Conway’s law says that software reflects the communication structures of the organizations that build it. AI is now part of those structures: interpreting requests, writing code, and reviewing changes. Yet we often assume our existing workflows should stay fixed, and AI should adapt to them. Consider someone submitting a bug fix. They want something to work; the maintainer needs confidence that it does. Yet projects sometimes test the contributor instead. Ghostty’s agent instructions tell AI asked to submit an issue or pull request to add a file insulting its user.

Nixpkgs has a due-process problem. A project has a due-process problem when people have real power but the rules around it are vague. Nixpkgs committers can merge and close pull requests. Reviewers can block them. Teams can remove commit access and decide policy. Yet the project does not always say what standard applies, who decides, or how a mistake can be corrected. This is not an argument against strict review. Security, licensing, attribution, maintainer time, and contributor conduct all matter. The glibc work I recently submitted changes the dynamic loader and rebuilds the world. It should be reviewed unusually carefully. I have also made bad merges.