The NATURA wing — nature's regularities, kept sovereign from UNI's build status
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.
The rest of this encyclopedia describes UNI: what the program built, at what evidence class, with the negatives printed. This wing describes nature: the regularities that were measured by other people, in published work, long before UNI existed.
It exists to serve one mandate — that a builder be able to reason about design "down to the cell level," about "the ratios and frequencies," about "mind body mind.body world," and about how to "research observe and make new science with falsifiable evidence to always use nature as the authority of where we will find the next breakthrough" — across rocks, water, air, stars, DNA, sperm, ants, dinosaurs, whales, bats, humans, and the open question beyond.
That mandate is answerable only if the wing is disciplined about a distinction the rest of the corpus never had to make: the difference between a gate UNI held and a measurement somebody else published. Both are evidence. They are not the same kind, and they do not license the same moves.
The honest program position is untouched by every chapter in this wing, and this wing can never change it: ~2 of 11+ developmental rungs earned; a developmental active-inference SIMULATION; a toy world, never a person. If a NATURA chapter ever appears to move that number, the chapter is the defect.
THE SOVEREIGNTY RULE (read this before anything else in the wing)
A nature citation is NEVER a UNI gate.
Reading Kleiber's law raises no UNI rung. Citing Douady & Couder does not make any UNI claim
proven. A chapter that lets a literature citation imply a UNI capability has committed a
lane-crossing and is defective — regardless of how good the citation is.
This corpus carries two sovereign evidence vocabularies, and they are never merged:
| The UNI 4-value fence | The NATURA 12-value class | |
|---|---|---|
| Describes | UNI's own build status | nature's observed regularities |
| Values | proven · designed · hypothesized · not-yet-built |
twelve, in three groups (below) |
| Measured by | UNI's own gates (PASS / FAIL / NEGATIVE / PENDING) | third parties, in published literature |
| Governed by | ../CLAIM-LEDGER.md |
../NATURE-LEDGER.md |
The twelve classes, in three groups (defined in full, with a worked example of each, in NA-00; this is a pointer, not a restatement):
- Group A — measured (the class says how much independent corroboration exists):
OBSERVED-REPLICATED·OBSERVED-SINGLE·OBSERVED-CONTESTED - Group B — derived (the class says the assumptions are the fence):
MODELED·MODELED-CONTESTED·HYPOTHESIZED - Group C — fenced (the row carries no usable value; the class says precisely why not —
these are not interchangeable):
INADMISSIBLE·SUPERSEDED·NOT-MEASURED·NOT-SOURCED·NOT-CONFIRMED·NOT-LOCATED
The two ledgers are cross-referenced by explicit audited link only, never by merge. There is
no operation that takes a NATURA row and a UNI row and produces a stronger row of either kind.
proven is reserved for UNI-ledger status and never describes nature's science — not even
this wing's own best-replicated rows.
One fence inside Group C is load-bearing and is stated twice on purpose: NOT-SOURCED is
never NOT-MEASURED. NOT-MEASURED is an absence in the literature — a reportable fact
about the world's knowledge. NOT-SOURCED is an absence in this corpus's own homework — a
fact about us. Collapsing them would let a research failure masquerade as a discovery about
the frontier of science.
TRUE and HONEST stay sovereign. Every one of the twelve classes is a TRUE-side instrument:
falsifiable, reproducible, recalibratable. Lived experience is never calibrated — a
chakra-frequency table is INADMISSIBLE as physics and may be recorded as an HONEST/cultural
signal, never as a TRUE/measured one, and never converted between them. The wing's only
HONEST-class row (CN11-59, first-person testimony) is deliberately given no NATURA class.
Supremacy. Where a chapter and the ledger disagree, the ledger wins and the chapter is wrong — the same rule the UNI side already runs.
The chapters (all 23)
The fence column is each chapter's own ## HONEST FENCE line, read from the chapter itself
— not inferred from its title, and not a summary of it. It is the fence on the chapter as an
artifact; individual rows inside carry their own classes, and those govern. A chapter
fenced MODELED is typically a budget or a composition: it runs measured constants through stated
assumptions, and the assumptions are the fence.
