Nav2 maintainer polls community on rejecting AI-generated PRs due to review burden

Nav2 maintainer Steve (”Your Friendly Neighborhood Navigator”) posted a community poll on May 19, 2026, asking how to handle the flood of AI-generated pull requests and issues in the open-source robotics navigation stack. He reports that over the past few months, contributions from AI agents or users copy-pasting AI outputs have become frequent. While some appear reasonable, many introduce subtle changes that fix problems no human would encounter, requiring hours of careful review. Steve has already started deprioritizing or closing such PRs, arguing that if the human contributor cannot be bothered to QA or write their own thoughts, maintainers should not waste time. The poll offers three options: treat AI contributions like human ones, accept until unreasonable, or outright reject if no human QA.

Community member chfritz voted for the last option and elaborated on the reasoning. He argues that open-source workflows designed for humans break with AI agents. The real question, he says, is what value AI-generated code provides when the same AI is available to everyone. Without human QA, the contribution has no unique value. He warns that contributors seeking resume-boosting “contributor” labels will find those meaningless if anyone can generate similar code via AI. Instead, chfritz suggests contributors focus on building AI automation tooling—like closed-loop engineering setups and test environments—rather than submitting functional code. This “going meta” approach would help evolve software engineering practices for the AI era.