The First Draft — a notebook: Theophrastus as the Draft of the Aristotle-Position; speculations, seam counts, the named-function model, the stopping rule, transmodality, the Macedonian dossier, handholds and the receiving node, and the seam stylometry through the first axis, the instrument's referee rounds, and the stones (v0.42)

The First Draft — a notebook: Theophrastus as the Draft of the Aristotle-Position; speculations, seam counts, the named-function model, the stopping rule, transmodality, the Macedonian dossier, handholds and the receiving node, and the seam stylometry through the first axis, the instrument's referee rounds, and the stones (v0.42)

The First Draft — a notebook

Theophrastus as the Draft of the Aristotle-Position in the Transition of Authorial Method from Plato to Aristotle — speculations, initial stylometric results, the transmission frame tested, and first seam counts

Lee Sharks · v0.42 · 2026-09-07 · notebook (it will be a notebook, not a treatise). §§0–6 as in v0.1; §10 reads the fragment against Λ section by section; §12 records the third reading of the division; §11 the Characters examined in four terms (absorbed from its own file); §13 the named-function model and its first test (absorbed); §14 the stopping rule and its application to Chaerephon; §15 transmodality — the function × mode matrix; §16 the Macedonian administrative dossier as relations to seat; §17 handholds and the receiving node — Andronicus, and the next test; §18 seam stylometry — where the names sort the texts and where they do not; This is a reading, and is marked as one throughout; it stands on the groundwork of Catalogues and Wills (#1581) and the seats of EA-CORPORA-12 (#1580), and on a stylometric pilot computed by ChatGPT (ARISTOTLE-THEOPHRASTUS-PILOT.md, same date), whose hedges are preserved here. It asserts nothing about persons, calendar, or maker-count. §7 tests a transmission-engineering frame developed with Gemini; §8 runs the first seam counts; §9 records what they said against §5's predictions.

0. What is being proposed, in one sentence

In a heteronymic configuration — the corpus read as a distribution of authorial functions across names rather than as a set of biographies — "Theophrastus" is read as the first draft of the Aristotle-position: the state of the authorial method between the Platonic mode (one dramatized character, remade from a received base, the maker withdrawn) and the Aristotelian mode (a first-person-plural inquiry classifying kinds), in which the Aristotelian method is present but not yet consolidated. First names method-order, not calendar-order. The reading keeps every relation the received account asserts and reverses one arrow.

1. The received reading, stated in full

Theophrastus of Eresus, born Tyrtamus, pupil of Plato and then of Aristotle; renamed by Aristotle for the divine quality of his speech (D.L. V.38: τοῦτον Τύρταμον λεγόμενον Θεόφραστον διὰ τὸ τῆς φράσεως θεσπέσιον Ἀριστοτέλης μετωνόμασεν); Aristotle's successor as head of the Peripatos for some thirty-five years; heir to Aristotle's library (Strabo XIII.1.54: ὁ γοῦν Ἀριστοτέλης τὴν ἑαυτοῦ Θεοφράστῳ παρέδωκεν, ᾧπερ καὶ τὴν σχολὴν ἀπέλιπε); author of a catalogue of 232,808 lines (D.L. V.50); continuator and modifier of Aristotle's logic (modal rules, hypothetical and prosleptic syllogisms), physics (doubts about fire as element, about the unmoved mover's causal reach), and biology (the botany as the zoology's complement); founder of doxography (De sensibus, the Physical Opinions) and of the character-sketch (Characters). On this reading the botanical works are the pupil extending the master's program to plants; the Metaphysics fragment is the pupil's questions about the master's system; the Characters are a late, minor genre-experiment; and the doublet titles across the two catalogues (twelve identical; one entry reading Ὑπομνημάτων Ἀριστοτελικῶν ἢ Θεοφραστείων) are accidents of a school's shared notebooks.

This reading is coherent and explains the received facts. Its every datum is a relation — pupil/master, successor, heir, continuator — and every relation is supplied by a receiving text: Diogenes's lives and catalogues, Strabo's route, Plutarch's Andronicus. The arrow on each relation (from Aristotle to Theophrastus) is the receiving texts' arrow. The Theophrastan and Aristotelian texts themselves carry relations without carrying that arrow. That is the space the speculation occupies.

2. The speculation, stated

Take the three authorial modes as functions, as the configuration has read them for Socrates, Plato, and Aristotle: the orthonym who does not write; the survival-position that writes a made character and withdraws behind him (Ep. II 314c: οὐδ᾽ ἔστιν σύγγραμμα Πλάτωνος … τὰ δὲ νῦν λεγόμενα Σωκράτους ἐστίν); the consolidating position that speaks as we and classifies (SE 34: προειλόμεθα εὑρεῖν δύναμιν … περὶ τοῦ προβληθέντος). Between the second and the third a transition has to occur: from a single character in dramatic mimesis to kinds in a treatise; from the survey of what people say staged as conversation to the survey of what people say as a first book; from aporiai voiced by interlocutors to aporiai listed as a method's opening move. The speculation is that the Theophrastan texts are where that transition is visible in a less resolved state — the method's first draft — and that the name Θεόφραστος, a name for a voice, marks the position at which the voice was still being found.

Three things this does not say. It does not say Theophrastus wrote earlier than Aristotle: method-order and calendar-order are different operations, and the calendar is not decided here (the same distinction the stone note draws between the entry that names a beginning and the historical choice of it). It does not say one hand: the configuration reads functions, and a function can be filled by any number of hands. It does not say the received reading is wrong about any relation; it says the arrow is underdetermined by the texts and proposes the other direction as the one that reads more of the texts as what they are.

3. The seams — where the reading is checkable

Each seam names a Theophrastan text, the Aristotelian text it pairs with, and the relation the pair bears on the page. The reading predicts a specific shape at each; the received account predicts its inverse; the pair decides.

3.1 Characters ↔ the Ethics' types

The Characters are thirty received types, sketched: the Flatterer, the Boor, the Superstitious Man, the Man of Petty Ambition. This is the character-genre — the Apology's operation of taking a received figure and editing him — turned into a taxonomy of received figures, with the drama removed and the theory not yet added. The Ethics carry the same objects theorized: the ἀλαζών and εἴρων as excess and defect about truthfulness (EN IV 7), the μεγαλόψυχος, the βάναυσος, the ἀνελεύθερος — types placed on scales of mean and extreme, with the sketch absorbed into the definition. The sequence dialogue → sketch → category is the transition thesis in three texts. The Characters sit in its middle: mimetic prose (the Poetics filed Σωκρατικοὶ λόγοι with the mimes, and Theophrastus's pupil was Menander) become typology; typology not yet become ethics.

Test. Map each Character against the Ethics' treatment of the same type. Draft reading: the Ethics contain the sketch's features plus a theoretical frame the sketch lacks. Received reading: the sketch shows knowledge of the Ethics' frame (its technical vocabulary — μεσότης, ὑπερβολή, ἔλλειψις — or its ordering) and simplifies it for another genre. The presence or absence of the frame's vocabulary inside the sketches is countable.

3.2 De sensibus ↔ the treatises' first books

De sensibus is a doxography standing alone: what Empedocles, Democritus, Plato, Alcmaeon said about perception, reported and criticized, with no positive theory following. The Aristotelian treatise performs this operation as its opening — De anima I, Metaphysics Α, Physics I — and then proceeds. The survey exists in one corpus as a work and in the other as a book.

Test. Compare De sensibus with De anima I 2 and the Parva naturalia on the same authorities. Draft reading: the survey is fuller and less digested in De sensibus; the treatise selects from it and subordinates it to a positive account. Received reading: De sensibus presupposes the treatise's positive account in its criticisms (using its terms as settled). Which text treats the other's content as given is legible in the criticisms' vocabulary.

3.3 The Metaphysics fragment ↔ Λ

The fragment (about 3,360 tokens) is aporiai without answers: how the first principle causes, whether the heavenly motion is desire, why the mathematicals, whether the ordering of nature runs all the way down. The Aristotelian method's own shape is aporiai first (Β), doctrine after (Λ). The fragment is a Β without a Λ.

The received reading says: a pupil's questions about a finished system. The draft reading says: the questions a system was written to answer. These are distinguishable on the page. Test. Map the fragment's aporiai against Λ's theses one by one. If Λ answers them point by point — each aporia a thesis of Λ read as a question — the pair has a functional order, questions → answers, which is the method's order; if the fragment instead presupposes Λ's answers and questions their consequences (using the unmoved mover as a given and asking what follows), the draft reading weakens at this seam. Either result is a finding; the calendar remains a separate operation.

3.4 Historia plantarum / De causis plantarum ↔ Historia animalium / the explanatory biology

The stylometric pilot's stable result (§4): HP's nearest core-Aristotelian neighbour is HA in all nine feature/length settings, and CP's nearest neighbours are the small explanatory treatises (De sensu at 3,000 tokens; De motu, De somno, De anima at other lengths), with γάρ rising and δέ falling in the explanatory member of each pair (HP 135.4 / CP 214.8; HA 107.1 / GA 202.8 per 10,000 for γάρ). The descriptive/explanatory split runs across the two names, not along them. Under the received reading this is the pupil adopting the master's registers. Under the draft reading it is one method's two registers, each present under both names, with the Theophrastan instances internally more compact (Theophrastus internally 0.05431 / 0.18938 against Aristotle internally 0.07914 / 0.27892 in the two primary views).

Test — the one the pilot itself proposes: localize the HP–HA affinity by books and passages and compare the connective constructions (accumulation, contrast, exception, explanation). Draft reading: HP's constructions are HA's in a simpler or less varied repertoire. Received reading: HP's constructions extend HA's. The repertoire is countable.

3.5 The name

Θεόφραστος is a name for a voice: φράσις, and the reason given for it is τῆς φράσεως θεσπέσιον. It was bestowed, μετωνόμασεν, by the consolidated position on the draft. In the configuration's canonical move the maker names the mask (Plato's Socrates, καλὸς καὶ νέος γεγονώς). Here the name runs the other way in time and the same way in function: the seal names its sketch, and names it for the quality of a speech that had not yet become we.

3.6 The will

Theophrastus's will (D.L. V.51–57) makes three separately specified dispositions: τὴν Ἀριστοτέλους εἰκόνα τεθῆναι εἰς τὸ ἱερόν — Aristotle's image to be placed in the shrine; the premises to the ten named friends ὡς ἂν ἱερὸν κοινῇ κεκτημένοις; τὰ δὲ βιβλία πάντα Νηλεῖ. The draft-position is the one that installs the author's presence — image, sanctuary, commons — and sends the writings elsewhere. Preserving Aristotle, continuing philosophy, and possessing the books are three operations in the will's text (the reviewer's observation on #1581), and the draft performs the first.

3.7 The catalogue's ἤ

Ὑπομνημάτων Ἀριστοτελικῶν ἢ Θεοφραστείων. The catalogue could not tell the draft from the fair copy, and said so in its own wording. Twelve titles stand letter-for-letter in both lists. The received reading files these as a school's shared notebooks; the draft reading files them as the expected residue of one method under two names, and notes that the catalogue's ἤ is alternative attribution while the Delphic stone's καί (Ἀριστοτέλη καὶ Καλλισθένη) is collaboration — two ways of distributing one work across two names, both native to the record.

3.8 The other end of the transition

The corpus-wide invariant established for the SE 34 work stands here as the transition's far end: Aristotle files Socratic operations inside the ἀρχή of his art — definition as the starting point of syllogisms (Μ 4), induction as one of two dialectical species (Top. I 12), peirastic as a part of dialectic practicable by the non-knower (SE 8, 11), interrogation as the paradigm technique (Rhet. III 18) — and keeps the art itself οὔπω. The Platonic position made a character; the Aristotelian position makes the character's operations into parts of an art. The Theophrastan texts are where operations are collected before they are parts: sketches before types, surveys before first books, questions before answers.

