'If you don't like AI, fork it or leave': Linus Torvalds pulls no punches

Linus Torvalds has returned to the debate surrounding the use of artificial intelligence in the development of the Linux kernel, taking a clear stance in favor of allowing developers to use tools based on language models (LLMs). Torvalds emphasized that Linux is not "anti-AI" and that technical decisions will continue to be made based on technological merit, not on suspicion of new solutions.

These statements come in response to a discussion that emerged on the official kernel mailing list, where the debate over the use of LLMs and tools like Sashiko AI for code review has intensified in recent months. The topic also arose after some recommendations published by the Software Freedom Conservancy regarding the use of AI in open-source projects.

Torvalds acknowledges that a part of the community harbors strong reservations about AI, but he asserts that he is willing to defend this position as the main maintainer of the project. According to the Finnish programmer, Linux will not become a project hostile to artificial intelligence, and those who do not share this approach can choose to create a fork of the kernel or abandon the project.

In his remarks, he emphasizes that LLMs should simply be regarded as a work tool, similar to other tools used daily by developers. In his view, the question about the actual utility of artificial intelligence is no longer a topic for discussion: those who continue to doubt it, Torvalds argues, likely have not practically experienced these technologies.

This does not mean that AI is without flaws. Torvalds admits that tools based on LLMs can increase the workload of maintainers and help identify bugs that can sometimes be uncomfortable to manage. However, the answer is not to reject the technology, but rather to improve how it is integrated into development workflows so that it acts as a support rather than a hindrance.

For this reason, the Linux project will not impose the use of artificial intelligence on developers, but at the same time, its founder states that he will openly ignore anyone who tries to prevent others from utilizing these tools. While acknowledging that LLMs are not perfect, Torvalds points out that human intelligence is also far from infallible and that mistakes are not a sufficient reason to dismiss a potentially useful technology.

In his statements, he reiterates that the Linux kernel has never been a project driven by ideological or social motivations. The primary objective remains the development of the best possible technology, and decisions will continue to be evaluated mainly based on their technical effectiveness.

The new statements mark an evolution from past declarations. In 2024, Torvalds had described the landscape of artificial intelligence as being composed 90% of marketing and only 10% of concrete results, suggesting that it would take about five years to understand its real value. Today, his judgment appears significantly more favorable, while maintaining a pragmatic approach.

The beginning of 2026 had already shown some signs of this change. Torvalds had revealed that he had started using AI tools to write code in a small personal project, specifying, however, that he did not directly use them in kernel development. Later, he also criticized some developers for submitting bug reports generated with the help of artificial intelligence just before a release candidate, noting how many of the proposed fixes were superfluous and unnecessarily increased the complexity of the kernel instead of addressing the issues at their root.

It is clear, however, that like AI, Torvalds' position has evolved as well, and he now considers LLMs to be mature and useful tools for software development, while still acknowledging that their use requires attention to avoid redundant code, ineffective fixes, and an increased maintenance burden on the kernel.