PRs NOT Welcome: How Top AI Open Source Projects Are Managing Thousands of Contributors

Top AI open source projects are changing how they handle contributions by shutting off external pull requests. Projects like Vercel’s AI SDK, Astro, Flue, and tldraw are instead using custom software factories and specialized AI agents to triage issues, reproduce bugs, write code, and review fixes.

This shift is driven by a massive influx of open issues and pull requests, many of which are AI-generated code that maintainers find unreliable or time-consuming to review. By trusting their own optimized agent setups, maintainers have regained control over massive backlogs. For example, Vercel reported that its system authors up to thirty-five percent of merged pull requests and closes up to eighty percent of issues, while Astro successfully transitioned from an unmanageable backlog to weekly prioritized issue handling.

Under this new model, projects like Flue and tldraw automatically close incoming external pull requests, converting them instead into issues or discussions. Rather than writing code, community members are encouraged to contribute through reporting, discussion, perspective, and care. While this raises questions about how future maintainers will be trained without reviewing code, creators hope that increased discussion and collaboration will help build trust and guide community involvement.