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.

Version 1.2 (March 2026)
This update adds three system diagrams clarifying Fathom’s architecture, recursive generative mechanism, and institutional positioning. Minor formatting improvements and an expanded “Relationship to Prior Work” subsection have also been included. No changes have been made to the core conceptual framework or system description.