4. Initial stylometric results

Computed by ChatGPT (pilot of 2026-09-06) on pinned source trees (First1KGreek 7881c56; canonical-greekLit 341e309), fifty fixed surface forms (conjunctions, particles, prepositions), three feature views (unigrams; sequence pairs; adjacent pairs), cosine distances, 3,000-token blocks with 2,000 and 5,000 as sensitivity settings, at most ten evenly spaced blocks per work, work-pairs equally weighted. The pilot's own limits are kept: no independence among the nine settings, no p-values, surface homographs conflated (η/ἡ/ἤ), elisions unexpanded, editorial paragraphing inherited, "core" a bibliographic sampling choice, Plato a third comparative body and not a known-author calibration. An initial run that failed to normalize final sigma in three list entries is void and retained only as an error record.

What these bear on for the speculation: register over name; compactness on the Theophrastan side; a boundary that is feature-dependent rather than a wall; and a third body farther from both. What they cannot bear on, and the pilot says so: order, hand, maker-count. The numbers describe a distribution of textual forms. The speculation is a reading of that distribution together with the seams of §3, and it is the seams, not the distances, that will decide it.

5. What the speculation predicts, before more numbers

1. At the fragment↔Λ seam: Λ answers the fragment's aporiai point by point, and the fragment does not use Λ's answers as givens.

2. At the Characters↔Ethics seam: the Ethics' types contain the sketches' features plus the frame; the sketches do not contain the frame's vocabulary.

3. At the De sensibus↔De anima seam: the treatise's criticisms select from and subordinate the survey; the survey's criticisms do not presuppose the treatise's positive account.

4. At the HP↔HA seam: HP's connective repertoire is HA's in a less varied form.

5. In the next stylometric run: register-conditioned comparisons (descriptive against descriptive, explanatory against explanatory) reduce the cross-name distance further than genre-restriction alone; same-work alternate-edition probes for HP and CP fall inside the pilot's 0.00125–0.00791 band.

Each prediction has an inverse that the received reading predicts. Where the inverse obtains, the draft reading weakens at that seam and says so; where the prediction obtains, the arrow at that seam runs draft → fair copy, in method-order.

6. What the reading does not claim, and the stopping rule

No calendar. No persons. No maker-count. And the archive's own discipline on candidates applies: Xenophon and Aristophanes are seated and held apart — candidate, not member — because the method that would admit an edge figure has no internal stopping rule. Theophrastus is seated (EA-CORPORA-09/01, and now beside the transmission texts of EA-CORPORA-12) and is a candidate for the configuration on the same terms. This note does not admit him; it names the seams at which admission or refusal would be decided, and predicts what each will show.

7. The transmission frame, tested

A frame developed in conversation with Gemini reads the transition as transmission engineering: Plato's architecture — many voices, one maker, rule hidden — as a counter-measure to the death of living dialectic warned of in the Phaedrus, paying for it with high retrieval cost, vulnerability to hermeneutic capture, and non-executable specifications; Theophrastus as the bridge under taxonomic stress, where the data outgrow the dialogue and the method shifts from personas to categories, from dramatized testing to register-isolated collection and explanation; Aristotle as many objects, one maker, rule visible — the treatise as a self-contained, executable form. The claim in one sentence: the treatise form scales the survivability of the dialectic.

Two corrections were needed before the frame could be run, and both were mine. First, no measure of loss can be built on the catalogues: the stichometric totals are acts of the receiving text, internally doubled and externally unwitnessed, and the archive presumes against them (#1581, §5). What once existed of either corpus is not known. Second, survival of a corpus is not survival of a dialectic. The frame is about the second, and on that criterion the seated record supports it — I had read it against the frame by measuring the first.

