MicroZig bans every LLM use, including spelling fixes and bug finding

MicroZig has added a strict seven-point LLM ban to its code of conduct. The project is an open-source toolbox for building embedded applications in the Zig language. It is maintained by the ZigEmbeddedGroup organisation and has 2,319 stars on GitHub. The post about the new policy had 51 points on Hacker News at the time of this item.

The policy sits under a new section of the project's code of conduct, titled "Strict No LLM / No AI Policy". The two named enforcement contacts in the document are Matthew Knight and Felix Queißner.

The seven rules

The section lists seven items. The verbatim text:

• "No LLM-generated content, whether it be code or prose."
• "No paraphrasing LLM-generated content."
• "No LLMs for editing, including fixing spelling or grammatical errors."
• "No LLMs for translation."
• "No LLMs for brainstorming and then sharing the results of that brainstorming."
• "No LLMs for finding bugs."
• "No talking about use of chatbot/LLM services."

The last point is unusual: it bans the discussion of LLM use in project spaces as well as the use itself. A spelling or grammar fix made with an LLM is also on the list. That rules out a quiet edit a reviewer could not spot from the diff.

What the maintainer told the Hacker News thread

Matthew Knight, the project's main maintainer, replied in the Hacker News discussion. He wrote that using LLMs on a project is valid for anyone who wants to. He added that those preferences and MicroZig's are not compatible. He said he had no hard feelings for anyone who forked the project to use LLMs. That reply is the clearest statement of the policy's intent on record. The code of conduct itself gives no reason for the ban.

How this compares to QGIS

A different open-source project has already banned AI-assisted code. QGIS, the open-source mapping project, replaced an earlier label-your-work rule with a strict ban. Its policy forbids AI, LLM or agent-assisted code in its main repository. The QGIS text covers code contributions only. MicroZig's text covers code, prose, editing, translation, brainstorming, bug finding, and talking about LLM use.

The question for maintainers

The two projects have picked different scopes. QGIS forbids AI code in the repository and leaves LLM use elsewhere to the contributor. MicroZig forbids any LLM use and any public discussion of LLM services. A maintainer thinking about their own project cannot read either policy as the standard. The two policies are already far apart. The MicroZig text sets the new far end.

Source

MicroZig Code of Conduct on GitHub.