Andrew Kelley didn’t mince words. On a recent JetBrains podcast the president of the Zig programming language called AI-assisted contributions “invariably garbage.” The remark landed like a slap. It wasn’t hyperbole. It reflected years of frustration with pull requests that added nothing but extra work.
Zig has long positioned itself as a better C. Simpler. Safer in key ways. Explicit about memory. The project runs under a nonprofit Zig Software Foundation with a serious budget and real users in production. Ghostty. TigerBeetle. Even parts of Bun before tensions arose. Yet success brought its own headaches. More interest meant more submissions. Many of them now come tainted by large language models.
The policy stands crystal clear. No LLMs for issues. No LLMs for pull requests. No LLMs for comments on the bug tracker, including translation. The Zig code of conduct spells it out. English is encouraged but not required. Contributors can post in their native language. The ban covers generated content, paraphrased material, edited drafts, brainstormed ideas, and debugged fixes. Everything touched by an LLM stays out.
Why such severity? Numbers tell part of the story. At the time of the podcast recording Zig carried 200 open pull requests. Reviewers remain scarce. Each low-quality submission steals minutes or hours from people who could fix real bugs or mentor serious contributors. “People are sending us contributions that have no value whatsoever,” Kelley said. “They have negative value, because they take review time away from the team…. We’ve wasted everybody’s time.”
Those words echo across the industry. Big Tech pushes hard for AI-written code. Companies tout percentages of output generated by models. Zig answers differently. It operates as a 501(c)(3). Efficiency at all costs isn’t the goal. Mentorship is. “We’re all trying to get better at programming,” Kelley explained. “People who are sending AI pull requests, those people are not helping this goal.”
The real bet sits on people, not single patches.
Loris Cro, vice president of community at the Zig Software Foundation, laid out the thinking months earlier. In an April 29, 2026 post on his blog he described “contributor poker.” The concept is straightforward. Open source works as an iterated game. The first pull request functions like an opening bet. Real value appears in later rounds once the contributor learns the codebase, earns trust, and ships consistently. You play the person, not the cards.
“Contributing to an open source project is an iterated game and the majority of the value that a contributor can bring to a project lies in the later iterations,” Cro wrote in his April 29 post. “You play the person, not the cards. In contributor poker, you bet on the contributor, not on the contents of their first PR.”
AI breaks that loop. It floods the queue with drive-by efforts. Some won’t compile. Others pass surface checks but hide hallucinations. A few look clean yet reveal LLM influence in follow-up discussion. Cro cataloged the damage. Worthless drive-by PRs full of hallucinations. Insane 10,000-line first-time submissions. Sneaky contributors who claim no LLM use but regurgitate its mistakes. Each one consumes reviewer attention without building long-term human capital.
The foundation sees this clearly. Users bet on Zig to deliver high-quality tools. The project in turn invests in an environment where competent engineers grow. “For us the ability to provide contributors with an engaging ecosystem where they can improve their systems thinking and interact with other competent, trusted and prolific engineers is a critical aspect of our business model,” Cro added. AI contributions undermine exactly that.
But the decision carries trade-offs. Some observers argue a blanket ban feels extreme. What about experienced developers who use AI as a thoughtful assistant? The policy draws no distinctions. Any touch of an LLM disqualifies the work. That strictness has already forced forks. Reports indicate the drama with Anthropic-backed Bun led to its own separate path after the ban took hold.
Recent coverage reinforces the tension. Business Insider detailed Kelley’s podcast remarks and the nonprofit structure on May 29, 2026. Slashdot amplified the “invariably garbage” line the next day, sparking fresh debate. No major reversal appears on the horizon. The project left GitHub partly over concerns about Microsoft’s AI focus and CI reliability. It now emphasizes independence.
And the results? Zig continues to attract serious adopters. Its zero-dependency cross-compilation toolchain stands out. Memory management stays explicit. No hidden garbage collection. Developers who value control and clarity keep choosing it. The ban signals that clarity extends to how the project accepts help.
Critics call it anti-progress. Supporters see preservation of craft. Either way the line is drawn. Zig won’t optimize for volume of contributions if those contributions erode the quality of its people pipeline. Other projects watch closely. Some may adopt similar rules. Others will lean harder into AI assistance and accept the noise.
Kelley and Cro bet that human relationships compound. A single strong contributor who returns for years outweighs dozens of one-off AI patches. Time will test that wager. For now the policy holds firm. No AI. No exceptions. Just code written by people who intend to stick around and get better.
The choice reveals something larger about open source in the AI age. Scale brings efficiency demands. Yet some teams refuse to trade their core mission for speed. Zig picked mentorship over volume. The 200 open requests may shrink slower than corporate roadmaps demand. But the ones that close will come from real collaboration. That, the project insists, matters more.
WebProNews is an iEntry Publication