Linux kernel development has always been a hive of activity, but in recent months a new player has emerged: artificial intelligence. Linus Torvalds, the creator of Linux, has expressed his frustration with the excessive use of AI tools to detect and fix bugs, especially when those bugs are trivial and clog up the development process. However, Torvalds himself has used AI on more than one occasion, and his latest experience demonstrates that, when used properly, it can be an ally.
The paradox is clear: while criticizing the spam of AI-generated patches, he himself uses AI to solve complex problems. The key lies in human judgment and oversight , something Torvalds strongly advocates. In this article, we review his complaints, his successful experiment with the Xe driver, and what this means for the future of the kernel.
Torvalds' frustration with kernel AI
In a message to kernel contributors, Torvalds expressed his frustration with the number of trivial fixes being submitted, many of them AI-generated. Kernel 7.1, in its fifth Release Candidate, is accumulating patches he considers irrelevant and time-consuming. "I'm not convinced it's worth the effort for an RC5," he wrote, referring to the fact that most are common driver fixes that could wait for the next development phase.
The problem isn't just the quantity, but the quality. There have been cases of multiple people fixing the same bug and reporting it separately, leading to confusion and overlap. Torvalds warned that he will be "more inflexible" and will begin rejecting pointless pull requests, especially given the Linux kernel's history of dealing with vulnerabilities and the role of AI. "Trivial fixes may be insignificant, but 'low probability' is not the same as 'zero probability,'" he concluded.
A success story with AI: debugging the Xe controller
But Torvalds' relationship with AI isn't all rejection. He recently encountered a bug in the Xe driver, which manages Intel's GPUs. The flaw involved incorrect rounding of the CCS memory address , causing random on-screen corruption. It took Torvalds more than 24 debugging patches and 18 kernel boots to isolate the cause.
In that process, he used an AI assistant, similar to how Claude Code is used as a programming assistant . The AI ​​even claimed the problem was unsolvable , but Torvalds persisted, and the tool continued generating debug code. Finally, the error was fixed by changing a misapplied function. Torvalds acknowledged the AI's work, though he joked, "I suspect those things were trained by people who might not be as stubborn as me."
The necessary balance
Torvalds is not against AI, but against its irresponsible use. He argues that AI tools should help maintainers, not cause them problems . "The solution is to make sure those LLMs help maintainers instead of just causing them problems," he stated. He also made it clear that he is not forcing anyone to use AI, but that he will ignore those who try to prevent others from using it.
This dual role of AI in the kernel reflects a broader reality: technology is a tool, not a substitute for human judgment . While some contributors create noise with automated patches, others, like Torvalds, use it to solve complex problems. The difference lies in oversight and knowing when to say "no."
Ultimately, Linus Torvalds's stance on AI is pragmatic: he uses it when it adds value, but he won't tolerate it becoming a source of noise. The Linux kernel needs collaborators who understand the context, not machines that generate patches without any discernment . The community will continue to debate this balance, but for now, Torvalds has made it clear that AI has a place, as long as it's used wisely.
