Redazione RHC : 26 July 2025 19:16
Sasha Levin, a longtime Linux kernel developer at NVIDIA and previously at Google and Microsoft, has proposed adding formal rules for using AI assistants in development to the kernel documentation. He also proposed a standardized configuration for tools like Claude and other AI assistants, which are already actively used to create kernel patches.
Levin has published an RFC (request for comments) proposing adding a special configuration file to the kernel repository that can be read by AI assistants. He also presented an initial set of rules describing how to properly use AI in kernel development, including requirements for formatting and commit attribution.
The proposed patch consists of two main components. The first adds a single configuration file referenced by tools such as Claude, GitHub Copilot, Cursor, Codeium, Continue, Windsurf, and Aider. This should ensure that AI behaves consistently throughout the kernel code. The second patch includes the rules themselves: adherence to Linux programming style, respect for established development processes, proper attribution in AI use, and license compliance.
The examples in the document demonstrate exactly how AI participation in commits should be formalized, including the use of the “Co-developed by” tag, which directly indicates co-creation of the virtual assistant. This approach, in the author’s opinion, will ensure transparency and fairness in the acceptance of patches into the main branch.
It’s not yet known how Linus Torvalds will react, but the discussion promises to be heated. The topic of AI’s role in creating critical system software has been a hot topic for some time, and now the community has the opportunity to develop clear and transparent rules.
Redazione