Ddos 
It appears the burgeoning wave of AI-generated code has yet to penetrate the Linux kernel as deeply as it has other software domains. Nevertheless, the discourse regarding the coexistence with Large Language Model (LLM) suggestions has finally reached Linus Torvalds, who responded with his characteristic and unvarnished bluntness.
The catalyst for this exchange was a thread within the kernel maintainers’ mailing list. Lorenzo Stoakes, an Oracle developer, challenged the popular refrain that AI is “another tool,” suggesting such a stance implies the Linux kernel is somehow immune to the systemic risks posed by LLMs. Torvalds, however, took umbrage at this line of inquiry. In his rebuttal, he essentially called for a cessation of the hand-wringing over “AI slop”—the deluge of dubious, machine-generated patches and prose—urging that the kernel development documentation not be transmuted into a political manifesto.
His argument is fundamentally pragmatic: documentation is intended for good-faith contributors, whereas those intent on submitting substandard, automated code are unlikely to transparently label their contributions as AI-generated. Consequently, Torvalds dismissed the demand for formal documentation of the issue as futile. He expressed a desire for the documentation to remain untainted by either the doomsday alarmism of the “all is lost” faction or the exuberant idealism of those claiming AI will imminently revolutionize development. To him, the neutral designation of AI as “just a tool” is preferable, as it avoids ideological grandstanding.
While the correspondence does not explicitly define Torvalds’ ideal etiquette regarding LLM usage, he clearly delineated the boundaries of what should be avoided: attempting to rectify the influx of “trash patches” through pedantic rules or turning developer guidelines into a manifesto. In an era of ubiquitous AI assistants, this approach is starkly realistic; formal prohibitions are easily circumvented, and proving an author’s reliance on a model is often an exercise in futility.
The subject has already transitioned from theoretical debate to practical application. Community members are reportedly deliberating on more precise recommendations for contributions aided by LLMs. Tangible examples are surfacing as well; for instance, Wikimedia developer Dmitry Brant recently recounted utilizing Claude Code to modernize a legacy driver nearly a quarter-century old.
Ultimately, Torvalds’ position reflects neither panic nor euphoria, but rather a concerted effort to maintain the kernel’s rigorous developmental momentum. While substandard code remains a persistent threat, he contends that it must be addressed not through sloganeering in the documentation, but through the kernel’s established tenets: meticulous peer review, individual accountability, and uncompromising standards for patch quality.
Donate