Fathom: A Runnable Testbed for Studying AI-Mediated Information Dynamics

This paper introduces Fathom, a deliberately constrained, runnable system designed to study feedback loops, visibility dynamics, and recursive generation in AI-mediated information environments. Rather than proposing a new platform, Fathom is framed as an institutional “organ” — a bounded research instrument that enables direct observation and experimentation with emergent informational behaviors. The system is open-source, local-first, and intended to support research, education, and public-interest exploration of AI-mediated systems.