What is 'Contributor Poker,' which clearly illustrates why pull requests containing AI-generated code should be banned across the board?
The general-purpose programming language '
Contributor Poker and Zig's AI Ban | Loris Cro's Blog
https://kristoff.it/blog/contributor-poker-and-ai/
The Zig project's rationale for their firm anti-AI contribution policy
https://simonwillison.net/2026/Apr/30/zig-anti-ai/
Open-source development comes with various advantages and disadvantages, and when addressing problems, it's necessary to consider the benefits that can be gained. In other words, if the disadvantages outweigh the benefits, the value of addressing the problem diminishes.
Open source is fundamentally a model of 'providing valuable software free of charge,' and in return, some users become 'contributors' and submit pull requests . A pull request is a mechanism in which a contributor shares the code they have developed with the project team and proposes its integration (merging) into the main codebase. Open source software improves as many contributors submit useful code through pull requests.
Unfortunately, many pull requests don't actually benefit the project; instead, they often cause extra work and friction. In fact, it's often less troublesome for maintainers to implement the changes themselves than to collaborate with pull request contributors to make the code mergible.
Furthermore, successful open-source projects like Zig will eventually receive pull requests that exceed their processing capacity. Based on what we've discussed so far, it might seem logical to stop incomplete pull requests, but Zig encourages incorporating code submitted by new contributors as much as possible.
When asked why he supports new contributors, Kuro argues that it's not only the right approach for an open-source project, but also a wise choice. To explain this, he uses the concept of 'contributor poker.'
Contributing to an open-source project is not a one-time event, but rather an iterative process that is repeated many times. The majority of the value a contributor brings to a project lies not in the initial pull request, but in the subsequent iterative process.
Even if the initial pull request has more disadvantages than advantages, investing some effort to welcome a new contributor and then seeing many benefits in later processes will ultimately benefit the project as a whole. In other words, the project initially pays a cost (a stake) to the contributor, and it's okay as long as the project ultimately recoups more than its initial investment.
The reason Mr. Kuro likens this exchange to 'poker' is that, just as in actual poker where the important thing is the negotiation with the 'players' rather than the 'hand' that is created, in open source projects, the relationship with the 'contributors' is more important than the 'pull request' itself.
Kuro stated, 'Having a clear understanding of these dynamics has brought immense value to the Zig project over the years. Building a compiler toolchain from scratch was an enormous undertaking and would have been impossible without the tremendous support from our contributors.'
Based on this 'contributor poker' concept, it becomes clear why AI-driven pull requests should be banned across the board. Large open-source projects like Zig already suffer from an overwhelming number of pull requests that they can't handle, and AI-driven pull requests would only exacerbate this problem.
To make a truly beneficial pull request for a project, contributors need to be familiar with the codebase and problem domain, and their commitment to pursuing the best approach by thoroughly reviewing all changes made in the pull request is trusted by the core team. As part of the process to further enhance trustworthiness, contributors must remain responsible for the code for a period after it has been merged and address any issues that arise. These follow-up discussions are also important for building lasting relationships with core team engineers who have been involved in open-source projects for a long time.
Unfortunately, according to Kuro and his colleagues, most of the AI-based pull requests they've seen so far have been negative. They've encountered pull requests that are useless and full of hallucinations, pull requests that suddenly send in 10,000 lines of code on the first submission, and contributors who claim they aren't using AI but send in AI-generated answers that are full of errors.
Of course, not all AI is bad, and these contributors are simply misusing AI, but the vast majority of AI-based pull requests submitted to the Zig project were useless. Kuro explains, 'Theoretically, there may be legitimate contributors using large language models, but from a contributor poker perspective, it's simply irrational to bet on users of large language models when there are many other contributors who don't bring this risk factor.'
While some complain that it's impossible to tell if a contributor is using AI, Kuro points out that these people don't understand the point of the policy to ban AI-generated pull requests and are unaware of the existence of contributor poker. 'Contributor poker is a crucial element of our strategy, and it's best for the project to stand firm against anything that hinders the effective execution of this strategy,' he said.
Open-source developer Simon Willison argues that the contributor poker concept makes sense. 'If the majority of pull requests are written by large language models, why should project maintainers spend time reviewing and discussing pull requests instead of launching their own large language models to solve the same problems?' he points out.
Related Posts:
