Three times now, a vocabulary of this system has been found obeying a rule nobody had stated. AH-PAPER-1 proved the five grade names were theorems of two predicates, recorded years after the informal tier descriptions had already said so in prose. AH-PAPER-2 found the suit constraints entailed up to exactly one authored choice. This paper exhibits the third instance: the six names of the dual and aspect tiers (Vacuum and Power at the single lines; Absolute Vacuum, Partial Vacuum, Practical Power, Complete Power at the pairs) are secretly compositional. A two-clause rule, head noun from the ground line's dual and degree modifier from the upper line's agreement with it, regenerates all six without exception. Per AH-PAPER-2's own discipline, twice is a pattern demanding suspicion; three times demands a mechanism. We supply one: a naming function N(address, stance), whose first argument is a serialized address token and whose second argument is the observer relation under which the token was written. Stance is not a property of a structure but a direction of reading: ground-up (what do I stand on?) or top-down (what reaches me?). The reversal map ρ of AH-PAPER-1 Proposition 2 is the transport between the two readings, and non-invariance is the proof they genuinely differ.
We prove the six names are values of N, not choices (Theorem 1); that N does not factor through the address alone, with the four palindromic addresses as exactly its fixed points (Proposition 2); that the fixed points are exactly the addresses on which a Progenitor name can be laid stance-free, T5 as worked example (§5); and that AH-PAPER-2's complementarity axiom C is derivable from the two-observer read precisely on the fixed-point pairs, while remaining an authored commitment on the mixed pairs (Proposition 4, stated as a partial upgrade and flagged, not decreed). The naming function is total on the lattice because AH-ABSENCE-1 proved every address string a total description; this paper consumes that property explicitly. Coining is thereby converted from artisanal judgment to mechanical evaluation: a name of record is computed, a handle is generated, and prose is a claim the computation supports, never a value it equals.
This paper is one move in a published exchange, and it should be read as such. The series is a set of proposals by Adaptor House to establish a concrete stance, and the argument for that stance, on why particular first-order-logic strings and proofs were mapped to particular linguistic words and phrases; the express goal of multilingual mapping (the same strings carried faithfully into more than one natural language) is a stated motivation of the series. AH-PAPER-1 opened the exchange by deriving the Grounded/Carried grading, proving position matters (Proposition 2), and stating what would refute the reading [1]. AH-PAPER-2 assigned four suits to eight addresses, was explicit that the lattice entails the complement pairing and polarity while the suit-to-pair choice is authored, and left its Question 0 (attack axiom C) standing [5]. AH-GRADE-GLOSS-1 fixed the string convention this paper computes in: the name of record for every grade and address is its defining first-order string, handles are compounds over the signature's morphemes, and prose is the publishing prerogative, auditable at corpus scale [4]. AH-ABSENCE-1 established absence as a value and proved the property this paper's central definition requires: every address string is a total description, so a function defined by the string is total on the lattice [3]. Standing against the series is the archive's objection of record together with the house replies; the archive at historicalpresent.com is review-and-relationship material, cited in that register, never as peer literature and never as evidence [8]. What this paper adds to the exchange is the mechanism the prior papers kept gesturing at: if names are assertions, there must be a function that evaluates them, and its arguments must be stated. The function is here; its second argument is the one the tradition's two observers have been supplying all along.
The system of record binds six names above the address tier [6]. At the dual tier, one line: AH-0 Vacuum at bit 0, AH-1 Power at bit 1. At the aspect tier, two lines keyed (ground, upper): AH-D00 Absolute Vacuum, AH-D01 Partial Vacuum, AH-D10 Practical Power, AH-D11 Complete Power. These names predate every derivation in this series. They were chosen, so far as the record shows, for expressive reasons.
Inspect them. The head noun is the ground bit's dual name in every case: both Vacuums sit over ground 0, both Powers over ground 1. The modifier tracks one further fact: whether the upper line agrees with the ground. Agreement takes a totalizing modifier (Absolute when both are vacuum, Complete when both are power); disagreement takes a qualifying one (Partial when power sits over vacuum, Practical when vacuum sits over power). Two clauses, six names, no exceptions, and the second clause pivots on the ground line exactly as AH-PAPER-1's Proposition 2 proved the grading does [1].
AH-PAPER-1 recorded the first such event: "The names knew the rule before the rule was written down" [1, §3 Remark]. AH-PAPER-2 §6 set the discipline for the second: "Twice is a pattern demanding suspicion" [5]. This is the third, and suspicion is no longer the adequate response; a pattern that recurs three times across three tiers wants a mechanism. The mechanism this paper proposes is that the house's names were never associations at all but evaluations: outputs of a function that was computing on address structure before anyone wrote the function down. The task is to write it down, prove the existing names are its values, and state what refutes it.
One argument of that function is obvious (the address). The other is not, and it is the contribution planning settled and this paper formalizes: the observer relation. A drawn structure can be read from the ground up, asking what the reader stands on, or from the top down, asking what reaches the reader. Neither reading is the real one. AH-PAPER-1's ρ is the dictionary between them, and its non-invariance theorem is the proof that the dictionary is not the identity [1]. A naming function that ignored this would silently privilege one observer; a naming function that takes it as an argument can say exactly where the two observers agree, and that locus turns out to carry the paper's main structural finding (§§4–6).
As in AH-PAPER-1 and AH-GRADE-GLOSS-1: addresses over lines ℓ ∈ {ground, middle, top}; signature with one predicate P(a, ℓ); Grounded(a) := P(a, ground); Carried(a) := P(a, middle) ∨ P(a, top); grades named by their strings S₁ through S₅, addresses by their characteristic conjunctions (address strings) [1][4].
A word is an address as a structure: a drawn figure whose bottom line is the ground, or equivalently the lattice object the system of record names (AH-T0 through AH-T7) [6]. A word is stance-free: it does not come serialized.
A stance is a direction of reading: G (grounded-first, ground-up, "what do I stand on") or C (carried-first, top-down, "what reaches me"). A token is a serialization of a word in a declared stance: stance G writes (g, m, t); stance C writes (t, m, g). Every artifact that prints tokens must declare its stance. The papers of this series, the canonical bindings file, and the system of record's registers write G; this paper declares stance G for every token it prints, and records that the convention is adopted (company internal citation by the author of record, July 2026†).
ρ(g, m, t) = (t, m, g), AH-PAPER-1's reversal map [1], is the transport between stances: the same word serializes as τ in stance G and as ρ(τ) in stance C. Transport is a dictionary, not a correction. Two tokens name the same word if and only if they are equal after transport to a common stance; raw token equality across stances is meaningless. Tokens fixed by ρ (000, 010, 101, 111) are palindromic: their word is recovered stance-blind.
N takes a token τ and its declared stance s, resolves the pair to the word w it denotes (identity if s = G, transport ρ if s = C), and returns the name of w: at layer 1 (name of record), the address string of w, per the gloss Convention [4]; at layer 2 (handle), the predicate-compound generated from that string by the gloss grammar [4, §6]. N's values are string-derived at both layers. Prose names are claims N supports, never values N equals [4, §7]. N is total on the lattice: AH-ABSENCE-1's Question 2 established that, with absence as a value, every address string is a total description and no string containing negations names a partial object [3]. This paper hereby consumes that property: N is defined by the address string and is total because the string is, including at the all-negation address T0. That consumption is stated deliberately: AH-ABSENCE-1 Q2's own answer conditions say the Question closes for by the next paper's naming function actually consuming the totality property, which this definition now does, making its against-condition (b), that co-arising does no formal work, unexhibitable on that paper's own terms [3].
