When Linus Torvalds decides to intervene in a public discussion, the free software community holds its breath. The creator of the Linux kernel, known for his aversion to soft consensus and his often brutal frankness, He has launched a new speech filled with violent criticism…
This time directed against attempts to convert the documentation kernel technique on a battlefield ideological opposition to content generated by Artificial Intelligence, referred to as "AI garbage".
The controversy arose from an exchange of emails with Lorenzo Stokes, an Oracle-affiliated developer, who maintained a cautious and critical stance versus Grand Language Models (LLMs). In response to an initiative to establish guidelines on bot-assisted contributions, Stokes suggested that treating LLMs as just another tool was a naive stance.
La Torvalds' reply was immediate and sharp.:
Technical documentation is not the place for moral statements or political manifestos.
Pragmatism versus symbolic activism
The central argument of Torvalds does not deny the existence of low-quality AI-generated codeRather, he questions the usefulness of regulating it through warnings in the documentation. In his view, developers who submit "AI garbage" aren't going to label their patches as such, regardless of what the rules dictate. Therefore, filling the documentation with ethical warnings is, in his words, a pointless pose that only serves to soothe the conscience of the rule-maker, without adding any real value to the quality of the code.
Torvalds insists that the documentation is written for honest developers. Trying to solve the problem of code quality through moral clauses is a mistake in approach. For the father of Linux, the kernel must remain neutral Given the technology used to create it, there are already enough divided opinions between those who see AI as the end of the world and those who see it as the ultimate solution; kernel documentation should not take sides, but rather limit itself to establishing technical standards of verifiability and accountability.
AI is already here: The AUTOSEL case
The irony of this debate moral Artificial intelligence has been operating for years. silently within the Linux ecosystem. The kernel stabilization process, a titanic task which requires deciding which patches to apply to the stable branches, sHe has long relied on AUTOSEL.. This tool, designed to automatically select relevant patches Based on statistical models, it has recently evolved to incorporate advanced AI techniques, including embeddings that allow it to analyze the semantic meaning of code and comments.
Sasha Levin, senior engineer at Nvidia, illustrated this reality during the Open Source Summit 2025. Levin explained how he used AI to write a complete routine. for git-resolve in kernel 6.16, limiting himself to reviewing and testing the resulting codeThis example underscores the prevailing view in high-level engineering: the tool accelerates error selection and correction, allowing humans to focus on validation. Even veteran developers, such as Dmitry Brant of Wikimedia, have publicly documented the use of assistants like Claude Code to modernize decades-old controllers.
The analogy of Autotune and human responsibility
To explain his philosophy on AI in creative development, Torvalds uses a musical analogy: Autotune. For him, Artificial intelligence is to programming what Autotune is to music. If used to allow untalented people to sing, the result is mediocre and artificial. However, in the hands of a skilled producer, it's a powerful tool for polishing and refining. True talent doesn't need AI for the creative process, but it can use it for organizational tasks, verification, and removing barriers to entry.
Ultimately, Torvalds' stance is a call for individual responsibilityTools, whether LLMs or compilers, amplify both the user's skills and their errors. What matters in a critical project like the Linux kernel is not the origin of the code (whether written by a sleepless human or a neural network), but rather that it is accurate, maintainable, and correct. Turning the documentation into a pamphlet against technology distracts from the only thing that truly matters: technical quality and rigorous human review.