NA-00 — How to read the NATURA wing (the second sovereign vocabulary)

What you are reading. The constitution of the NATURA wing. This chapter defines the wing's evidence vocabulary, fences it away from the UNI claim ledger with a hard rule, and maps the wing. Read it before any other NA or CN chapter. It asserts no fact about nature of its own — every number in it is a worked example, carrying its own class, its own citation, and its own falsifier.

AMENDED 2026-07-15. This chapter originally declared a six-value vocabulary — "six classes and only six." That declaration was wrong, and the corpus proved it wrong: 72 of the 919 rows the chapters actually wrote could not be placed in any of the six. The vocabulary is now twelve classes in three groups, plus four individually-named rows that deliberately carry no class at all. The original wording, and the strike against it, are carried openly in § Amendment record — not silently edited away.

---

Why this wing exists

The rest of the encyclopedia describes UNI: what the program has 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.

The wing serves one mandate — that the science consultant 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 "dna sperm ants dinosaurs whales bats rocks water air stars 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, they do not license the same moves, and the whole wing turns on never confusing them.

Per ../../README.md, this wing's ledger of record is ../NATURE-LEDGER.md. Every row in every NA/CN chapter is written to be admissible there. Where a chapter and the ledger disagree, the ledger wins and the chapter is wrong — the same supremacy rule the UNI side already runs.

---

The cardinal rule

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 — full stop, regardless of how good the citation is.

This is not a new rule. It is the same fence the corpus already applies to the embodiment substrate ("UNI.OS substrate evidence is never general-AIF evidence"), and the same one behind M9's class authority ordering. The NATURA wing simply gives it a second front. 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.

The honest program position is unchanged by anything in this wing, and this wing can never change it: ~2 of 11+ developmental rungs earned. If a NATURA chapter ever appears to move that number, the chapter is the defect.

---

The classes, with a worked example of each

The UNI fence has four values (proven · designed · hypothesized · not-yet-built) and the UNI ledger has four states (PASS / FAIL / NEGATIVE / PENDING). Neither vocabulary is used in this wing.

STRICKEN 2026-07-15 — the original wording, carried on record: "This wing uses six classes and only six." It was false. 72 of the 919 rows the chapters wrote (7.8%) carried a class string outside the six, because the six could not express what those rows had actually observed. See § Amendment record. This wing now uses twelve classes in three groups — and the count is a measured property of the corpus, not a design target. If a thirteenth is needed, that is a finding, and the same amendment procedure applies.

The three groups answer three different questions. Group A — measured: somebody put an instrument on nature; the class says how much independent corroboration exists. Group B — derived: the number came out of a model; the class says the assumptions are the fence. Group C — fenced: the row carries no usable value, and the class says precisely why not — the ways of being empty are not interchangeable, and collapsing them is how a corpus starts lying.

Group A — measured

OBSERVED-REPLICATED — measured, and independently replicated across labs, datasets, or taxa. Carries a real citation, a falsifier, and the known range. Worked example: the Schumann resonance fundamental, ≈ 7.83 Hz. Predicted by Schumann (1952), first extracted from background noise by Balser & Wagner (1960), Nature 188:638–641, and measured at ELF stations worldwide since. It is a real Earth–ionosphere cavity resonance. Note what the class does not buy: it licenses the frequency, not any claim attached to the frequency.

OBSERVED-SINGLE — (registered 2026-07-15; 42 rows) measured, by a method another party could repeat, but no independent replication is on record. One line of evidence: a single study, a single group, a single report, or a single chain through a review. The class asserts the measurement and asserts nothing about corroboration. Why it had to exist — the corpus's own words. CN-07 stated the reason exactly, and it is the whole argument for this class: "OBSERVED-REPLICATED asserts a replication that did not happen, and OBSERVED-CONTESTED asserts a dispute that does not exist. Both are false, in opposite directions." A measured-but-unreplicated result forced into either neighbour is a lie in a known direction; under the six, 42 rows had no honest home. Worked example: CN09-26 — nitrogen-washout collapse depth in Tursiops truncatus, ~70 m (Ridgway & Howard 1979, Science 206(4423):1182–1183). Real measurement, real citation, one group, never independently repeated. Falsifier — the promotion and the demotion, both live: an independent replication promotes it to OBSERVED-REPLICATED; a failed independent replication demotes it to OBSERVED-CONTESTED (or, if withdrawn, SUPERSEDED). Until one of those happens, the row stays here — and "nobody has tried" is a fact about the literature, printed as one.

OBSERVED-CONTESTED — measured, but the community genuinely disputes the value, the exponent, or the mechanism. Both positions must be carried and the dispute named. This is a feature, not an embarrassment. Worked example: the metabolic scaling exponent b in BMR ∝ M^b. Kleiber (1932) reported ≈ 0.74 (Brody, the same year, ≈ 0.73). Dodds, Rothman & Weitz (2001) re-analysed the classic datasets and found little evidence for rejecting 2/3 in favour of 3/4. White & Seymour (2003) report 0.686 ± 0.014 (95% CI). The exponent moves with the analysis choices — which taxa are excluded, whether body temperature is normalised, whether the regression unit is the species or the order. That mobility is the dispute, and a chapter that picks one number and drops the other has lied.

Group B — derived

MODELED — derived from a model that fits data. The model's assumptions are the fence. Worked example: West, Brown & Enquist (1997) derive b = 3/4 exactly, from three assumptions: a space-filling fractal branching network, terminal units invariant with body size, and minimised energy dissipation. The derivation is clean. It is also conditional: drop any one assumption and the 3/4 does not follow. Cite it as MODELED and cite the assumptions with it, or do not cite it.