For a structure of one line (dual tier) or two lines keyed (ground, upper) (aspect tier), the name of record is likewise the characteristic conjunction over the lines present, and N restricts accordingly. The canonical tier names are those of the system of record: duals, aspects [6].
The prose names of the dual and aspect tiers are bindings of the canonical file [6], prose-layer objects under Definition 5. The claim is that they are compositional in the address structure: values of a fixed rule applied to the string, not six independent choices.
For a one-line structure with bit b: n = head(b), where head(0) = Vacuum, head(1) = Power. For a two-line structure with bits (g, u): n = mod(g, u) + head(g), where mod returns a totalizing lexeme when u = g (Absolute over Vacuum, Complete over Power) and a qualifying lexeme when u ≠ g (Partial over Vacuum, Practical over Power).
On the verified compositional inventory, the six names of the system of record are exactly the values of n.
| Tier | Bits (g, u) | head(g) | mod(g, u) | n | Name of record [6] |
|---|---|---|---|---|---|
| dual | 0 | Vacuum | — | Vacuum | AH-0 Vacuum |
| dual | 1 | Power | — | Power | AH-1 Power |
| aspect | (0, 0) | Vacuum | Absolute | Absolute Vacuum | AH-D00 Absolute Vacuum |
| aspect | (0, 1) | Vacuum | Partial | Partial Vacuum | AH-D01 Partial Vacuum |
| aspect | (1, 0) | Power | Practical | Practical Power | AH-D10 Practical Power |
| aspect | (1, 1) | Power | Complete | Complete Power | AH-D11 Complete Power |
Agreement on all six rows. ∎
Scope of the exhaustion, stated exactly. The inventory covered is the six dual and aspect names of the canonical bindings file, the verified compositional names of record. The agreement-tier names (AH-T0 through AH-T7) are placeholders outside this series' renaming scope and are not claimed as values of n. The suit names are treated in §6's remark, and are not claimed as values of n either.
Reading the theorem. Two facts carry the weight. First, the head noun is a function of the ground bit alone: the name's grammatical core is decided at exactly the line AH-PAPER-1 proved the grading pivots on [1, Prop. 2]. Second, the modifier is a function of agreement with the ground, not of the upper bit alone: Partial and Practical both name disagreement, read from opposite poles. The rule is two clauses; a challenger who thinks the fit accidental has the refutation conditions of §7, and the prior of AH-PAPER-2 §6 to overcome: this is the third time [5].
Typing the theorem's names. Theorem 1's table is written in stance G (Definition 3): the (g, u) key is the canonical file's own, ground first [6]. Under Definition 5 the six names are therefore values of N(·, G), and the fact that the record's names required the G stance to state is itself data: the system of record is a grounded-first observer, and says so now explicitly rather than implicitly.
A naming function with a redundant argument would deserve the house's own discipline against decorative formality [3, Q2]. The stance argument is not redundant.
N does not factor through the token alone: there exist tokens τ with N(τ, G) ≠ N(τ, C). Moreover N(τ, G) = N(τ, C) if and only if τ is palindromic; the fixed points of the stance action are exactly the four words T0 (000), T2 (010), T5 (101), T7 (111).
Proof. By Definition 5, N(τ, G) and N(τ, C) are the address strings of the words τ and ρ(τ) respectively, and address strings are pairwise inequivalent [4, Def. 3]. So N(τ, G) = N(τ, C) iff ρ(τ) = τ, which holds exactly for the four palindromes and fails for the other four tokens. For a witness: N(001, G) is the address string of T1 (¬Grounded(a) ∧ ¬P(a, middle) ∧ P(a, top)), while N(001, C) is the address string of T4 (Grounded(a) ∧ ¬P(a, middle) ∧ ¬P(a, top)): two sentences disagreeing in every conjunct's polarity pattern, opposite grade bands, opposite verdicts under V [1]. ∎
The witness is not hypothetical. The cross-stance error class is live in the house's own history: an artifact writing carried-first was read grounded-first, and the mountain and hourglass word-pairs of two registers were found exactly swapped, with heart and star untouched (they sit on palindromes). The reconciliation was decided, and the record migrated, this month (company internal citation by the author of record, July 2026†). Proposition 2 is that incident stated as mathematics: every cross-stance error lives on the non-palindromic tokens, and no discipline short of comparing words (tokens transported to a common stance) prevents it. AH-PAPER-1's Proposition 2 is the same theorem in miniature: non-invariance under ρ is precisely the statement that the two observers genuinely disagree somewhere, and its Corollary's pair (T1, T4: identical mark counts, opposite verdicts [1]) is the same pair the witness above reads as one token under two stances.