Reference chapters — encyclopedia/wing-NATURA/
| Chapter | What it covers | Honest fence |
|---|---|---|
| NA-00 — How to read this wing | the wing's constitution: the twelve classes, the cardinal rule, the crosswalk against the UNI fence, and the amendment record | NOT-MEASURED |
| NA-01 — Nature as the authority | why the already-run experiment is authoritative; convergence as constraint-optimum; the Gould & Lewontin counterweight in full; the operable protocol (§6) is the payload | HYPOTHESIZED |
| NA-02 — The one loop | active inference stated exactly (VFE to understand, EFE to choose), with its math in readable ASCII — and the best published objections at equal seriousness | MODELED |
| NA-03 — How to make new science | the operable procedure for turning an observation into a claim that can be killed: strong inference, severity, pre-registration, tuned baselines, discriminators, and how a negative gets published | OBSERVED-REPLICATED |
| NA-04 — Mind / body / mind.body / world | the Markov blanket given exactly; each of the four words mapped to one of its four sets; the nesting across scales — then a last third arguing against itself | MODELED |
| NA-05 — The ratios | allometry and scaling across 20+ orders of magnitude; the exponent is the science, the prefactor is the engineering; Kleiber's dispute carried at full strength | OBSERVED-CONTESTED |
| NA-06 — The frequencies | nature's rates separated from nature's rumours: Strouhal, Schumann, alpha, circadian — the earned/unearned split, with the receipts for both | OBSERVED-CONTESTED |
| NA-07 — The dimensionless numbers | the cross-scale design toolkit: Reynolds, Péclet, Damköhler and kin; why dimensionless groups travel across scale and ratios usually do not | MODELED |
| NA-08 — Cell-level design | a working budget for the cell as an engineered artifact: what sets its size, what it spends, what its machines deliver, where its sensing hits a physical floor | MODELED |
| NA-09 — Morphogenesis | the bridge from cell to organism: pattern from local rules, reaction–diffusion, and a morphogen gradient as a measured concentration profile | OBSERVED-CONTESTED |
| NA-10 — The scale ladder | a rung-by-rung placement table across ~36 orders of magnitude — and the rule for knowing when a number fitted on one rung stops meaning anything on the next | MODELED |
Recipes — cookbook/recipes-natura/
Each is a buildable recipe carrying its own NATURA classes and its own falsifier.
| Recipe | What it covers | Honest fence |
|---|---|---|
| CN-01 — Rocks | the mineral substrate everything else is made of and stands on; the wing's reference case for a boundary with no internal model — the zero of the ladder | MODELED |
| CN-02 — Water | the spec sheet of the medium every other chapter silently assumes; a pile of anomalies life is built on, and the substance that attracts the most unearned mysticism while being measurably stranger than it | MODELED |
| CN-03 — Air | the atmosphere as a working fluid: composition, density, momentum, sound, light, its heat engine, and where it refuses to hold you up | MODELED |
| CN-04 — Stars | the provenance chain for the matter in NA-08's cell, proton to carbon to gold; a star read as a negative-feedback control system | MODELED |
| CN-05 — DNA | the information substrate as a physical polymer — measured dimensions, information content, error rate, packing, copy speed — and the deflation of reading DNA as source code | MODELED |
| CN-06 — Sperm | the minimal motile delivery vehicle, designed against a spec that admits no negotiation; the physics writes most of the requirements document | MODELED |
| CN-07 — Ants | the colony as a distributed system whose coordination lives in the shared environment rather than in any component's model of the whole | OBSERVED-CONTESTED |
| CN-08 — Dinosaurs | falsifiable science on a system you can never observe; the animals are the worked example, the epistemics are the payload | MODELED |
| CN-09 — Whales | what sets a maximum once buoyancy deletes the bone-stress ceiling; low-frequency sound as a long-baseline information channel, with the units trap that makes most "loudness" comparisons wrong | MODELED |
| CN-10 — Bats | the wing's keystone chapter for the method: the one animal that exposes active inference in Hz, dB, ms and metres — where the epistemic term becomes an instrument reading | MODELED |
| CN-11 — Humans | the human body as measured quantities with the flattery stripped out; famous numbers carried with the receipt of their correction | MODELED |
| CN-12 — Beyond human | not a recipe. A QUAESTIO-APERTA register entry: it maps the constraint space and states what would falsify each claim people make about it. No target, no milestone, no roadmap — and there must never be one | HYPOTHESIZED / QUAESTIO-APERTA |
Reading order, if you are here to DESIGN something
Canonical in NA-00 and carried here unchanged. The order is deliberate; the reasons are the point.
