Linus Torvalds Now Uses AI to Write and Review Linux Kernel Code
Linux's creator has begun relying on AI tools for kernel development, marking a major shift in how the world's most important open-source project is maintained.
Linus Torvalds, the creator and principal maintainer of the Linux kernel, is now using AI to help develop and review code for the project, according to ZDNet. The disclosure marks the first time Torvalds has publicly acknowledged integrating AI into his own kernel workflow, a notable reversal for a developer who has historically been skeptical of new development fashions.
The Linux kernel is the foundation of most servers, Android devices, and cloud infrastructure, with thousands of contributors submitting patches every release cycle. Torvalds has spent more than three decades personally reviewing and merging those patches, and his adoption of AI assistance suggests the tools have become useful even at the highest level of systems programming. ZDNet reported the move as a turning point in his relationship with AI, though specific tools and the scope of their use were not detailed.
Torvalds has not been an eager adopter of every trend in software development. He has criticized overly complex abstractions and resisted changes he considered unnecessary, so his use of AI carries weight beyond his own workflow. When the person who sets the tone for kernel development changes tools, maintainers and contributors across the ecosystem tend to take notice, even if adoption is not mandated.
The disclosure lands as AI coding assistants become standard in many development environments, from autocomplete tools to patch-review bots. In the kernel world, where a single bad commit can affect millions of machines, the bar for accepting machine-generated or machine-reviewed code is higher than in most projects. How Torvalds uses AI, and whether he treats its output differently from human submissions, will shape expectations for the kernel's roughly 1,000 active contributors.
Open-source governance already faces questions about AI-generated contributions, including licensing, attribution, and the risk of plausible but subtly wrong patches. Torvalds's move does not resolve those questions, but it pushes them from theoretical debate into the daily routine of the kernel's most important reviewer. The next few kernel release cycles should show whether AI assistance changes patch volume, review times, or the kinds of bugs that slip through.
What to watch next is whether Torvalds describes his AI setup in more detail and whether other top-tier maintainers follow. If AI-assisted review becomes normal inside the kernel, it will be the strongest signal yet that these tools can operate at the most demanding end of software engineering, not just in application code. The open question is where Torvalds draws the line between assistance and delegation.