Andrew Kelley, who created Zig and leads the Zig Software Foundation, says AI-generated code is not welcome in the project. During a recent JetBrains podcast appearance, he explained that pull requests written with chatbot assistance are often rejected because they add little value and take time away from the core team’s review process.
Kelley described code produced through large language models as consistently poor in quality. He noted that the Zig project already has hundreds of pending commits waiting for review, leaving a small group of maintainers to work through them under limited time. In his view, AI-generated patches often slow things down further because they tend to show little understanding of the project’s structure or design.
Kelley compared the review process to “contributor poker,” where code reviews help the Zig team identify developers with long-term potential. In some cases, promising contributors are invited to join the project more closely. He argued that a rise in machine-generated submissions makes that process harder by flooding reviews with low-value contributions, weakening signals about who is genuinely learning and improving.
This is basically Zig putting its policy into words after drawing the line for a while. The project has made it clear that code influenced by LLM-assisted editing, brainstorming, or debugging is not welcome. Kelley’s comments also echo a bigger argument from AI skeptics in software development, who believe relying on AI too much means skipping the hands-on learning that open-source work is supposed to teach.
Zig was introduced by Andrew Kelley in 2016 as a free, open-source systems programming language inspired by ideas first established in C by Dennis Ritchie. The project is designed to serve as a lightweight replacement for C and C++, with an emphasis on reducing complexity so developers can focus more on debugging applications. Because the Zig team operates with limited maintainer capacity, review time is treated as a limited and valuable resource.
Kelley’s rejection of AI-assisted code adds to an ongoing discussion about the role of AI in software development. Some studies suggest that “vibe coding” may slow experienced programmers in certain situations, even as technology leaders continue promoting agentic AI tools. Concerns over costly errors have also led some developers to introduce hidden instructions into projects that are intended to interfere with or misdirect AI-assisted coding workflows.
Maybe you would like other interesting articles?