TL;DR
Linus Torvalds pushed back against efforts to make kernel documentation address 'AI slop', arguing that documentation should remain practical and neutral. He said contributors who rely on large language models are unlikely to label their patches as AI‑generated and warned against using docs to make a statement on the technology.
What happened
In an exchange sparked by a message from Oracle‑affiliated developer Lorenzo Stokes, Linus Torvalds criticized proposals to make kernel development documentation address low‑quality output from large language models. Torvalds said there is no point turning documentation into a battleground over what he called "AI slop," reasoning that developers who use LLMs would not mark their contributions as such. He emphasized that kernel documentation should be aimed at responsible contributors and should avoid taking either an alarmist or evangelistic stance on LLMs, preferring a neutral "just a tool" framing. The discussion follows reporting that a team is drafting clearer guidelines on LLM‑assisted contributions, and comes amid examples of developers using AI coding assistants in kernel work, such as a reported effort to modernize an old driver with a tool called Claude Code.
Why it matters
- Kernel documentation sets expectations for contributors; framing it as a policy battleground could affect community cohesion and contribution practices.
- Torvalds' stance signals resistance to making documentation a venue for enforcing opinions about LLM use, which could influence how maintainers and contributors handle AI tools.
- If AI‑assisted patches are not explicitly labeled, enforcement and traceability of tool use become difficult, complicating any potential rules.
- Ongoing work to draft guidelines means policy could evolve; the outcome may shape how widely LLMs are accepted in core development.
Key facts
- Linus Torvalds responded directly to comments from Lorenzo Stokes about LLMs and kernel documentation.
- Torvalds argued documentation should be for 'good actors' and should avoid making statements about AI being either catastrophic or revolutionary.
- He said contributors who produce 'AI slop' are unlikely to document their patches as AI‑generated.
- A team is reported to be preparing clearer guidelines on LLM‑assisted kernel contributions, according to Phoronix coverage of a related email thread.
- The discussion follows documented instances of developers using LLM coding assistants on kernel code, including a reported case using Claude Code to modernize a legacy driver.
- Torvalds has previously described much AI marketing as hype and has expressed conditional tolerance for casual 'vibe coding' in earlier comments.
- The article that reported the exchange also speculated about potential industry shifts such as rising model costs or an AI industry correction, though those are presented as opinion.
What to watch next
- Publication or release of the team’s LLM‑assistance guidelines for kernel contributions (reported as in progress by Phoronix).
- Whether kernel documentation will adopt any explicit wording about LLM use or remain neutral: not confirmed in the source.
- Any proposals for mechanisms to label or verify AI‑assisted patches and how maintainers would enforce them: not confirmed in the source.
Quick glossary
- Linux kernel: The core part of the Linux operating system that manages hardware, processes, and system resources.
- Large language model (LLM): A machine learning model trained on large amounts of text to generate human‑like language and code suggestions.
- Documentation: Written guidance and rules that describe development practices, coding standards, and contribution processes for a project.
- AI slop: Informal term for low‑quality or unreliable output produced by generative AI systems.
Reader FAQ
Did Linus Torvalds ban the use of LLMs in kernel development?
Not confirmed in the source.
Are official kernel guidelines about LLM use already in place?
A team is reported to be working on clearer guidelines for LLM‑assisted contributions, according to Phoronix; a final policy is not confirmed in the source.
Did Torvalds say contributors must label patches created with LLMs?
Torvalds argued people who use LLMs are unlikely to mark their patches as such and said documentation won’t solve that issue.
Are developers already using AI coding assistants on kernel code?
Yes; the article cites a reported example where a developer used Claude Code to update a decades‑old driver.
OSES Linus Torvalds: Stop making an issue out of AI slop in kernel docs – you're not changing anybody's mind 'Because the AI slop people aren't going to document their…
Sources
- Linus Torvalds: Stop making an issue out of AI slop in kernel docs – you're not changing anybody's mind
- Linus Torvalds: "The AI Slop Issue Is *NOT* Going To Be …
- Linus: Stop making issue of AI slop in kernel docs
- Linus Torvalds talks AI, Rust, & why Linux is the only thing …
Related posts
- Digital Red Queen: Adversarial Program Evolution in Core War with LLMs
- Dynamic Large Concept Models: Compressing Tokens into Adaptive Concepts
- Since 2023 Chinese AI models have trailed the US frontier by seven months