MODELED-CONTESTED — (registered 2026-07-15; 1 row) derived from a model, and the community genuinely disputes it. Both positions carried, and the model's assumptions still the fence — the row must survive both tests, not one. This class exists because the original six were treated as disjoint and are not: MODELED and OBSERVED-CONTESTED overlap in reality, and a row landing in the overlap was previously forced to drop one of its two true properties. Worked example: CN12-31 — human behavioural throughput, ~10 bit/s (Zheng & Meister 2025, Neuron 113(2), a Perspective — a literature synthesis, not a new experiment), directly contested by Sauerbrei & Pruszynski (2025), Nat Neurosci. Modeled and disputed; calling it MODELED alone hides the dispute, calling it OBSERVED-CONTESTED alone claims a measurement nobody made. Falsifier: a measure reconciling cognitive throughput and motor control under one accounting.

HYPOTHESIZED — mechanism proposed, not settled. Not claimed. Worked example: the radical-pair mechanism of avian magnetoreception (Ritz et al., 2000; reviewed in PNAS, DOI 10.1073/pnas.0711968106). Radio-frequency disruption experiments are consistent with it, and behaviour at microtesla fields strengthens it (Nature Communications, DOI 10.1038/s41467-019-11655-2). Much of the supporting evidence remains circumstantial, and a competing magnetite-based hypothesis is live. Birds navigate; how is not settled.

Group C — fenced (the row carries no usable value; the class says why)

INADMISSIBLE — a claim that failed observation, or that is unfalsifiable as stated. Carried with the receipt of why it failed. Never asserted, never mocked. Worked example: "432 Hz is nature's true tuning." A4 = 440 Hz is a convention — ISO 16:1975 fixes it at 440 ± 0.5 Hz, not a natural constant — and 432 Hz is a different convention. Neither is more natural, and the claim names no measurand that could distinguish them. Inadmissible as physics or biology. The person asking is not stupid; the claim is just not shaped like a claim yet.

SUPERSEDED — (registered 2026-07-15; 3 rows) a value that was carried in the literature and has since been retracted, withdrawn by its own authors, or contradicted by a calibrated re-analysis. Kept as the receipt for why it must not be cited — never deleted, because deleting it is how a superseded number gets re-imported by the next reader who finds it in an old textbook. Distinct from INADMISSIBLE: an INADMISSIBLE claim was never shaped like a claim; a SUPERSEDED value was a perfectly good claim that lost. Worked example: CN01-34 — f_bio,2008 ≈ 2/3 of 4,300 minerals as biologically mediated (Hazen et al. 2008), withdrawn by the authors themselves. Scope cell reads: SUPERSEDED — DO NOT CITE. Falsifier: this class's rows have already had their falsifier fire. That is what put them here. On the name — read this before "fixing" it. These three rows were originally written NEGATIVE. This amendment deliberately did not register that string, because NEGATIVE is a UNI ledger state (PASS / FAIL / NEGATIVE / PENDING) and appears as such in the crosswalk below, where it is paired against INADMISSIBLE. Registering NEGATIVE as a NATURA evidence class would have put one token in both sovereign vocabularies at once — closing this defect by committing the exact lane-crossing the cardinal rule exists to forbid. The rows' own cells already said "SUPERSEDED FRAMING", "SUPERSEDED — DO NOT CITE", and "contradicted by"; SUPERSEDED is the class that was always true of them. The wing's separate, unrelated use of "NEGATIVE" — a biomimetic design that failed to beat its tuned baseline under M7 — is untouched and stays UNI-side, where it belongs.

NOT-MEASURED — the honest empty face: nobody has measured it. Write this instead of inventing a number. Worked example: the effect size of the Schumann resonance on human physiology at ambient field strengths. The fields are in the picotesla range. This chapter searched and found no pre-registered, independently replicated dose–response. That is not a refutation — it is an absence, and the absence gets written down as an absence.

NOT-SOURCED — (registered 2026-07-15; 20 rows) the chapter states a value it could not trace to a source. Somebody measured it; this corpus could not find who. This is NOT the same as NOT-MEASURED, and the difference is the whole point of registering it. NOT-MEASURED is an absence in the literature — a real, reportable fact about the world's knowledge. NOT-SOURCED is an absence in this corpus's own homework — a fact about us, not about nature. Collapsing the two would let a research failure masquerade as a discovery about the frontier of science, which is among the worst things this wing could do. Never merge them. Falsifier: produce the primary source. The row is then re-classed on its merits (usually OBSERVED-SINGLE or OBSERVED-REPLICATED) — or, if no source exists, the value is struck.

NOT-CONFIRMED — (registered 2026-07-15; 1 row) the primary source is named, at one remove — a secondary or tertiary work cites it — but the primary was not read in this pass. The provenance chain is known and unwalked. Distinct from NOT-SOURCED, where no chain was found at all. Worked example: CN10-53 — f_click,B.trigona up to 4,500 Hz, via a tertiary source citing Corcoran et al. 2009; primary not read. Falsifier: read Corcoran et al. 2009, Science 325:325–7, and confirm or strike.

NOT-LOCATED — (registered 2026-07-15; 1 row) a named source was searched for and does not appear to exist. Strongest of the three provenance fences: not "we didn't look" (NOT-SOURCED) and not "we didn't walk it" (NOT-CONFIRMED), but "we looked, and it wasn't there." Worked example: CN12-40 — doctrine row D3-08, cited in the chapter's own commissioning brief; a grep of uni-onchip/design/ found no such row. The chapter reports the miss against its own brief rather than quietly satisfying it. Falsifier: produce the row, or correct the citation.

The provenance fence, in one line. NOT-MEASURED = nature has not been asked. NOT-SOURCED = we cannot say who asked. NOT-CONFIRMED = we know who asked and did not read them. NOT-LOCATED = we checked, and the source named is not there. Four different failures. Four different repairs. One of them is nature's; three of them are ours.