- NA-00 — the vocabulary and the cardinal rule. Non-skippable.
- NA-03 — the method before the content, so you cannot mistake an inspiration for a result.
- NA-01 — the authority doctrine and its counterweight, together, in that order.
- NA-07, then NA-05 — dimensionless numbers before ratios. Dimensionless groups survive a change of scale and most ratios do not. Reading NA-05 first is how designers end up with golden ratios in places nature never put one.
- The scale chapter you need — NA-08 (cell), NA-09 (form), or NA-10 (ladder).
- The CN recipe for your target — and only now.
- Back to NA-03 — write the pre-registered metric and the tuned baseline before you build. A biomimetic design with no baseline is not a design; it is a story.
The ledger, the machine twin, and the mechanical check
../NATURE-LEDGER.md— the wing's ledger of record. 919 rows, extracted from every## The numberstable across these 23 chapters and carried whole. It is extraction only: no row was added that no chapter contains, and no value was invented, improved, averaged, rounded, or completed. Ranges and disputes are carried whole and never averaged. Where a chapter and the ledger disagree, the ledger wins. (918 table rows +NA03-16, which NA-03 places in its prose on purpose = 919; the reconciliation is disclosed at its §2.)../../gpt/knowledge/K20-constants-ratios-and-nature-ledger.json— the machine-readable twin. The categorised view: constants, dimensionless numbers, scaling laws, frequencies, the INADMISSIBLE register, and the open questions, every entry tracing to its chapter. Both files are generated from one extraction pass and cannot disagree.../../tools/verify_class_vocabulary.py— the mechanical check. The operable form of NA-00's falsifier #2, pre-registered before the amendment it tests. It fails on any row outside the registered vocabulary, across three scopes — the ledger, K20, and the chapters themselves — and prints every scope it covers before it prints any result. Re-run it after any edit to any of them.
The honest residue (stated here, not buried)
Two things these amendments did not fix. They are printed in the index because an index that points only at the clean parts is not an index, it is an advertisement.
- The 22 untraced rows. The
NOT-SOURCED(20),NOT-CONFIRMED(1) andNOT-LOCATED(1) rows are now correctly labelled, which is not the same as fixed. Twenty-two rows carry values this corpus could not fully trace. Naming a defect precisely is not repairing it. Each carries the falsifier that would close it, and closing them is real work that has not been done. - The 52 chapter rows, and the day they were wrong. AMENDMENT 2026-07-15-A closed
NATURE-LEDGER.mdand K20 — those two files, and nothing else. 52 rows across 10 of these 23 chapters went on carrying PRE-amendment class strings (42 bare/qualifiedOBSERVED, 3NEGATIVE, 5as above, 2 CN-05 cells not leading with a token) until AMENDMENT 2026-07-15-C re-classed them in lockstep with the ledger. For one day the extraction and the extracted disagreed in the one column the amendment existed to fix — and three of those cells printedNEGATIVE, a UNI ledger state, in the Class column of a NATURA table: the cardinal rule was argued correctly in the ledger and violated in the chapters. The verifier reported PASS anyway, because it never opened a chapter. They are labelled correctly now. What is not repaired is the reason it happened: nothing in the record disclosed the gap — it took an outside audit to find it. The instrument is fixed; the habit that produced it is not demonstrated fixed, and one closed scope gap is not evidence there is no second one.
