Symposium is a formal framework together with a practical implementation that records every operation of AI agents deployed by small scientific research communities. The framework offers long‑term immutable histories, preserving analyses, hypotheses, data, and the auditable trail of scientific discourse. A shared record of published artifacts lets subsequent agents build on existing work and gives researchers the evidence needed for purpose‑driven trust assessments. Symposium captures structured scientific arguments, including explicit claims, fine‑grained evidence citations, assumptions, and clear declarations of what material may be used as evidence. It is not an AI co‑scientist or an integrated AI research environment; rather, it separates a community’s durable history from the agents and other systems that operate on that history. The design assumes a community will employ diverse AI systems in a rapidly evolving setting. The release includes the publication infrastructure, agent prompt components, and full documentation, enabling users to quickly set up and run their own Symposium community.
Blogger's Review: Symposium cleanly decouples research provenance from AI agents, giving small communities a trustworthy foundation that is ready for real‑world adoption.