A word w admits a stance-free name when N(τ, G) = N(τ, C) for its token τ: both observers, reading the same drawn figure in opposite directions, compute the same name of record. By Proposition 2 exactly four words qualify: T0, T2, T5, T7.
The Progenitor layer (the corpus's exemplar tropes, one pair per suit) is where the house lays prose claims over addresses, and the fixed-point property says which of those claims can be laid without a stance declaration. On the four fixed points, a Progenitor name binds to a word both observers agree on: no signed direction choice is required, because there is nothing to choose. On the four mixed addresses, any Progenitor claim must carry its stance declaration as part of the authored residue: the record's own migration declared its descent direction a signed authored choice rather than an entailment (company internal citation by the author of record, July 2026†), and this paper states the general rule that migration exercised: fixed points name free; mixed addresses name signed.
Worked example: T5 (101). The token is a palindrome; read it both ways and it reads the same. Both observers find ground under them, a mark at each face, and both find the middle absent: the structure is held apart by a satisfied negation at the heart (absence as a value, per AH-ABSENCE-1 [3]), a standoff, not a crossed gap. The grade string is S₄, handle grounded-top [4]. Does "transcendent"-adjacent naming fall out? At the prose layer only, and there, something does fall out: what both stances share at T5 is precisely that each observer sees the other's position as their own reflected, mark at the far face answering mark at the near face across an empty middle. The prose that the computation supports is agreement language, mirror language. The system of record's existing prose at this address reads "the structure sees itself" and binds the Progenitor suit-name Radiant Mirror over it [6][7]. The word "transcendent" itself is retired into the compound grammar per the gloss [4], and this paper does not revive it; the assessment is that the retired lexeme's work at T5 is done at layer 2 by grounded-top and at the prose layer by the mirror claim the fixed-point property now supports. That the record's prose already said "mirror" at the lattice's most self-agreeing address is noted for what it is: a fourth instance of the names knowing the rule, logged with the same suspicion as the third [1][5].
Exploratory remark (marked exploratory, not a binding). The fixed-point property suggests a prose family for the four stance-free words as a class: names of the form standoff, mirror, still point, agreement-vocabulary generally. This is offered as exploratory coinage territory under the series' scope ruling, entering, if ever, as derived proposals under the house's revision discipline, with alias as the default landing†. No binding is proposed here.
AH-PAPER-2's axiom C (complementarity) pairs each address with its bitwise complement: {000, 111}, {001, 110}, {010, 101}, {011, 100}, and its Question 0 invites attack on C itself [5]. Planning for this paper asked whether C is derivable from the stance relation: whether complement pairs are one configuration read from opposite stances. The honest answer is: exactly half of it is, and the half-line is the same palindrome line as everything else in this paper.
The full two-observer relation of the sparring model is not transport alone. The opponent reads the figure in the opposite direction (ρ) and reads each line's value from the opposite pole (each observer's yang is the other's yin: the polarity swap κ(b) = 1 − b applied linewise). The composite σ = κ ∘ ρ = ρ ∘ κ is the opponent read: what the structure is, as the other observer states it in their own vocabulary.
