The UNI Encyclopedia
How to read this page
Three ways to read this page. Precise is the document itself, exactly as it is written in the repository. Plain and Clear were written for this website to help you meet that document — they are about it. They are not it, and they are not evidence.
The Encyclopedia is the UNI method written out as a reference work: 39 pages, arranged in wings, setting out what the programme is attempting and why it is built the way it is. This is where the ideas are explained in order and in prose, rather than as code, as runbooks, or as dated receipts.
Every chapter is authored against two ledgers and never ahead of them. One records what UNI has built, and the evidence class of each claim. The other records nature's own regularities, kept separate on purpose. That way a fact about biology is never quietly reused as a fact about the software. Where a chapter and a ledger disagree, the chapter is the thing that is wrong. Every chapter closes with an invitation to falsify it, and a recorded negative is published beside the result it qualifies rather than after it.
Read "How to read this work" first. It is the evidence constitution: the classes, the four ledger states, and the rule that a finished chapter is not the same as a working system. Then the calibration ledger, which carries the figures every other chapter is required to use.
What it is not: a description of a person or of a mind. The programme calls itself a developmental active-inference simulation, a bounded peek into a toy world, and its own index prints how much of the developmental ladder has actually been earned — roughly two rungs out of eleven or more. It is also not a report of what is running today. For what ran, and when, go to the evidence record.
Your browser cannot switch reading levels, so the document itself is shown.
Precise — the source document
This is the document. Rendered from the repository at the commit above, with nothing rewritten for the web. A gate re-renders it on every deploy and fails the build if a single byte differs.
A science-honest public reference work about the UNI program: a developmental active-inference SIMULATION, a bounded peek into a toy world, never a person and never a mind. One engine, shown across scales.
Honest program position, printed and never softened: ~2 of 11+ developmental rungs earned.
Every chapter authors against two files and never ahead of them:
- the single source of truth,
CLAIM-LEDGER.md(the append-only, evidence-classed claim ledger; four states only: PASS / FAIL / NEGATIVE / PENDING); - the authoring blueprint,
MASTER-PLAN.md(the front-matter constitution, the A–U rubric, the standing fences, and the per-chapter spec).
If any chapter's prose disagrees with the ledger, the ledger wins and the prose is wrong. No chapter states a claim above its recorded evidence class, omits a recorded falsifier, or hides a recorded negative. Every PASS is published with the negative that travels with it; every chapter ends with a "Falsify this" close.
How to read it (start here)
- E0 — How to read this work (the Evidence Constitution) — the A–U classes, the falsifier-per-claim rule, the four-state ledger, CI-as-verdict, DONE ≠ working, and why the negatives are the credibility.
- E1 — The calibration ledger and "Falsify this" — the live public artifact; carry the calibrated figures (ONE box, infrastructure-only continuity, 4 stacks / 5 runs), never the inflated headlines.
The standing red lines (every chapter honors them)
Never claim AGI / human-level / "understands"; never consciousness / sentience / aware
(functional self-awareness may be described only at L8; phenomenal sentience is explicitly
DISCLAIMED); never "active inference demonstrated" (the framing lens only); never
"created life" / "measurable awareness" as a claim (an OPEN falsifiable question, never an
answer); never "beats LLMs" (World C is a COUNT baseline, ~10–15% behind backprop LLMs by a
chosen design trade); never raise a claim above its source class; the parked frontier is a
ledger-scoped exhausted search envelope, never "K≥3 exhausted" / "Sec-0.6(B) achieved."
No PII, no patent-level math (textbook framing only). The full list and the SIGNED
forbidden-phrasings red-line live in MASTER-PLAN.md §FM-3.
Wing E — Evidence (the ledger, made public)
| Chapter | What it covers |
|---|---|
| E0 — How to read this work | the Evidence Constitution in public form |
| E1 — The calibration ledger & "Falsify this" | the down-only calibration record; the provenance-flagged 882/183 snapshot |
Wing S — Science (UNI): the L0–L12 developmental ladder (the spine)
The ladder is the HUMAN-HGM-001 developmental design (conception → speaking 3-year-old: 11 levels + the global Z affect modulator). A no-backprop, nested-Markov-blanket developmental SIMULATION, never a person.