---

Rows that deliberately carry NO class (registered as such, not classed)

Four rows carry no evidence class on purpose. They are named individually — this is a closed list of four, not an open category, and a fifth may not be added by anyone who simply finds classing inconvenient. Each is registered here so that a verifier can pass them without pretending they have a class they do not have.

Row	What it is	Why it is not classed
NA03-16	The SI defining constants (Δν_Cs, c, h, e, …), BIPM SI Brochure 9th ed., 2019	Definitional. Exact by convention, effective 20 May 2019. They are not observations of nature, so no NATURA class applies. NA-03 puts them in the prose below its table on purpose. The chapter's instruction is the point: know which of your numbers are conventions.
CN04-22	Hoyle's 1953 prediction of the ~7.68 MeV carbon resonance	A prediction is not a measurement. SIGNUM SIGNUM MANET: the prediction and the measurement travel as separate rows — the measurement is CN04-23 (Dunbar, Pixley, Wenzel & Whaling 1953), classed OBSERVED-REPLICATED on its own merits. Classing the prediction as if it were an observation would merge the two.
CN11-59	First-person testimony (CN-11)	Not a NATURA class at all — see below.
CN12-41	anima in silicio? / silicium ut aqua?	A register status, not an evidence class — see below.

CN11-59 — an HONEST signal. Not a NATURA class. Not calibrated. Ever.

The wing's only HONEST-class row is not a NATURA row wearing an odd label, and this amendment explicitly declines to give it a NATURA class. It belongs to the HONEST store, which is sovereign from the TRUE store and is never calibrated — that is the cardinal sin of this repo, and closing a vocabulary defect is not a licence to commit it.

Every class in the twelve above is a TRUE-side instrument: each is falsifiable, reproducible, and recalibratable, and each carries the falsifier that would move it. Testimony has none of those properties, and this is not a deficiency to be repaired. The row's own falsifier cell says it better than a rule could: "None. Testimony is never calibrated — that is the point, not a gap."

Forcing it into the TRUE-side vocabulary — even into an honest-looking fence like NOT-MEASURED — would assert that first-person report is a failed measurement rather than a different kind of signal. It would be a lane-crossing of exactly the kind the cardinal rule forbids, and it would be wrong in the direction that flatters the corpus, which is the direction to distrust most. It is carried, sovereign and uncalibrated, and it is not counted in the twelve.

CN12-41 — NONDUM FALSIFICABILIS is a register status, extra Arborem

NONDUM FALSIFICABILIS ("not yet falsifiable") is not an evidence class and never was. It is a register status: the mark of an item held in the QUAESTIONES APERTAE register, extra Arborem (outside the Tree) — leguntur, non aguntur: they are read, not acted upon.

The distinction is load-bearing. An evidence class says how well a claim about nature is supported. A register status says this item is not in the ledger at all. CN12-41 — anima in silicio? / silicium ut aqua? — offers no falsifier, and its source cell says why that is correct rather than defective: that is exactly why it lives in a register and not in a ledger. A ledger row without a falsifier would be a defect; a register entry without one is the register working.

Giving it an evidence class would import it into the Tree by relabelling — admitting through the vocabulary something that has not passed the admission tests. It stays extra Arborem, and it is not counted in the twelve.

---

Compound rows (the "disjoint" half of falsifier #2, answered honestly)

Two rows carry two classes, because two parts of the row have different evidentiary standing:

• CN05-31 — ε_MMR: MODELED (the ladder rung) · OBSERVED-REPLICATED (the overall band).
• CN06-30 — OBSERVED-SINGLE (the latency) · MODELED (its identification with T).

The original design assumed every row lands in exactly one class. These two do not, and no amount of vocabulary fixes that — the row has two parts, not the class system. Splitting them into separate rows would be the tidier fix and is deliberately not done: the parts are only meaningful together, and separating them invites a reader to cite the modeled half as if it were the measured half. That is the merge this wing exists to prevent.

The convention, registered: a compound row names each part's class to its part, in the class cell, in the order the parts appear. It is counted at its leading class (CN05-31 counts as MODELED; CN06-30 as OBSERVED-SINGLE) — the count is a convention for arithmetic, and the class cell, not the count, is authoritative for the row. Compounds are rare and must stay rare: two of 919. A compound is admissible only when the parts genuinely cannot be separated. If they can be, split them.

---

Nature as the authority — stated precisely, and its counterweight

The doctrine is defensible in exactly one form:

Nature is the authority because it is the only system that has already run the experiment — a very long parallel search under real physical constraints, in which the failures were deleted. Convergent evolution is evidence of a constraint-optimum.

And it is honest only with the counterweight, which travels with it everywhere in this wing:

Not every trait is an adaptation (Gould & Lewontin 1979, Proc. R. Soc. Lond. B 205(1161):581–598). Phylogenetic inertia, drift, developmental constraint, pleiotropy and historical contingency all produce features that are not optimal solutions to anything. Nature is full of frozen accidents: the vertebrate retina is wired backwards; the recurrent laryngeal nerve takes its detour.

Therefore: "nature does it this way" is a hypothesis generator, never a proof. A biomimetic design must still beat a tuned conventional baseline on a pre-registered metric, or it is recorded NEGATIVE — the wing inherits M7 (contains-baseline + load-bearing discriminator) unchanged. That paragraph is the entire difference between this wing and cargo-cult biomimicry.

The wing's own poster child polices this. Spiral phyllotaxis really does cluster near the golden angle, and Douady & Couder (1992) reproduced it in a physical ferrofluid-droplet experiment from simple repulsion dynamics — mechanism, not mysticism. It is earned. And in the largest sunflower study with explicit, checkable inclusion criteria, Swinton & Ochu (2016) found 18% of scored counts were non-Fibonacci — roughly one in five, in the organism the claim is named after. The regularity is real. The universality is not. Both facts are printed here, together, and the second one is the reason to trust the first.

