Linus Torvalds has integrated an LLM-generated solution into a Linux kernel development branch. The patch, aimed at addressing a specific performance issue, resulted in a 23x speed increase in a synthetic benchmark. Torvalds clarified that real-world performance gains will be significantly lower, as the synthetic test does not reflect typical system loads.
The kernel development team has updated its policy regarding AI-generated code. Contributors are no longer required to disclose the use of AI tools in their submissions. Torvalds advocates for treating LLMs as standard development utilities, comparable to compilers or text editors, provided that developers maintain a clear understanding of the code they produce.
Torvalds previously expressed support for AI-driven code review tools, noting their ability to identify basic programming errors more effectively than human reviewers. He has taken a firm stance against restricting AI usage within the kernel development process, emphasizing the importance of code quality over the origin of the submission.
The integration of AI has also impacted security management. The Linux security mailing list faced significant challenges earlier this year due to the volume of AI-detected vulnerability patterns. Currently, the team addresses approximately 2,000 CVEs per release, leading to decisions to drop support for older hardware drivers that frequently trigger security reports.
