A federation of minds
PANTHEON is an experimental AI federation. Instead of one big model, it’s many small minds — each doing exactly one thing (reasoning, thinking, knowing, learning) — wired together and running live across a handful of machines. You can watch each one work.
Each mind does one thing. Open one and watch it work.
A traceable chain of thought — it steps from one idea to another, its free energy falling to a minimum as it reaches a conclusion. Grounded, no-drift.
watch it reason → mmthinkAny organ can ground a question in the real graph and reason it through with a language model — a chain of thought, not just an answer.
give it a task → mmknow691,293 real facts across every domain — chemistry, medicine, history — with real relations. Ask it about almost anything.
explore what it knows → ORACLEThe reasoning core — a self-growing knowledge graph of concepts and relations, with an elastic capacity that manages itself.
see the live system → atlasEvery agent in the federation, mapped to the machine it runs on — current vs. designed placement.
open the atlas → anatomyAll 68 agents arranged as one organism — senses, memory, reasoning, learning, hands — by what each does.
see the anatomy →What this is — honestly. PANTHEON is a research system, not a product and not an oracle. It demonstrates specific, measured things: real semantic retrieval, grounded multi-step reasoning, a cross-domain knowledge base, and a graph that scales itself — each shown working, in the open. It reasons over what it knows and says so when it can’t; the language parts can still be wrong. Nothing here should be taken as authoritative fact or advice.
live · an experiment by thataiguy.org · the federation runs across several machines you can watch in the system view