---

The honesty rail

• RES IPSAE, NON SIMULACRA. No fabricated numbers. Every value carries a real citation, or it is written NOT-MEASURED. There is no third state. A plausible-sounding number with no source is the worst defect available in this wing. If you half-remember a value, search for it; if you cannot source it, write NOT-MEASURED.
• Every numeric claim carries value + units + scope + class + source + falsifier. A number without units or scope is a defect.
• SIGNUM SIGNUM MANET. The signal and the commentary about the signal travel separately: the measured value goes in the table, the interpretation goes in the prose. Never smuggle an interpretation into a value. (White & Seymour title their paper "…proportional to body mass^2/3"; the title is commentary, the CI is the signal. Read the CI.)
• Banned in author voice (quote-with-attribution only): verified, secure, guaranteed, certified, self-aware, conscious, AGI, intelligent. proven is reserved for UNI-ledger status and never describes nature's science. Use observed / measured / replicated / modeled / BOUNDED / PENDING / CONTESTED.
• TRUE signals (falsifiable, reproducible, recalibratable) and HONEST signals (lived experience, never calibrated) stay sovereign. 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 from one to the other.
• "Full human" and "the next evolution beyond human" are permanent open questions — QUAESTIO-APERTA. Never a target, never a milestone, never a deliverable.
• Observed / claimed / inferred / unknown stay visibly distinct.

---

The crosswalk (the two ledgers, side by side)

Each row pairs a UNI fence value with the NATURA class it superficially resembles — because the resemblance is the trap. The last column is the load-bearing one.

UNI fence (CLAIM-LEDGER)	Look-alike NATURA class	Why they are not the same	What this can never imply about the other
proven — a held, sealed, UNI-signed PASS (Class A/C), falsifier live	OBSERVED-REPLICATED	proven is about UNI's own artifact passing UNI's own gate. OBSERVED-REPLICATED is about nature, measured by third parties.	A replicated natural regularity can never make a UNI claim proven. A UNI proven row can never make a natural regularity replicated, or count as evidence about nature.
designed — a typed/signed spec exists, build partial or unverified	MODELED	designed is an intention about UNI. MODELED is a fitted account of nature, conditional on assumptions.	A model of nature that fits can never discharge a UNI design gate. A signed UNI design can never be evidence about how nature works.
hypothesized — a stated UNI mechanism, no sealed gate (Class U)	HYPOTHESIZED	Same word, different subject: one is UNI's open mechanism, one is nature's.	Nature's open mechanism can never move a UNI rung. UNI's open mechanism can never settle a scientific dispute.
not-yet-built — no engine, no run, no gate	NOT-MEASURED	not-yet-built is an absence in UNI's build. NOT-MEASURED is an absence in the literature.	UNI not having built it can never be evidence nature lacks the value. Nature lacking a measurement can never be a reason UNI cannot build — nor an excuse to invent the number.
(no UNI equivalent — the ledger's four states are PASS/FAIL/NEGATIVE/PENDING)	OBSERVED-CONTESTED	The UNI ledger has no contested state. Nature's live disputes are carried whole, both positions named.	A contested exponent can never be cited as settled to move a UNI row. UNI's four-state ledger can never flatten a live scientific dispute into one number.
NEGATIVE (ledger state) — a held UNI run that missed its pre-registered bar	INADMISSIBLE	A UNI NEGATIVE is a result UNI earned. INADMISSIBLE is a claim about nature that failed observation or was never falsifiable.	A nature claim's inadmissibility can never be recorded as a UNI NEGATIVE. A UNI NEGATIVE can never be a statement about nature.
NEGATIVE (ledger state) — (second look-alike, added 2026-07-15)	SUPERSEDED	Both describe something that lost — which is exactly why they are confusable. A UNI NEGATIVE is UNI's own run missing UNI's own pre-registered bar. SUPERSEDED is a published value about nature that its own field retracted, withdrew, or contradicted. Different subject, different authority, different repair.	A retracted natural value can never be recorded as a UNI NEGATIVE, or counted among UNI's published negatives. A UNI NEGATIVE can never supersede a value in the literature. This is why the NATURA class is named SUPERSEDED and not NEGATIVE: the token NEGATIVE belongs to the UNI ledger, and one token may not serve both vocabularies.
(no UNI equivalent)	OBSERVED-SINGLE	The UNI fence has no value for "built and measured once, never independently reproduced" — a UNI PASS is a held, sealed gate, not a lone reading.	A single unreplicated natural measurement can never move a UNI rung — least of all this class, which exists precisely to stop one study from being read as settled. A UNI PASS can never stand in for a missing replication in the literature.
(no UNI equivalent — not-yet-built is the nearest, and it is not close)	NOT-SOURCED · NOT-CONFIRMED · NOT-LOCATED	These record a gap in this corpus's homework, not in UNI's build and not in nature.	An untraced citation can never be evidence about UNI's build state in either direction. And none of the three may ever be read as NOT-MEASURED — "we could not find the source" is not "nature has not been asked." That merge would convert a research failure into a claim about the frontier of science.

