Foundation models were trained on open source code. The developers who wrote that code received nothing in return — not attribution, not compensation, not influence over how the models were built or deployed. The agent sessions that now run over those same codebases generate reasoning traces that are the most valuable new artifact in AI-assisted development. If those traces are captured in proprietary formats by proprietary tools and used to train proprietary models, the extraction continues. If they are captured in an open format, stored in a community-owned dataset, and used to train open models, the relationship between the open source community and the AI industry becomes reciprocal for the first time. This is not a sentimental argument. It is a structural argument about who should benefit from the value generated by open source code.