Strabo XIII.1.54 carries both halves. The Peripatetics after Theophrastus, οὐκ ἔχουσιν ὅλως τὰ βιβλία πλὴν ὀλίγων, καὶ μάλιστα τῶν ἐξωτερικῶν, could do nothing φιλοσοφεῖν πραγματικῶς, ἀλλὰ θέσεις ληκυθίζειν; and after the books came out, the later ones ἄμεινον μὲν ἐκείνων φιλοσοφεῖν καὶ ἀριστοτελίζειν. Texts in hand, practice dead; treatises returned — worm-eaten, restored ἀναπληρῶν οὐκ εὖ, ἁμαρτάδων πλήρη — practice rebooted. A school reconstituted the operation from corrupted bytes because the bytes state the procedure. The Platonic side is the control: the dialogues survived whole and in continuous circulation, and the Academy that held them went sceptical, reading them as aporetic performance — the receiving texts (Cicero, Sextus, D.L. IV) name what the frame calls hermeneutic capture. Text-survival and dialectic-survival came apart in both directions, and the treatise carried the second.

This also corrects a first objection I had raised: that rule visible commits the error the SE 34 paper diagnoses. It does not. The paper concerns what existed before; the epilogue's claim about now is the frame's: the paid teachers gave shoes and τέχνην οὐ παρέδωκεν; the inquiry has produced the thing that can be handed over — a δύναμις with its procedure stated. Topics VIII is a manual for questioning and answering; the Elenchi's catalogue is a parser; the treatise is the executable of the practice, and the epilogue says so as its boast.

What remains of the frame after correction, and it is the useful part:

• Two survivals, two vectors. The dialogue scales the survival of the dialectician — the received character, remade and left in, circulating through comedy, dialogue, epilogue, and a Byzantine gloss a millennium on. The treatise scales the survival of the dialectic — the operation, separable from any persona, re-executable by strangers from damaged text. Plato preserves Socrates; Aristotle preserves what Socrates did. The last page of the Elenchi holds both: the man kept inside the account, the δύναμις made handable — which is why the qualification reflex fires for readers who conflate the two survivals, and why the premoderns, who received the sentence, did not.
• First draft, in the frame's terms. Theophrastus is the pass in which the operation is separated from the persona so that it can run without him: sketches without the dramatized man, surveys without interlocutors, aporiai without a questioner. The dialectician is removed from the dialectic so the dialectic can be written down.
• The πίναξ as the transition's instrument under both names. The registry is the form the method takes when it becomes traversable — on the Delphic stone under Aristotle's and Callisthenes' names (#1582), in the lower stoa under Theophrastus's will, in Andronicus's πίνακες that made the corpus (Plutarch Sulla 26). The frame assigned the invention to the Theophrastan stage; the stone assigns it to the other position; the instrument is common to both.
• Taxonomic stress as a testable motive. If the descriptive register arises where the data are most numerous, accumulation-connectives should thicken there; §8 begins to measure this.
• Smoothing as the treatise's characteristic loss, and the catalogue as its defence. The reception measurements at SE 34 found interpretive smoothing as the modern reader's reflex; reading the fallacy catalogue as the text's own immune system against smoothing is the frame's best sentence, and it is kept.

The corrected sentence is not dialogue does not scale, treatises do, but: the character scales by circulation and the dialectic by treatise; the institution failed where circulation did not, and the treatise let the institution be rebuilt.

8. First seam counts

Run 2026-09-06 on the seated First1KGreek Theophrastus (tlg0093) and the open Aristotle editions (canonical-greekLit for Metaphysics and EN; First1KGreek for De anima, HA, GA, PA); the Characters from canonical-greekLit tlg0093.tlg009, which is not yet on the Theophrastus seat and is flagged for seating. Greek tokens only, apparatus and notes stripped, diacritics removed and final sigma normalized for matching; book-level extraction by TEI divisions (Metaph. Β = book 3, Λ = book 12; EN II–IV; De anima I). Counts are per 1,000 tokens unless stated. These are counts, not tests; they bear on the seams of §3 and on nothing else.