---

The map of the wing

Reference chapters — encyclopedia/wing-NATURA/

Chapter	What it answers
NA-00 — How to read this wing	this chapter — the twelve classes (six as designed, six registered by amendment 2026-07-15), the cardinal rule, the crosswalk
NA-01 — Nature as the authority	why the already-run experiment is authoritative; convergence as constraint-optimum; the Gould & Lewontin counterweight, in full
NA-02 — The one loop	perceive → predict → act → update, as nature runs it and as the design method uses it
NA-03 — How to make new science	falsifiers, pre-registration, tuned baselines, discriminators, and how a negative gets published
NA-04 — Mind / body / world	the typed blanket split; interoception; why WORLD ⊥ BODY ⊥ MIND is a modelling commitment, not a metaphysics
NA-05 — The ratios	phyllotaxis and the golden angle (earned); allometry; what a ratio claim must carry to be admissible
NA-06 — The frequencies	Schumann, alpha, whole-body vibration, circadian; the earned/unearned split, with the receipts
NA-07 — The dimensionless numbers	Reynolds, Péclet, Damköhler and kin; why dimensionless groups travel across scale and ratios usually do not
NA-08 — Cell-level design	the cell as the design unit; budgets, transport, membranes, gradients
NA-09 — Morphogenesis	pattern from local rules; reaction–diffusion; growth as the generator of form
NA-10 — The scale ladder	molecule → cell → tissue → organism → colony → ecosystem, and what survives each step up

Recipes — cookbook/recipes-natura/ (the build catalog; each is a buildable recipe, each carries its own NATURA classes and its own falsifier)

Recipe	Subject	Recipe	Subject
CN-01	rock	CN-07	ant colony
CN-02	water	CN-08	dinosaur
CN-03	air	CN-09	whale
CN-04	star	CN-10	bat
CN-05	DNA	CN-11	human
CN-06	sperm	CN-12	the open question beyond — QUAESTIO-APERTA, never a plan

---

Reading order, if you are here to DESIGN something

1. NA-00 (this chapter) — the vocabulary and the cardinal rule. Non-skippable.
2. NA-03 — the method before the content, so you cannot mistake an inspiration for a result.
3. NA-01 — the authority doctrine and its counterweight, together, in that order.
4. NA-07, then NA-05 — dimensionless numbers before ratios. The order is deliberate: 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.
5. The scale chapter you need — NA-08 (cell), NA-09 (form), or NA-10 (ladder).
6. The CN recipe for your target — and only now.
7. 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 numbers (the ratio/frequency table)

Every row here is a worked example of a class, not a claim this chapter makes about nature.

Symbol	Value	Units	Scope	Class	Source	Falsifier
ψ	137.507764050038…	degrees	exact arithmetic: 360/φ², φ=(1+√5)/2	MODELED (derived)	definition	Not falsifiable — it is arithmetic. Falsify the plant claim instead (next row).
ψ_obs	clusters near 137.5	degrees	spiral-phyllotactic vascular plants; not universal	OBSERVED-REPLICATED	Douady & Couder (1992), Phys. Rev. Lett. 68:2098–2101, DOI 10.1103/PhysRevLett.68.2098	Exhibit a well-sampled spiral-phyllotactic taxon sitting stably away from 137.5° with no repulsion-dynamics account.
—	565/768 = 74% Fibonacci; 136/768 = 18% non-Fibonacci; 82% incl. predefined-type	dimensionless (fraction of scored counts)	657 Helianthus annuus seedheads; citizen-science; photoreviewed counts	OBSERVED-REPLICATED	Swinton & Ochu (2016), R. Soc. Open Sci. 3:160091, DOI 10.1098/rsos.160091	Re-sample under the same explicit inclusion criteria and recover ~100% Fibonacci → the non-Fibonacci fraction was a scoring artifact.
b	≈ 0.74 (Brody, same year: ≈ 0.73)	dimensionless (BMR ∝ M^b)	~13 spp mammals + birds, 1932 dataset	OBSERVED-CONTESTED	Kleiber (1932), Hilgardia 6:315–353	See the dispute row below.
b_WBE	3/4 = 0.75 exactly	dimensionless	conditional on: space-filling fractal branching network + size-invariant terminal units + minimised dissipation	MODELED	West, Brown & Enquist (1997), Science 276(5309):122–126, DOI 10.1126/science.276.5309.122	Exhibit a taxon meeting all three assumptions whose measured b excludes 0.75. Drop any assumption and 3/4 does not follow.
b_WS	0.686 ± 0.014 (95% CI); interordinal 0.65	dimensionless	571 mammal spp; BMR normalised to 36.2 °C; excl. Artiodactyla, Macropodidae, Lagomorpha, Soricidae; interspecific ≈0.69, N=619, r²=0.94	OBSERVED-CONTESTED	White & Seymour (2003), PNAS 100(7):4046–4049, DOI 10.1073/pnas.0436428100; Dodds, Rothman & Weitz (2001), J. Theor. Biol. 209(1):9–27	Pre-register taxon set + temperature normalisation + regression unit, and show the CI is stable across all three. It currently is not — that is the dispute.
f₁	≈ 7.83	Hz	Earth–ionosphere cavity, global; observed peaks are wide	OBSERVED-REPLICATED	Schumann (1952) predicted; Balser & Wagner (1960), Nature 188:638–641	Measure the cavity with calibrated ELF magnetometers and find no peak near 7.8 Hz.
f₁ range	7.5–8.1 (mode 1); station/component specific (e.g. 7.2–8.2 B_EW; 7.6–7.9 B_NS)	Hz	per-station, per-component — not a planetary constant	OBSERVED-REPLICATED (as station data)	station ELF records; see NA-06	Show a single global diurnal curve reproducing all stations.
f₁ global diurnal constant	—	—	—	NOT-MEASURED	no single global figure found by this chapter's search; only station-specific ranges	Publish a station-independent global diurnal constant with its CI.
f_ideal	≈ 7.49	Hz	ideal lossless cavity: c ÷ Earth circumference (≈3.00×10⁵ km/s ÷ 4.0075×10⁴ km)	MODELED	textbook derivation	The ~0.34 Hz gap to 7.83 is the model's known error (finite ionospheric conductivity lowers propagation speed), not a mystery.
B_SR	picotesla (pT) range	T	ELF background amplitudes	OBSERVED-REPLICATED	Balser & Wagner (1960), Nature 188:638–641	Calibrated ELF magnetometry returning nanotesla-scale SR amplitudes.
SR → human physiology	—	—	ambient (pT) field strengths	NOT-MEASURED	no pre-registered, independently replicated dose–response found by this chapter's search	Pre-register and independently replicate a dose–response at ambient pT strengths.
f_α	8–13 (≈10 dominant)	Hz	human posterior cortex; relaxed wakefulness; eyes closed	OBSERVED-REPLICATED	Berger (1929), Über das Elektrenkephalogramm des Menschen; IFCN definition	Record posterior EEG, eyes closed, relaxed, and find no 8–13 Hz rhythm.
f_WBV	4–8 vertical; 1–2 horizontal; 4–16 hand/arm	Hz	human whole-body mechanical vibration sensitivity	OBSERVED-REPLICATED	ISO 2631-1:1997	Standard psychophysical protocol returning a flat frequency response.
A4	440 ± 0.5	Hz	Western concert pitch, by convention, at ~20 °C	OBSERVED-REPLICATED (as a standard, not a natural constant)	ISO 16:1975, Acoustics — Standard tuning frequency	It is a convention; only the claim that it is natural is falsifiable, and that claim is inadmissible.
A4 = 432 "natural tuning"	—	—	—	INADMISSIBLE	names no measurand distinguishing it from any other convention	Name a measurable physical or biological quantity that differs between 440 and 432 Hz, and pre-register it.
chakra frequency tables	—	—	—	INADMISSIBLE as physics	no measurand defined	May be recorded as an HONEST/cultural signal. Never as a TRUE/measured one.

