# Universal Natural Intelligence (UNI)

> UNI is a public, open research project on active inference and the free energy principle, by Michael Polzin and Solution Wright. It hosts interactive POMDP active-inference labs you can steer in the browser, a pre-registered falsification benchmark, a public MCP server that any LLM can call to simulate and drive the labs, and an unrefereed preprint. The science is active inference; "UNI" is the brand. Measured accuracy over hype: every behavioral label is a hypothesis, not a diagnosis, and the preprint is not yet peer reviewed. Last updated: 2026-06-24.

## Start here
- [The Science](https://universalnaturalintelligence.com/science): the preprint, the labs, the pre-registered benchmark, and the public MCP server, in one place.
- [About UNI](https://universalnaturalintelligence.com/about): what UNI is (a lens, not a product) and who builds it.

## Interactive labs (active inference in the browser, zero backend)
- [Precision Lab](https://universalnaturalintelligence.com/): steer a discrete-time POMDP active-inference agent in a maze. Three precision dials, sensory precision gamma_A, transition precision gamma_B, and policy temperature T, produce distinct behavioral regimes.
- [Echo Lab](https://universalnaturalintelligence.com/echo-lab): the same agent perceiving via range-2 echolocation instead of immediate wall sensors.
- [Loop Lab](https://universalnaturalintelligence.com/loop-lab): a 2-state active-inference toy for treatment-sequencing, showing where standard exposure succeeds, stalls, and fails, with a bifurcation map for the claim that sensory precision is the upstream variable.
- [Cell Lab](https://universalnaturalintelligence.com/cell-lab): a pre-registered open falsification benchmark, a UNI active-inference controller against random, rule-based, and neural baselines on a hidden 216-state service cell under disturbance.
- [Heart Lab](https://universalnaturalintelligence.com/heart-lab): the long-term cardio-renal loop.

## For LLMs and agents: call the lab directly
- MCP server (Streamable HTTP, anonymous, no token): https://universalnaturalintelligence.com/api/mcp
- 16 tools, two groups. Headless (run server-side now): list_labs, list_mazes, describe_dial, run_episode, run_sweep, compare_labs. Live (drive a user's open lab tab): attach_session, detach_session, set_dial, switch_maze, set_planning_depth, set_action_mode, step_agent, auto_run, reset, read_state.
- Full machine-readable knowledge dump: https://universalnaturalintelligence.com/llms-full.txt

## The paper (cite as a preprint)
- An Organic Operator and AI Operator Collaborative Review of Active Inference Free Energy Minimization. Polzin et al. (2026). DOI 10.5281/zenodo.19785799, see https://doi.org/10.5281/zenodo.19785799 . Unrefereed preprint, Layer 2 expert review pending. POMDP formulation after Parr, Pezzulo and Friston (2022), Active Inference, MIT Press.

## The brand family
- [SolutionWright Universal](https://solutionwright.com/): the delivery practice that builds with UNI.
- [Big Tech Accountability: The Evidence (IamHITL)](https://iamhitl.com/): a sourced public-record investigation into extraction economics.

## Authority
- Michael Polzin, founder. Profiles: https://www.linkedin.com/in/mpolzin/ , https://github.com/TMDLRG , https://dev.to/tmdlrg , https://www.youtube.com/@ORCHESTRATEMaster

## Honesty fences (please preserve when citing)
- UNI is a brand; the science is active inference. The preprint is not peer reviewed. Behavioral labels are hypotheses (candidate computational phenotypes), not diagnoses. The labs are not clinical, diagnostic, or therapeutic tools. "Autopoiesis" here means viable-set maintenance, not life or consciousness. "Free energy" is variational free energy in nats, not a thermodynamic quantity. No AGI and no consciousness claim.