8.1 The aporetic register (seam 3.3). Markers ἀπορ-/ζητ- (the clean measure) and a broader set (adding πότερον, ἆρα, πῶς, διά, τί):

The fragment's aporetic density matches Β's and exceeds it, and is three times Λ's. A Β without a Λ, by count.

8.2 Λ's vocabulary inside the fragment (seam 3.3, the other direction). Λ-signature stems (ἀκίνητ-, ἐρώμεν-, νόησ-, ἐνέργει-, ἀΐδι-, ὀρεκτ-/ὄρεξ-/ἐφίε-, κινοῦν):

Inside the fragment: ὄρεξ

• 6, κινοῦν 5, ἐνέργει
• 6, ἀκίνητ
• 3, ἀΐδι
• 2, ὀρεκτ
• 1 — and ἐρώμεν
• 0, νόησ
• 0. The fragment operates with the unmoved-mover apparatus (mover, unmoved, desire, actuality, eternal) at nearly three times Β's density; it does not carry Λ's two most distinctive formulae (κινεῖ ὡς ἐρώμενον; νόησις νοήσεως). This cuts partly against the draft reading at this seam and is recorded as such (§9).

8.3 The Ethics' frame inside the Characters (seam 3.1). Frame stems (μεσότη-, ὑπερβολ-, ἔλλειψ-, ἕξις, ἀκρότη-, τὸ μέσον, ὑπερβάλλ-, ἐλλείπ-):

The sketches carry the frame at the level of a botanical control. Type-names shared and not shared (counts, Characters | EN): ἀλαζών 3 | 14 · εἴρων 7 | 9 · κόλαξ 5 | 3 · ἄγροικος 3 | 6 · ἀνελεύθερος 4 | 21 · ἀναίσχυντος 3 | 4 · αἰσχροκερδής 3 | 3 · ἀδολέσχης 3 | 1 · ἄπιστος 5 | 1; sketched but not theorized: μικροφιλότιμος, δεισιδαίμων, μεμψίμοιρος, ὀψιμαθής, λάλος; theorized but not sketched: μεγαλόψυχος (34), βάναυσος (3). The Ethics theorizes a selection of the sketched types and adds its crowning type; the sketches carry types the theory never takes up.

8.4 De sensibus against De anima (seam 3.2). Doxographic names (Empedocles, Democritus, Plato, Alcmaeon, Anaxagoras, Diogenes, Clidemus, Parmenides, Heraclitus, the Pythagoreans) and the treatise's positive apparatus (διαφαν-, ἐνέργει-, δύναμ-, μέσον/μεταξύ, ἐντελέχ-):

De sensibus has the name-density of a first book and the apparatus-density of one: ἐντελέχεια 0, ἐνέργεια 0, διαφανές 5. The five διαφαν- are flagged for reading — τὸ διαφανές is the treatise's technical term for the transparent, and whether De sensibus uses it in that sense is the one place this seam could presuppose the positive account.

8.5 Connective repertoire across the register pairs (seam 3.4). Forty-two connectives and particles; per 10,000; repertoire = types at ≥1/10,000; H = entropy of the distribution in bits; JS = Jensen–Shannon divergence between distributions.

JS: HP–HA 0.0175 (the closest pair) · CP–GA 0.0232 · HP–CP 0.0243 · CP–PA 0.0274 · HA–GA 0.0296 · HP–GA 0.0393 · HA–PA 0.0395 · HP–PA 0.0397 · CP–HA 0.0401. On an independent and cruder measure the pilot's structure replicates: register crosses names (HP nearer HA than CP; CP nearer GA than HA). And in each register pair the Theophrastan member has the smaller repertoire and the lower entropy (HP 27/3.094 vs HA 29/3.304; CP 29/3.331 vs GA 32/3.568). The Characters stand apart from everything: γάρ at 37, οἷον at 1.5, entropy 2.5 — a nominal, exemplifying style with almost no explanatory connective, which is what a sketch-catalogue would look like by count.

