The UNI Workshop

$75,000. Built to teach UNI.

The Workshop teaches UNI as a buildable active-inference architecture, not LLM tooling. Comparable AI and large-language-model programs commonly price near $30,000; we make the cost negotiable through the grant.

Apply for a scholarshipWhat we do not claim
What it teaches
UNI, a buildable active-inference architecture (not LLM tooling)
Price
$75,000
Common comparison
AI and LLM programs commonly price near $30,000
What we optimize for
A durable understanding of UNI and active inference, not the lowest sticker price
Scholarships
Partial to full ride, through the EducateWright grant
Prerequisite
Curiosity, and the willingness to be corrected
A durable understanding of UNI and active inference is what we aim for. It is a promise about how we teach, not a measured outcome we are claiming today.

What is inside

The science, and the discipline that keeps it honest.

Active-inference foundations

The established mathematics of active inference and the Free Energy Principle (Friston and colleagues), the standard science UNI is built from. Not new mathematics, and not an established field of its own.

The correctness discipline

The Polzin layer: bound directions, model versus process, the limits of the analogies. It is correctness discipline, not new mathematics. You learn to state things exactly.

Build-grade practice

You build UNI, not just watch it: typed model specifications, generative-model components, genome-style encodings, POMDP-style agents, and validation tests. Hypothesis, prediction, observation, provenance, and the loop that keeps your own claims honest.

What we have seen, so far

Preliminary observations, offered for reproduction.

In UNI simulations, agents configured with minimal engineered priors, weak or uniform priors, or learnable hyperpriors have shown learning and viability-preserving behavior under defined test conditions; agents with learned priors have shown changes in behavior and model organization. These are observations to be reproduced and documented, not yet peer-reviewed results. Some remain unpublished until they can be presented with adequate methods, tests, and limitations.

The honest boundaries

What the Workshop is, and is not.

  1. ×

    A preprint and reproducibility materials are archived on Zenodo. Expert peer review is invited and pending; until it completes, treat the work as a preprint and build-audit artifact, not an authoritative scientific result.

  2. ×

    We do not claim the Workshop produces achieved general intelligence, consciousness, biological equivalence, or a finished system.

  3. ×

    On some standard benchmarks a larger general model scores higher, and we say so.

The paper is on Zenodo as a working preprint. Read it, and try to break it. That invitation is the whole point.

Apply for a scholarshipExplore UNI