Linus Torvalds responded on the Linux kernel developers forum to critics of using AI tools in the project. «Linux is not one of those anti-AI projects,» he wrote, adding that anyone with problems with this «can fork. Or leave».

The discussion arose around Sashiko, a Linux Foundation tool that uses LLMs to automatically review code proposals. Torvalds said AI is «clearly useful,» even though a year ago it wasn’t so obvious, and «anyone who doubts it hasn’t used it.» He also acknowledged that AI can be «painful» — surfacing embarrassing bugs, for example — but the answer isn’t to ignore it. The project makes decisions based on technical merit, «not fear of new tools».

Why it matters to you. An authoritative voice shifts the debate from ideology to practical use. If you work with code, Torvalds’ position reinforces a principle that already guides this site: AI is a tool, and your job is to understand when you need it, not whether it’s permissible to use it. The agentic coding course starts from there.

In detail

What came before.

The Linux Foundation had launched Sashiko, a code review system that uses kernel-specific prompts and a dedicated protocol to automatically analyze patches. Sashiko can import change proposals from mailing lists or local Git repositories and works with different LLM providers. Some developers criticized the tool, and Roman Gushchin flagged a «strongly anti-LLM position» that risked undermining the goal of lightening the load on maintainers.

What Torvalds said.

Torvalds’ response was unambiguous. He acknowledged that an anti-LLM stance exists among some people involved, but clarified it doesn’t represent the Linux project as a whole. He wrote that he’s willing to «put his foot down as a top-level maintainer» on this point, and that AI is «clearly a useful tool».

He also anticipated objections: yes, AI can create extra work for maintainers, it can find bugs that make the people who wrote them feel stupid, but the answer isn’t to «bury your head in the sand.» AI isn’t perfect, but neither is human intelligence. No one is forced to use it, but Torvalds will «very loudly ignore» anyone who tries to discourage others from using it.

What it means in practice.

Torvalds’ position carries weight because Linux is critical global infrastructure, and his voice shifts the debate from «is it permissible to use AI?» to «how do we use it well?». For developers, this means AI tools in development workflows are no longer a novelty to justify, but a possibility to evaluate on technical merit.

Sashiko is documented on GitHub, and guidelines for kernel maintainers who want to use it are public. If you use coding agents, the story confirms a direction: open projects adopt AI tools where they work, with clear controls and boundaries, not on principle but for results.