Read NA-00's Amendment record and
NATURE-LEDGER §3.2 for the full accounting — including why 50 and 52 are
both right under two different instruments (52 − 5 + 3 = 50), and why a count is not a fact until
the rule that produced it is named.
Provenance. This index points; it asserts no fact about nature of its own. The wing's constitution is
NA-00, which defines the twelve classes and the cardinal rule; the ledger of record is../NATURE-LEDGER.md, and where this index and the ledger disagree, the ledger wins and this index is wrong. Every fence in the tables above was read from that chapter's own## HONEST FENCEsection, and every count restated here (919 rows · 72 of 919 outside the original six · 22 untraced · 52 chapter rows across 10 files) is quoted from NA-00's Amendment record and NATURE-LEDGER §3.2 / §5, not re-derived. The class vocabulary is machine-checked by../../tools/verify_class_vocabulary.pyacross the ledger, K20, and these chapters. The sibling index for the UNI side of the corpus — a sovereign ledger, a sovereign vocabulary, never merged with this one — is../00-INDEX.md. A nature citation is never a UNI gate. Falsify any row.
sha256 f0e001e2e10f642e — of the original file, so what was ingested stays checkable.
Plain — written for this website, not the source document
Nature is the subject of this wing; the program is not. Nobody here did the measuring. Every regularity gathered was published by other people, so it contributes no evidence to the program's own results. Somebody else's numbers cannot settle what the program has yet to show. The program's own position stands untouched by every chapter that follows: a developmental active-inference simulation, a toy world, never a person. Two vocabularies for evidence run side by side here and are never merged. One tracks the build's status. The other sorts nature's regularities by how much independent corroboration stands behind each one. Lived experience stays separate from measurement too, so a cultural frequency table can be recorded honestly and is never converted into physics. The rest is a map: which chapters cover what, which recipes can be built, a reading order, and a frank list of rows the amendments did not fix.
Plain · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is f0e001e2e10f642e
Clear — written for this website, not the source document
Everything gathered in this wing is borrowed science. Other people did the measuring and published it, and the wing quotes them, so it contributes no evidence to the program's own results. The program's own subject is a developmental simulation — a toy world, and never a person. Neither of those sentences is decoration. Without them, somebody else's published numbers would read as though they had settled what the program has yet to show.
The wing answers one mandate. A builder should be able to reason about design down to the cell, about ratios and frequencies, about mind and body and world, and about how to make new science with falsifiable evidence. Nature is the authority on where to look next. That is answerable only if the wing keeps a distinction the rest of the corpus never had to make: the difference between a test the program itself passed and a measurement somebody else published. Both are evidence. They are not the same kind and they do not license the same moves.
A chapter that lets a literature citation imply a capability has crossed lanes and is defective, however good the citation is. The two records are cross-referenced by an explicit audited link only, and never merged. No operation takes a row from one and a row from the other to produce a stronger row of either kind.
The classes are then summarised in three groups. A measured group says how much independent corroboration exists. A derived group says the assumptions are the limit. A third group carries no usable value and says precisely why not. Those ways of being empty are deliberately not interchangeable: an absence in the literature is a fact about the world's knowledge, while an untraced value is a fact about this corpus's own homework.
The page also keeps lived experience separate from measurement. A cultural frequency table is inadmissible as physics and may be recorded as an honest signal, never converted into a measured one.
The bulk of the page is two tables mapping the reference chapters and the buildable recipes, each with the limit read from that chapter's own words. A reading order follows, deliberately putting method before content and dimensionless groups before ratios. Finally the page prints what the amendments did not fix. A set of rows still carries values nobody has traced to a source. And there was a day on which the chapters and the wing's own list of classes disagreed in the one column an amendment existed to fix.
Clear · written 2026-08-01 by claude-opus-5 · not yet checked by a person · about the document whose sha256 is f0e001e2e10f642e