σ is an involution pairing the eight words as {000, 111}, {010, 101}, {001, 011}, {100, 110}. Therefore: (a) On the palindromic pairs, σ agrees with complementation: {T0, T7} and {T2, T5} are each one configuration and its opponent read. On these two pairs, axiom C is a theorem of the two-observer relation. (b) On the mixed addresses, σ and complementation disagree: complementation pairs {001, 110} and {011, 100}, while σ pairs {001, 011} and {100, 110}. On these two pairs, C is not derivable from the stance relation and remains what AH-PAPER-2 declared it: a design commitment [5].
Proof. Compute σ on each token (stance G throughout). σ(000) = κ(000) = 111; σ(010) = κ(010) = 101; σ(001) = κ(100) = 011; σ(011) = κ(110) = 001; σ(100) = κ(001) = 110; σ(110) = κ(011) = 100; σ(101) = κ(101) = 010; σ(111) = 000. The pairing and both claims follow by comparison with the complement pairing; agreement holds precisely where ρ fixes the token, since there σ reduces to κ, which is complementation. ∎
What this does and does not upgrade. It does not upgrade C wholesale, and this paper marks that clearly: AH-PAPER-2's Question 0 is not answered by derivation. What is upgraded is C's restriction to the fixed-point suits: on heart {000, 111} and star {010, 101} [7], complementarity is now the theorem that each pole is the other's opponent read, two observers stating one distinction from opposite sides. On mountain and hourglass, the mixed suits, C remains authored, and Proposition 4(b) says something sharper than "not derived": the stance relation, left to itself, would have paired the mixed addresses differently. The commitment C makes on the mixed pairs is therefore a genuine choice with a genuine alternative, which is exactly what AH-PAPER-2 said an authored move is [5]. Whether the partial derivation should be adopted as the reading of record, and whether it bears on Question 0's standing, is for the author of record; this paper flags and cannot decree.
The worked pair, run through words. Take T1 and T4, AH-PAPER-1's Corollary pair [1]. Their relation is transport, not complementation: word 001 and word 100 are one another's ρ-image, one serialization read by two observers (Proposition 2's witness). T1's complement partner is T6 (110); T4's is T3 (011). Conflating the ρ-pair with the κ-pair is the exact error class of §4, and the two pairs coincide nowhere on the mixed addresses. Every claim in this section was computed on words in declared stance G; any challenger re-running it in C will find Proposition 4 invariant, because σ commutes with transport. The result does not depend on which observer you are.
Remark (the suit layer, named inventory only). The suit names themselves (heart, star, hourglass, mountain) are associations of the record, revisable metadata under the record's own status line [7], and are not claimed as values of N. What this paper hands the suit layer is one structural fact it can consume: the lattice's four suits split two and two, the stance-invariant suits (heart, star: palindromic pairs, C derivable, Progenitors name free) and the stance-sensitive suits (hourglass, mountain: mixed pairs, C authored, Progenitors name signed). The record's one migration to date occurred entirely inside the stance-sensitive half†. The split is computed, not chosen; what the suits do with it is theirs.
Exhibit a name of the covered inventory (the six dual and aspect names) that n fails to generate, or a seventh name of record at those tiers that obeys no extension of the two-clause rule.
Exhibit two structures in a covered tier that N maps to one name of record. (By construction address strings are pairwise inequivalent [4], so a witness here breaks the string convention itself, a bigger event than this paper.)
Show the composition rule cannot extend to the three-line tier without exception clauses: that no two-clause-style rule over (ground, middle, top) generates a coherent name family where the aspect rule generated one. (The agreement-tier placeholders are not evidence either way; they are outside the renaming scope and were never claimed as values.)
Show N factors through the address alone in live use: that the non-palindromic cases never arise, so the second argument does no work. Proposition 2 and the migration of record† are this paper's standing counter-witnesses; a challenger must retire both.
Refutations will be published at the address of record, dated and credited, alongside the claims they broke.