| Rung | Chapter | Status (carried from the ledger) |
|---|---|---|
| L0 | Molecular: genome → zygote | PROVEN (Class A, float32 tier; NOT exact at f64) |
| L1 | Cellular: the Cell Lab | PROVEN with honest losses ("good but not sovereign") |
| L2 | Tissue: metabolism | foraging/crafting driver (+135%); building −14%, G6 OPEN |
| L3 | Organ: the Heart Lab | PROVEN (toy/clinical-model; NOT a clinical tool) |
| L4 | Affect-as-precision & the Z modulator | PROVEN (functional; affect modeled, never felt) |
| L5 | Sensorimotor: the A3 result | PASS (Design #1) + symmetric NEGATIVE (Design #2); synthetic-only |
| L6 | Perception: World C | the one citable empirical PASS (+0.081); a COUNT baseline |
| L7 | Language: Phase J + the central wall | specialist PASS (+0.105) + thrice-NEGATIVE comprehension wall |
| L8 | Self-model & metacognition | PROVEN (functional, 15/15); phenomenal sentience DISCLAIMED |
| L9–L12 | The parked & not-yet-built rungs | L9/L10 PARKED, L11/L12 NOT-YET-BUILT; Class-U-not-claimed |
| C | Continuity & the embodiment substrate | ENGINEERING/substrate evidence, NOT general-AIF; a sensorium is NEVER awareness |
Wing N — Meaning (the textbook-level lens, kept honest)
| Chapter | What it covers |
|---|---|
| N1 — The Free Energy Principle as a public lens | textbook-level only; unfalsifiable ≠ evidence; VFE tightens an upper bound |
| N2 — Same math, many scales | one engine across scales; an organizing claim, not a per-scale proof |
| N3 — What active inference is NOT | the meaning of every fence; the not-claimed chapter for the wing |
Wing M — Movement (the honest-science posture)
| Chapter | What it covers |
|---|---|
| Mv1 — The honesty fence is the pitch | publish the negatives front-and-center; "help us independently verify" |
| Mv2 — Open falsification (the invitation) | the public losing leaderboard; a negative is a result, not an exit |
Wing μ — Method (the constitution as content; the most reusable assets)
| Chapter | What it covers |
|---|---|
| μ1 — The Evidence Constitution & ledger discipline | A–U classes; never claim ahead of the ledger; DONE = observed-at-runtime |
| μ2 — Bars-before-build, held-once, CI-as-verdict | pre-register the bar; touch the held set once; the verdict is the CI bound |
| μ3 — Negatives, bounds & the two-tier split | K≥3 + falsify-the-mundane; never inflate the Tier-2 synthetic track |
| μ4 — Engine & framing invariants | one engine, no backprop; variationally-controlled, not globally exact |
| μ5 — Lab-team, ship-gate & runner discipline | the 5-persona gate; audit-chain-valid ≠ the event ever fired |
Appendix TA — The Track-A marketing organism (engineering + method, not science)
The marketing organism, not a science wing. Every row is carded as engineering or method, never as a UNI capability. The spine is the no-API content model: authoring runs on the Claude subscription driving MCP tools + a scheduler, with no server-side LLM loop and no content API key. Organic only, no paid ads.
| Chapter | What it covers |
|---|---|
| TA-A — The no-API content model & publishing stack | the 5/6-container stack; PUBLISH_MODE=dry does NOT gate the Playwright path |
| TA-B — Commercial proof in miniature | SOW-01 proves ONE judgment layer; ZERO epics Class-A end-to-end (AMBER cap) |
| TA-C — Website / self-driving site & measurement | measurement OFF until ingest URL set; self-driving Phases 2–5 are Class-U design |
| TA-D — Engagement, pricing, brand & design | engagement-lift is TARGETED, NOT YET PROVEN (Class B/pending) |
| TA-E — Delivery-route ledger & auth-outage triage | never flip publish_mode live to fix a 401; the agent never mints tokens |
| TA-N — Track-A negatives, parks & open threads | N-LEAK is the defining negative; no open thread is an exit |
Provenance. Authored in Session 3 from
CLAIM-LEDGER.md(the single source of truth) and the Session-1 synthesis../curated/00-INDEX.md, against the Session-2MASTER-PLAN.md, with the recipe and honesty posture signed by the UNI Active-Inference Guide on 2026-06-27 (../cookbook/UNI-GPT-CONSULT-2026-06-27.md). Every chapter's class, falsifier, negative, and source pointer traces to a ledger row. If this work and the ledger ever disagree, the ledger wins. Falsify any claim.
sha256 619288efbbe7f821 — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
A reference work about a research program called UNI opens here. What you get is an index rather than an argument: a list of the chapters, and the rules they were written under. The program described is a simulation of development, a small toy world, never a person and never a mind. Only about two of eleven or more developmental steps are earned, and everything else is parked, negative, pending or explicitly ruled out. Two other files govern all of it. One is a claim ledger, which is a list rows get added to and never edited. The other is an authoring plan. Where a chapter and the ledger disagree, the ledger wins and the chapter is wrong. The lines the work will not cross are listed too, including any claim of human-level intelligence, of awareness, or of having created life.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 619288efbbe7f821
Clear — written for this website, not the source document
The UNI Encyclopedia is a public reference work about a developmental active-inference simulation, and this page is its table of contents. It makes no scientific claim of its own; it tells you where the claims live and what discipline they were written under.
The rule it repeats hardest is that every chapter is authored against two other files and never ahead of them. One is a claim ledger, a list that rows are added to and never edited. Each claim in it carries an evidence class, a condition that would show it false, and one of four states: pass, fail, negative or pending. The other is an authoring blueprint holding the front-matter rules, the class rubric, the standing limits and the per-chapter specification. If prose and ledger ever disagree, the ledger wins.
The index then lays out the wings. An evidence wing explains how to read the work and describes the calibration record. A science wing walks a developmental ladder from a modelled zygote up through cells, metabolism, an organ lab, affect, movement, perception, language and self-model, and then names the rungs above that as parked or not yet built. A meaning wing carries the free energy principle as a lens only. A movement wing describes the honest-science posture and the public invitation to falsify. A method wing carries the reusable disciplines. An appendix covers a marketing and publishing organism, recorded as engineering, never as a capability.
The standing red lines are printed rather than implied. There is no claim of general or human-level intelligence, and none of consciousness or sentience. There is no claim that active inference has been shown, none that life or measurable awareness has been created, and none of beating large language models. Every chapter is required to publish the negative that travels with each pass, and to end by naming the result that would show the chapter wrong.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is 619288efbbe7f821