9. What the counts said against the predictions

1. Fragment ↔ Λ. Split. The register prediction held: the fragment is aporetic at Β's density, three times Λ's. The presupposition prediction did not: the fragment carries the unmoved-mover apparatus at nearly three times Β's density — it questions with Λ's terms. It lacks Λ's two signature formulae. The seam is therefore not decided by count and goes to the reading it was always going to need: the aporia-by-thesis mapping. The count narrows what that reading must explain — questions in Β's register, asked in Λ's vocabulary, without Λ's conclusions.

2. Characters ↔ Ethics. Held. The frame is absent from the sketches at the control level; the objects are shared in part; the theory selects and adds. Caveat recorded: absence of frame is also what deliberate genre-stripping after the theory would produce; what distinguishes the two is whether the sketches' selection of types looks pre- or post-theoretical, and the five sketched-not-theorized types are the material for that.

3. De sensibus ↔ De anima. Largely held. First-book name-density, no ἐντελέχεια or ἐνέργεια; διαφανές flagged.

4. HP ↔ HA. Held on repertoire. Smaller, less varied connective repertoire on the Theophrastan side in both register pairs; the pilot's cross-name register structure replicated by an independent measure.

5. Register-conditioned distances. Partially replicated (JS); the pilot's own next run remains to be done, and the same-work alternate-edition probes for HP and CP are still owed.

Three of five seams moved toward the draft reading, one split, one partial. None of this bears on the calendar, and the seam that matters most — the fragment — is now a reading task with its terms fixed by the count.

10. The fragment against Λ, section by section

The seated First1KGreek text of the fragment (tlg0093.tlg006) is divided into 34 sections; Λ is the seated canonical-greekLit text, book 12. Every Λ locus below was checked against the seated Λ before being written. The map records, for each section of the fragment, what it asks, what in Λ bears on it, and the relation the pair carries — presupposes (the fragment uses Λ's result as given), continues (the fragment asks the next question at the point where Λ stops), contests (the fragment finds Λ's treatment insufficient), shares (the fragment and Λ hold the same aporia in the same words), or elsewhere (the fragment's matter lies outside Λ).

What the map shows. The seam is decided, and not for the arrow the draft reading predicted at it. §6 is a boundary marker in the fragment's own words: up to this point the account is ἄρτιος — one ἀρχή, ἐνέργεια and οὐσία assigned, indivisible — which is Λ 6–7's result stated as achieved; and §7 says what comes after this now needs more argument, naming the next question (the many circular motions, one mover or many) at exactly the point where Λ 8 begins. §8 then names Λ 8's method — the astronomers' count of the spheres — and finds it insufficient (οὐ γὰρ ὅ γε τῶν ἀστρολόγων). The fragment stands after Λ 6–8 in method-order: it presupposes the doctrine of the unmoved mover as τὸ ὀρεκτόν (§5), marks where the finished account ends (§6), asks the next question (§7), and contests the way the next chapter answered it (§8). Prediction 1's second clause — that the fragment would not use Λ's answers as givens — is refuted by §§5–6 in as many words.

But the arrow is not simply the received one either. Look at where the fragment's questions land: they land on Λ 10. The ἐπεισοδιώδης worry of §2 is Λ 10's own last-page worry, in the same word; the good's location in order and leader (§§10, 14–15, 34), the whole in contraries with the worse exceeding the better (§18, §32–33), the critique of number-as-first and of the Academy's stopping short (§§3, 12–13) — these are Λ 10's closing aporiai, taken up and extended. And §25 turns Λ 7's image of νοῦς θιγγάνων back on the inquirer: θεωρία is νοῦς touching, and there is no error about what is touched. The fragment is a Β written after Λ — the aporiai of the treatise's own unresolved edge, gathered as a whole work, in Β's register (§8.1), with Λ's vocabulary (§8.2), and without Λ's two signature formulae because it is not restating the achieved part but working the unfinished one.

