. The controversy wasn’t really about AI in the abstract — it was about whether Sashiko should automatically email patch authors with AI-generated feedback before any human had reviewed it. That’s the line Torvalds drew: AI tools should help maintainers handle their workload, not add to it by generating noise that contributors then have to respond to.
His actual position is more measured than the headlines suggest. He’s not telling anyone to use AI. He’s saying he won’t penalize those who do, and he won’t let ideological opposition to AI tools drive technical decisions. “The kernel project is technology-driven, not ideologically driven” is the gist. Decisions get made on technical merit. That’s it.
What makes this interesting is how far he’s traveled to get here. In late 2023, Torvalds was dismissing AI as “90 percent marketing” and joking about being replaced. By late 2025 he’d warmed up enough to try an AI IDE on a personal project. In April 2026, Linux 7.0 shipped with an official AI Coding Assistants policy — licensing rules, responsibility tags, human accountability requirements. By July 2026, he’s telling critics to fork the kernel. That’s a real arc.
He’s still not naive about the costs. He acknowledged that AI-generated submissions increase maintainer workload and that hallucinated bug reports are a real problem — he’d previously complained about AI-generated vulnerability reports flooding the security mailing list. His answer isn’t to ban the tools; it’s to make the tools better at helping rather than burdening the people who have to review everything.
The backlash to the backlash is where it gets a little muddled. Some developers read Torvalds’ statement as a green light for vibe-coding kernel patches, which it clearly isn’t. Omar Cornut put it well: “stating ‘AI is just a tool’ doesn’t give anyone a mandate nor an encouragement to naively vibeslop the kernel.” The Linux project already requires human review, proper tagging of AI-assisted work, and the ability to defend every line you submit. None of that changed.
The compiler analogy Torvalds floated is worth sitting with: developers don’t credit compilers for writing their code, so the framing of “AI wrote this” is doing something weird. The accountability still sits with the contributor. That’s the actual policy, and it’s pretty reasonable.
The broader Linux ecosystem is less unified. Ubuntu has been embracing AI tooling. The Fedora community voted down an “AI Developer Desktop” proposal. Red Hat and open-source security firm nolabs have both echoed Torvalds’ line that human accountability matters more than whether AI helped write the code — but that consensus doesn’t extend everywhere.
What Torvalds actually said is defensible. What some people heard — that the kernel is now open to AI-generated slop as long as you tag it — is not what he said. The gap between those two readings is where most of the current discourse lives.

Trends

Linus Torvalds responded sharply on the Linux kernel mailing list to developers pushing back against AI tooling, stating Linux is not an anti-AI project. The controversy was sparked by Sashiko, an AI code review tool proposed to automatically email patch authors with AI-generated feedback. Torvalds' stance: AI tools are just tools, nobody is forced to use them but nobody can ban them either. He compared LLM-assisted code to compiler output, rejecting the framing that AI 'writes' code. The ideological debate he largely shut down; the practical concern — automated AI reviews adding noise and burden to already-stretched maintainers — remains an open and legitimate question.

Linus Torvalds to AI critics: use it, fork it, or leave

This discourse is fascinating. Very valid points from both sides.
Thread starts here: <a href="https://lore.kernel.org/linux-media/20260530103004.6fe2ffa7@foz.lan/" target="_blank" rel="noopener nofollow">https://lore.kernel.org/linux-media/20260530103004.6fe2ffa7@foz.lan/</a>

AI is just a new compiler. If your compiler outputs a bug, it’s still your fault. Developers remain 100% responsible for their code.