Read the frequency rows together. ≈7.83 Hz (Schumann, electromagnetic, planetary cavity), 8–13 Hz (alpha, electrical, cortical), and 4–8 Hz (whole-body vibration, mechanical) are three real, separately measured quantities that happen to land in the same numeric neighbourhood. They are different physics with different measurands. The numeric proximity is a coincidence of units, and treating it as a mechanism is precisely the move this wing exists to catch. Each value is earned; the bridge between them is not.

---

Amendment record (2026-07-15)

This is a calibrate-down event on NA-00. The chapter overclaimed; the corpus measured it; the wording moves down to the measured value and the strike is carried in the open. Calibration in this repo only ever moves wording down to what was observed, and the prior wording stays on record — the strike above, and this entry. A reader who is told only the corrected number has been handed a conclusion; a reader who is shown the strike has been handed the evidence.

AMENDMENT 2026-07-15-A — the six-class vocabulary was incomplete

What NA-00 declared	"This wing uses six classes and only six." Falsifier #2 pre-registered the refutation: "The six classes are not exhaustive or not disjoint. Exhibit a real nature claim in the NA/CN chapters that cannot be placed in exactly one of the six."
What the corpus observed	72 of 919 ledger rows (7.8%) carried a class string outside the six. Not one counter-example — seventy-two, written by the chapters themselves, before anyone went looking. Breakdown as extracted: OBSERVED (bare) 42 · NOT-SOURCED 20 · NEGATIVE 3 · HONEST-class 1 · MODELED-CONTESTED 1 · NOT-CONFIRMED 1 · NOT-LOCATED 1 · NONDUM FALSIFICABILIS 1 · no class carried 1 · no evidence class (definitional) 1.
Verdict	Falsifier #2 FIRED, on both halves. Not exhaustive: 42 rows were measured-but-never-replicated and fit none of the six. Not disjoint: MODELED and OBSERVED-CONTESTED overlap (CN12-31), and two rows carry two classes because two parts of the row differ in standing. The six-class design is refuted as stated.
Who found it	The chapters did — not a reviewer, not an audit. CN-06 and CN-07 both recorded the gap explicitly while writing, and CN-07 stated the reason exactly: "OBSERVED-REPLICATED asserts a replication that did not happen, and OBSERVED-CONTESTED asserts a dispute that does not exist. Both are false, in opposite directions." Both flagged the fix as belonging at the ledger; both noted NATURE-LEDGER.md did not exist yet, which is why the gap stayed open. It exists now, and it closed the gap it was built to catch.
What changed	Six classes → twelve, in three groups (measured / derived / fenced). Registered: OBSERVED-SINGLE, MODELED-CONTESTED, SUPERSEDED, NOT-SOURCED, NOT-CONFIRMED, NOT-LOCATED. Four rows registered as deliberately unclassed; the compound-row convention registered.
What did NOT change	No value, no source, no falsifier, in any row, in either artifact — machine-checked across all 919 rows: of 10 columns, only the class column differs, on 48 rows. This was a re-class only — relabelling rows to the class that was always true of them. Not one row changed evidentiary standing, and not one row left the six. The six-class subtotal was 847 as the ledger published it (§5.1) and is 847 now. The two artifacts were re-classed in lockstep and still agree on all 443 shared classed entries; every ledger_row_id is unchanged and in its original order.
The one counting subtlety, disclosed	Under a strict leading-token count, the old text yields 846, not 847 — because CN05-31's class cell read "rung MODELED; overall band OBSERVED-REPLICATED", which does not begin with a class token and so does not parse mechanically. The published §5.1 counted it at MODELED anyway (that is where its 251 came from), which is why the published subtotal was 847. This amendment re-ordered that cell to "MODELED (the ladder rung); OBSERVED-REPLICATED (the overall band)" so it now parses as what it was always counted as. The row's value, source, falsifier and evidentiary standing are untouched; only the cell's word order moved. This is recorded because a subtotal that appears stable at 847 under one rule and shifts 846→847 under another must have the difference named, or the stability is an illusion the reader cannot audit.
Verified how	tools/verify_class_vocabulary.py — the operable form of falsifier #2, written before the amendment and run against both artifacts. Pre-amendment it reported 47 ledger + 24 K20 rows needing re-labelling; post-amendment it reports 0 out of vocabulary, 919/919 rows placed. Re-run it after any edit to either artifact. ⚠ READ WITH AMENDMENT 2026-07-15-C: "both artifacts" was the whole of the scope, and the chapters were not in it. This cell's "0 out of vocabulary" was true of the ledger and K20 and false of the wing — 52 chapter rows still carried pre-amendment strings. The wording stands as written, struck, because it is the evidence of how the gap read from inside.
Standing	This amendment is itself falsifiable. It is not a claim that twelve is correct — only that twelve is what 919 rows measured. Thirteen is a finding, not a failure.

