Linus Torvalds on AI Coding in Linux: 'Fork It, or Just Walk Away' — and How 6 Open Source Projects Compare

On July 17, 2026, Linux creator Linus Torvalds told critics of AI-assisted coding in the kernel to "fork it, or just walk away" — one of the most permissive public stances any major open source project has taken on AI-generated contributions.

The comment, posted to the Linux kernel mailing list, carries outsized weight because the kernel is among the most conservative, review-intensive codebases in the world, underpinning most of the planet's servers and cloud infrastructure. It lands in the same year AI coding startups reached tens of billions of dollars in valuation and several other major open source projects moved the opposite direction, banning AI-generated contributions outright.

Linus Torvalds AI Coding: What He Actually Said, and Why

Torvalds wrote that "Linux is not one of those anti-AI projects," adding that he would "absolutely put my foot down as the top-level maintainer" on the point. Critics who object, he said, "can do the open-source thing and fork it. Or just walk away." He framed the kernel's openness to AI tools as a decision made "based on technical merit. Not fear of new tools" — reported by Ars Technica on July 16, 2026, with the underlying mailing-list post dated July 17.

The dispute that prompted the comment centered on Sashiko, a Linux Foundation AI code-review system that imports patches from mailing lists or Git repositories and reviews them against kernel-specific prompts using various LLM providers. Developer Roman Gushchin had objected to what he described as a broadly "anti-LLM" attitude among some kernel contributors, arguing it undermined Sashiko's goal of easing maintainers' review workload. Torvalds' position is a personal stance rather than a change to formal kernel policy, and plenty of individual maintainers remain skeptical of AI-generated code regardless of what he says publicly.

Is This a Reversal? Torvalds Also Pushed Back on AI-Generated Bug Reports in May 2026

Two months before "fork it, or just walk away," Torvalds was criticizing AI's effect on the kernel, not endorsing it. In his Linux 7.1-rc4 release notes on May 17, 2026, he said a flood of AI-generated bug reports had made the kernel's security mailing list "effectively unmanageable," with maintainers spending their time forwarding duplicate reports or replying that an issue had already been fixed weeks earlier, since different people were running the same AI tools against the same code and surfacing the same problems.

His fix wasn't a ban — it was a documentation update, led by kernel contributor Willy Tarreau, telling AI-tool users to "ask your tool to propose a fix and test it before reporting the issue" rather than leaving verification to maintainers. Read together, the two comments are consistent rather than contradictory: Torvalds objects to AI-generated noise that offloads work onto maintainers, not to AI-assisted contribution itself — a distinction that lines up with the Linux kernel's actual rule, which requires a human to add real value and take responsibility for any AI-assisted patch, tagged "Assisted-by:".

How Big Has the 2026 AI Coding Funding Boom Gotten?

Torvalds' comment lands in the middle of a funding surge for the companies selling AI coding tools. Cognition, maker of the AI software engineer Devin, raised more than $2 billion in a Series E announced September 9, 2026 at a $48 billion valuation, nearly double the $26 billion mark it set four months earlier; Andreessen Horowitz and Accel led, joined by Founders Fund, General Catalyst, Avenir, and Nvidia. The company's run-rate revenue grew from $492 million in May 2026 to almost $900 million by the round's close. The quarter's biggest number came from an exit, not a round: Cursor maker Anysphere was acquired outright, with SpaceX closing an all-stock, $60 billion deal for the company on August 14, 2026 — the largest startup acquisition on record. And Poolside was reported in late 2025 to be in talks for Nvidia to invest up to $1 billion as part of a roughly $2 billion round, valuing the AI-coding startup at $12 billion pre-money — about four times its $3 billion valuation from October 2024. Further down the market, AI app-builder Emergent raised a $130 million Series C in July 2026 at a $1.5 billion valuation, up five-fold from the $300 million mark its January 2026 Series B set, on roughly $120 million in annualized revenue and more than 200,000 paying customers.

The adoption numbers behind that funding are real but contested. An analysis of 4.2 million developers between November 2025 and February 2026 put AI-authored code at 26.9% of production code, up from 22% the prior quarter — the most rigorously measured figure available. Broader developer surveys that count code merely suggested or lightly edited by AI report higher shares, around 41-42%, using a looser definition that inflates the comparison. On the distribution side, Microsoft disclosed more than 4.7 million paid GitHub Copilot subscribers on its January 28, 2026 earnings call, up 75% year over year — the best-documented adoption number among the major AI coding tools, since Microsoft and GitHub don't separately disclose monthly or daily active usage.

How Do Other Major Open Source Projects Handle AI-Generated Code?

The Linux kernel's permissiveness is not the industry norm. Several other major open source projects moved toward outright bans on AI-generated contributions in the same year, mostly citing copyright exposure, reviewer burden, and code-quality risk rather than opposition to the technology itself.

How we counted: 6 major open source projects' published AI-contribution policies as reported through October 2026; each project name links its primary news source.

GCC's development team rejected AI-generated contributions beyond roughly 15 lines of code on July 29, 2026, citing the GNU project's own threshold for a "legally significant contribution" and copyright exposure above that size; smaller AI-assisted changes are still accepted if tagged "Assisted-by:". OpenJDK's governing board took a harder line in April 2026, banning any contribution containing content generated "in part or in full" by an LLM, citing reviewer burden, safety, and IP uncertainty from ongoing litigation over AI-generated output. Codeberg's members voted 358 to 144, with 14 abstentions, to ban "mostly AI-generated" repositories platform-wide, in a vote that closed July 22, 2026.

LLVM took a middle path in January 2026, requiring a human to review all LLM-generated code before it reaches maintainers, banning fully autonomous agent submissions, and disclosing substantial AI-generated content with an "Assisted-by" commit trailer — the closest policy on this list to the Linux kernel's own disclosure-plus-permission approach. GraalVM, built by Oracle alongside OpenJDK under the same contributor agreement, takes the opposite stance from its sibling project and permits AI-generated submissions outright, without a published rationale.

What the headline misses

Torvalds' "fork it, or just walk away" is a personal position from the kernel's top-level maintainer, not a binding kernel-wide policy vote like Codeberg's or a governing-board rule like OpenJDK's — it can be overruled by subsystem maintainers in practice, and several have stayed openly skeptical of AI-generated patches regardless of what Torvalds says publicly. The 26.9% AI-authored-code figure is also a general measurement of production software, not a kernel-specific number — nobody has published a comparable study measuring what share of actual merged Linux kernel commits are AI-assisted, so the connection between "a lot of code everywhere is AI-written" and "the kernel is meaningfully different because of it" is inference, not something this data directly shows.