Why Zig Bans LLM Contributions and Puts People Over Code

Zig language author Simon Willison reiterates that Zig prioritizes contributors over code, viewing each contributor as an investment of the core team’s review effort.

Recent concerns about AI‑generated code have led Zig to adopt one of the strictest anti‑LLM contribution policies among mainstream open‑source projects: LLM use is prohibited in pull requests, issues, and bug‑tracker comments, even for translation.

“Prohibit LLM use in pull requests.” “Prohibit LLM use in issues.” “Prohibit LLM use in bug‑tracker comments, including translation.”

While Zig encourages English, it allows posts in native languages and relies on community translation tools.

The most prominent Zig‑based project, the Bun JavaScript runtime, was acquired in December 2025 and uses AI‑assisted coding. Bun runs its own Zig fork; after adding parallel semantic analysis and multiple code‑generation units to the LLVM backend, it achieved a four‑fold performance improvement. However, the upstream Zig team does not plan to adopt these changes because of the strict anti‑LLM stance.

A core contributor explains that the policy, dubbed “contributor poker,” means betting on the contributor, not on the first PR’s content. The name reflects the idea that the project values the person’s long‑term potential over any single code contribution.

I call it “contributor poker” because, like the card game, you’re playing with people, not cards. In contributor poker you bet on the submitter, not on the content of their first PR.

Zig’s migration from GitHub to Codeberg at the end of 2025 reflects a broader operational choice tied to complaints about bugs, code bloat, and GitHub Copilot.

Overall, Zig’s policy aims to maximize the return on review effort by fostering contributor growth rather than merely filtering imperfect PRs.