The Zig Software Foundation has a message for developers who want to use AI to churn out pull requests: don't bother. In a detailed blog post published this week, the organization pulled back the curtain on its controversial policy banning AI-generated code contributions—and dropped some genuinely useful frameworks for thinking about open-source sustainability in the process.

The Contributor Poker Framework

The heart of Zig's reasoning comes down to what they call "contributor poker." It's a concept that sounds clever until you realize it's just honest accounting. When you accept an external contribution, you're not just getting code—you're initiating a relationship. And like any investment, the real returns come later, after you've spent time onboarding someone who initially costs more energy than they'd save you. "You play the person, not the cards," the post explains. "In contributor poker, you bet on the contributor, not on the contents of their first PR." This framing reframes what looks like generosity—helping new contributors get patches merged—as pure strategy. The Zig project has benefited enormously from this approach. Ryan Liptak contributed Windows resource script (.rc) file compilation support. Frank Denis built out substantial portions of std.crypto. These weren't one-off wins; they were relationships that compounded over time.

Why AI Breaks the Model

Here's where it gets interesting for anyone building infrastructure around LLM-generated code. The contributor poker model assumes a human being on the other end—someone who will stick around, take responsibility when their code breaks production six months later, and develop deeper domain expertise with each interaction. AI contributions break all of these assumptions. According to Zig's experience, LLM-based pull requests have been "mostly negative"—flooding maintainers with hallucinated code that doesn't compile, 10,000-line first-time PRs from strangers, and polished-looking submissions where follow-up discussions immediately reveal the author has no real understanding of what they submitted.

The Trust Problem Nobody Talks About

There's a subtler issue at play here. When a human contributor submits code, they're accountable for it. If their change introduces a regression, there's someone to ping on Slack, someone who understands the broader context and can help course-correct. This iterated relationship is where the real value of open-source collaboration lives. With AI-assisted contributions, that accountability evaporates. The person who submitted the PR might not understand the code they just pasted in. When things go sideways three versions later, there's no established relationship to fall back on—just a stranger with an LLM subscription and plausible-sounding explanations. "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," Zig's post notes. AI users are simply a worse bet in this framework—higher variance, lower accountability, no compounding relationship value.

The Hard Truth About Open Source Economics

Zig doesn't pretend their approach is ideologically pure. They've acknowledged that good PRs have gone un-reviewed for extended periods, potentially losing valuable contributors to frustration. The project has grown past the point where they can invest in every newcomer. AI has made this worse by flooding the queue with noise. "The people who remarked on how it's impossible to know if a contribution comes from an LLM or not have completely missed the point of this policy," the post pushes back directly. It's not about detection—it's about risk modeling. Even if you can't always tell, rational actors should recognize that AI users present different (worse) expected value profiles than organic contributors.

Key Takeaways

  • Contributor poker frames open-source contribution as relationship investing, not code acquisition
  • AI contributions destroy the accountability and iteration assumptions that make external contributions valuable
  • The policy isn't about catching LLM users—it's about optimizing for long-term trust relationships
  • Even with a ban in place, Zig acknowledges review bottlenecks remain an unsolved problem

The Bottom Line

Most projects don't have the luxury of turning away contributors, but Zig's explicit framework forces a useful question: what are you actually optimizing for? If you're building infrastructure where long-term code stewardship matters, AI-generated PRs might cost you more than they're worth—no policy required.