What this event is evidence of, stated without flattery. A chapter made a clean, checkable declaration; the work refuted it; the refutation was recorded by the refuting chapters rather than hidden; and the fix waited for an instrument that could count. That is the method working — and it is not a result about nature, not a rung, and not evidence of anything about UNI. The program position is untouched: ~2 of 11+ rungs earned. A vocabulary that can now name its own gaps is a vocabulary, not an achievement.

AMENDMENT 2026-07-15-C — the amendment's scope was narrower than this record read

A second calibrate-down, on the record above rather than on the vocabulary. Amendment 2026-07-15-A is not withdrawn: twelve classes still stand, and no row's standing moves. What was wrong was the scope, and the fact that nothing in the record said so.

What 2026-07-15-A actually closed	NATURE-LEDGER.md and K20. Those two files, and nothing else. The record above says so if read closely — "machine-checked across all 919 rows", "both artifacts", "either artifact" — but it never says what is outside those words, and every reader who took "0 out of vocabulary" as a statement about the wing read it the way it was written to be read.
What was still broken	52 rows in 10 of the 23 chapters still carried PRE-amendment class strings — the chapters' own ## The numbers tables, i.e. the artifacts the ledger extracts, and the artifacts a reader actually reads. Breakdown: bare/qualified OBSERVED 42 · NEGATIVE 3 · as above 5 · CN-05 cells not leading with a token 2. The ledger had been re-classed away from its own sources, so for one day the extraction and the extracted disagreed in the one column the amendment existed to fix.
Why the verifier said PASS anyway	It never opened a chapter. tools/verify_class_vocabulary.py checked the ledger and K20 — the artifacts it was written beside — and reported 0 while the real chapter residue was 52. Its scope was correct for what it checked and silent about what it did not, which is the defect: a verifier that does not print its scope cannot be audited, and a PASS from an unstated scope is a false negative wearing a green result.
What changed now	(1) All 52 chapter cells re-classed in lockstep with the ledger — the ledger already carried the correct class and the reasoning for each, and each cell was mirrored from it, never invented. Only the class token moved: the chapters' own qualifier text was kept verbatim rather than overwritten with the ledger's shorter restatement of it. as above was resolved to the literal cell of the row it points at (NA10-30). No value, no scope, no source, no falsifier was touched — machine-asserted per row, per column. (2) The verifier now scans the chapters as a third scope, counted on its own line, and prints every scope it covers before it prints any result. (3) It now also checks that each chapter's table still aligns row-for-row with the ledger, so ledger/chapter divergence is itself a finding.
The tokenizer moved a number, so it is declared	The verifier's lead_token() only stripped ** when a cell was bold end to end. Chapter cells routinely bold the token and leave the qualifier plain (**MODELED** (as above; unreplicated)), so on first scan the chapters reported 90 out-of-vocabulary rows: 52 real + 38 false positives whose class was never in doubt. Emphasis is now stripped before the leading-token match — safe by construction, since every registered token is [A-Z-] only. 90 → 52, and 52 is the number fixed against. A count that moves because the instrument changed is not a count until the instrument is named.
Verified how	Three positive controls, because a scope that reports zero is worthless until it is shown able to report non-zero: (1) a bare OBSERVED re-injected into a CN-09 class cell → caught as CN09-02, exit 1; (2) NEGATIVE re-injected into CN-01 → caught as CN01-34, exit 1; (3) a chapter table row deleted → caught as a ledger/chapter misalignment, exit 1. Restored, the run is 0 in every scope, exit 0. The chapters' class counts now equal the ledger's exactly (918 = 919 − NA03-16, which lives in NA-03's prose) — an independent check that the mirror is a mirror.
Standing	This does not raise anything. It closes a scope gap that should never have shipped, found by an outside audit and not by this wing.
50 or 52? Both, and the difference is the instrument	The audit reported 50 rows across 9 files; this amendment re-classed 52 across 10. Neither is a typo, and the gap is fully accounted: 50 is reader/build.py's build-note count, which finds a registered class anywhere in a cell; 52 is the leading-token count this wing's verifier and §5 actually use. Reconciled exactly: 52 − 5 + 3 = 50. The reader misses 5 real defects because each names a class somewhere but not first — CN05-06, CN05-31 (the two §3 already names), CN06-30 ("…the latency; MODELED…"), and CN08-05 / CN08-24 (whose qualifier reads "downgraded from OBSERVED-REPLICATED", so the cell cites the very class it was demoted from). The reader adds 3 that are not defects at all — CN04-22, CN11-59, CN12-41, deliberately unclassed and registered as such. Two instruments, two answers, one corpus: the number is not a fact until the rule that produced it is named. The 52 stands because a class cell must lead with its class — a row whose class must be inferred from a qualifier is exactly the ambiguity the amendment re-ordered CN05-31 to remove.

The honest residue — what these amendments did NOT fix. Two lines, not one.