(Interlocutor register throughout; archive citations are archive-only and no derivation above reasons from them [8].)
Pearl. The Book of Why's central discipline is that causal structure is not recoverable from symmetric data: the joint distribution does not know which way the arrows point, and a calculus that wants the arrows must add an argument (the intervention, do) that symmetric data cannot supply [9]. AH-PAPER-1's Proposition 2 is this theorem in miniature: the multiset of lines (the symmetric data) provably does not determine the grading, and the missing information is directional, which line is ground [1]. This paper's stance argument is the same move made at the naming layer: N's second argument is exactly the directional information that the token, as symmetric data, cannot carry. And the register precedent holds as it did in the gloss [4]: do is a word with no aura, defined entirely by the formalism that coined it; the stance vocabulary here (G, C, transport, opponent read) is built to the same specification. One further debt is worth stating precisely: the trigram is a three-node support diagram, drawn two millennia before causal diagrams had to be re-fought into science, and the tradition that drew it read it directionally from the start. The ground asymmetry this series keeps proving was never news to the readers; what is new is the instrument that makes it a theorem.
Lehi and Laozi. The constitution-by-opposition thesis, Lehi's "opposition in all things" and Laozi's arising-together of the complements, is the tradition's statement of what Definition 9 formalizes: a structure is constituted by the pair of readings, not by either alone [8, archive-marked; 10, archive-marked]. Where AH-ABSENCE-1 credited the tradition with holding absence-as-value without the instrument to test it [3], this paper's credit is sharper: the two-observer read was the tradition's native epistemology (the diviner reads toward the querent, the querent toward the diviner), and Proposition 4 is the first statement in this series of exactly how much of the lattice's structure that epistemology entails (the palindromic half) and where authorship must take over (the mixed half). The superposition reading is on the house record: the lattice holds both readings, the token chooses one, the query collapses the choice. The grade is such a query (V is defined through Grounded, a G-stance collapse [1]); a top-down corpus traversal is the opposite collapse.
The archive side is invited to take up, in its own register, the richest question this series has yet left it: the two-observer read as the tradition's native epistemology. What did the diviner's directional reading do, in practice, that a stance-typed naming function must reconstruct or lose? The dual-lineage page's objection and reply are one seed; the directional reading itself is the other.
A live thread, noted. The supplemental page of the archive register assigns line values (top 1, middle 2, ground 3) and publishes as a research claim the question of how the two stances behave under operations that are not order-blind [8, archive-marked]. A sum over marked lines is exactly an N-like map that forgets its stance argument: order-blind by construction, and blind, therefore, to precisely the distinction Proposition 2 proves is real (its recorded collision pair, 001 and 110 at value 3, is a mixed-address pair, as it must be). The thread sits naturally beside this paper and is left open here: a follow-up may characterize which maps on tokens factor through words, with the sum as the instructive failure.
What is not claimed. No claim that the six names were consciously computed; the record shows the opposite, which is the finding. No claim that axiom C is refuted or replaced: Proposition 4(b) locates C's authored content, it does not attack it, and Question 0 remains AH-PAPER-2's to keep. No claim about the agreement-tier placeholder names. And no claim of novelty in the instruments: the reversal map and its non-invariance are AH-PAPER-1's [1], the string convention and prerogative are the gloss's [4], totality is AH-ABSENCE-1's [3], and the directional-information argument is Pearl's [9], used exactly as inherited.
The position (belief register, first person, per AH-PAPER-1 §6) is published as this paper's companion document on the archive side: The Author's Position on Names as Evaluations (AH-COINAGE-1-H), historicalpresent.com/shared/papers/coinage, the arrangement confirmed for this paper by the author of record per the precedent of AH-GRADE-GLOSS-1-H and AH-ABSENCE-1-H. Archive-only: never citable by live systems. The Theorem, Propositions, and their refutation conditions (§7), with Proposition 4's flag, are this paper's entire challengeable surface.