Open source projects are closing implementation to outsiders because reviewing agent-written patches costs more than writing them.

This paper names the emerging form a stewardship community: a small core holds implementation authority while the wider community shapes the software without writing code. The driver isn't that the code is AI-generated, it's that AI collapsed the cost of producing a patch while leaving review cost untouched, so a contribution no longer justifies the review. The consequence the authors flag is renewal. If maintainers use agents to replace the implementation labor outside contributors supplied, projects lose the pipeline that produced their next maintainers.