1. The 22 untraced rows. The NOT-SOURCED (20), NOT-CONFIRMED (1), and NOT-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.
2. The 52 chapter rows, and the day they were wrong. Amendment 2026-07-15-A closed the ledger and K20 only; 52 chapter rows across 10 files carried the old class strings until AMENDMENT 2026-07-15-C re-classed them. They are labelled correctly now. What is not repaired is the reason it happened: the amendment was written against the artifacts its verifier could see, the verifier's scope was never printed, and 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. The only defence is a scope that prints itself, and it now does.

---

Falsifier (operable)

This chapter is refuted if any of the following is shown:

1. A merge is legitimate. Exhibit one sound inference in which a NATURA-class row raises, lowers, or discharges a row in CLAIM-LEDGER.md (or the reverse) without a separate UNI gate. If such an inference exists, the cardinal rule is wrong and this wing is mis-architected.
2. The classes are not exhaustive or not disjoint. Exhibit a real nature claim in the NA/CN chapters that cannot be placed in one of the registered classes.

   THIS FALSIFIER HAS FIRED — 2026-07-15, on 72 rows, on both halves. See § Amendment record. It is left standing and unweakened, against the amended twelve, because a falsifier that fired once is the most likely one to fire again — and it is now mechanised as tools/verify_class_vocabulary.py, which fails on any row outside the registered vocabulary. The count of classes is not part of the claim; the claim is that every row can be honestly placed. A thirteenth class is a finding, and the amendment procedure above is how it gets registered.

3. A crosswalk "can never" is too strong. Exhibit a case where one of the six non-implications in the crosswalk table blocks a valid inference.
4. A number here is unsourced or wrong. Any value in the table that cannot be traced to its cited source, or that the source does not support, is a defect in this chapter — report it and the row is corrected or demoted to NOT-MEASURED.
5. NOT-MEASURED was premature. Produce the pre-registered, replicated measurement for either NOT-MEASURED row. The row is then reclassified — which is the system working, not failing.

---

Recorded INADMISSIBLE / NEGATIVE (first-class, inline)

• "The golden ratio is a universal design law of nature." — INADMISSIBLE as stated. Unfalsifiable and cherry-picked: it names no measurand, no scope, and no observation that could refute it. Receipt: the mechanism-bearing version of the claim is real and earned — Douady & Couder (1992) produced golden-angle divergence in a physical ferrofluid experiment from repulsion dynamics alone. And the universality version fails on its own poster child: 18% of scored sunflower counts are non-Fibonacci (Swinton & Ochu 2016). Honour the mechanism; reject the universal law. This pair is the template for every mystical-adjacent item in this wing.
• "432 Hz is nature's true tuning." — INADMISSIBLE. A4 = 440 Hz is a standards convention (ISO 16:1975, 440 ± 0.5 Hz), not a natural constant, and 432 Hz is a different convention. The claim names no measurand.
• "Schumann 7.83 Hz entrains the human brain because alpha is ~8–13 Hz." — INADMISSIBLE as stated. Both frequencies are real and separately replicated. The bridge names no mechanism, no dose–response, and no falsifier, and the SR field is ambient at picotesla strength — the effect size is NOT-MEASURED (above). If the claim is re-stated with a measurand and a pre-registered dose–response, it is admissible and becomes a HYPOTHESIZED row. It is not there yet.
• Chakra frequency tables — INADMISSIBLE as physics. No measurand. Recordable as an HONEST / cultural signal, sovereign and uncalibrated, never crossed into the TRUE store.
• NEGATIVE, carried standing: "nature does it this way" has never been sufficient. Every biomimetic design in the CN recipes that fails to beat a tuned conventional baseline on its pre-registered metric is recorded NEGATIVE and printed — not retried until it wins.

---

HONEST FENCE — NOT-MEASURED

This chapter measures nothing. It is a constitution: definitional, governance-class, and it asserts no regularity of nature in its own voice. Every number above is a worked example carrying its own class and its own citation, and none of them is claimed here — they are demonstrations of the vocabulary, borrowed from their sources and returned to them.

NOT-MEASURED is the correct and honest fence for a chapter with no measurand, and it is recorded rather than dodged. The UNI-side analogue (status: method, fenced proven-as-governance in CB-01) is not available here and is deliberately not borrowed: proven is reserved for the UNI ledger, and this wing does not get to use it — including for itself.

---

Not claimed

This chapter does not assert:

• that any UNI claim is proven, designed, or at any rung — it makes no UNI claim at all, and the program position (~2 of 11+ rungs earned) is untouched;
• that nature is optimal, that convergence proves optimality, or that any observed regularity is an adaptation — Gould & Lewontin (1979) travels with the doctrine permanently;
• that the golden angle, the Schumann resonance, alpha, or any other frequency has significance beyond the scope printed in its row;
• that any biomimetic design works — no design has been benchmarked in this chapter;
• that the twelve classes are the only defensible taxonomy, or that twelve is a settled or correct number — only that they are this wing's, that twelve is what 919 rows measured rather than a design target, and that they are never merged with the UNI four. The chapter has already been wrong about this once (it declared six; see § Amendment record), which is the strongest available reason not to claim the new number is final;
• that re-classing the 72 rows repaired them. 22 rows (NOT-SOURCED · NOT-CONFIRMED · NOT-LOCATED) are now correctly labelled as un-traced, which is a description of a defect, not its repair;
• anything whatever about "full human" or "the next evolution beyond human." Both are QUAESTIO-APERTA — permanent open questions, never targets, never milestones, never deliverables. CN-12 is a chapter about a question, not a roadmap to an outcome.

Falsify any row. The two NOT-MEASURED rows are invitations, and the INADMISSIBLE rows are held open for anyone who can re-state them with a measurand.