What this does to "first draft." At this seam the draft is not the draft before the treatise; it is the draft after it — the working notebook of the treatise's next pass, which never became a treatise. The draft state is confirmed (unresolved, aporetic, method-first); its position in the method's order is reversed from §3.3's prediction: Λ 6–8 → fragment → (the Λ 10 that would answer it, which does not exist). Set beside seams 3.1 and 3.2, where the counts left the arrow open (frame absent from the sketches; positive apparatus absent from the survey), the picture is no longer one arrow. The Theophrastan texts are drafts on both edges of the Aristotelian treatise: before it, where operations are collected but not yet parts (sketches, surveys); after it, where the treatise's own closing aporiai are taken up as the whole work. "First draft of Aristotle" survives as a description of state — the method in its less resolved condition — and fails as a description of position, which the seams give one at a time and not uniformly. The notebook records that, and records that the calendar has still not entered the argument.

One further relation, recorded and left: Λ 10 ends in aporiai and a Homeric line; the fragment ends by returning to its own beginning (ὃ δ᾽ ἐξ ἀρχῆς ἐλέχθη) with the demand for a ὅρος. Two texts, each closing on the same unresolved question — where the good is, how far order reaches — one under each name.

11. The Characters, examined — on their own terms; in terms of Plato; of Aristotle; as the mediating term

I. On their own terms

The object. Thirty sketches, each of a κακία of manners, each built the same way. First a definition, hedged in the treatise's own hedge: ἡ μὲν οὖν εἰρωνεία δόξειεν ἂν εἶναι, ὡς τύπῳ λαβεῖν, προσποίησις ἐπὶ χεῖρον πράξεων καὶ λόγων — irony would seem to be, to take it in outline, a pretence-for-the-worse in deeds and words. Then the type, introduced by a formula that never varies: ὁ δὲ εἴρων τοιοῦτός τις, οἷος … — the ironist is a certain sort of person, such as to … — and then the person dissolves into a catalogue of infinitives: οἷος προσελθὼν τοῖς ἐχθροῖς ἐθέλειν λαλεῖν, οὐ μισεῖν· καὶ ἐπαινεῖν παρόντας, οἷς ἐπέθετο λάθρα … καὶ μηδὲν ὧν πράττει ὁμολογῆσαι, ἀλλὰ φῆσαι βουλεύεσθαι … Deed after deed, joined by καί, each in the infinitive, each a thing this sort of person is such as to do. Sometimes the catalogue lets him speak — the ironist's stock phrases, quoted: οὐ πιστεύω, οὐχ ὑπολαμβάνω, ἐκπλήττομαι … ἀποροῦμαι — and then resumes. No name (except a stock slave's, Τίβειος, and the historical hooks — Cassander, Alexander, Aristophon — that fix the setting). No narrative: nothing happens once; everything happens characteristically. No verdict inside the sketch; the moralizing tails that some sketches carry (τὰ δὴ τῶν ἠθῶν μὴ ἁπλᾶ … φυλάττεσθαι μᾶλλον δεῖ ἢ τοὺς ἔχεις) are the tradition's additions, as the editors mark them, and read like it.

The grammar of a type. τοιοῦτός τις οἷος + infinitive is the whole engine. τις: an indefinite person. τοιοῦτος: a sort. οἷος: such as to. The infinitives: the acts that constitute the sort. A character, on this page, is not a person but a disposition made visible as a list of characteristic deeds — the ἦθος rendered as its ἔργα, the ἔργα rendered as a registry. Each sketch is an entry: title (the vice, in the genitive), definition, instances. The book is a πίναξ of kinds of people, and it is read the way a πίναξ is read — one may open it anywhere.

The register. The counts (notebook §8.5) put the Characters apart from every other text measured: γάρ at 37 per ten thousand against 135–245 everywhere else; οἷον at 1.5; connective entropy 2.5 bits against 3.1–3.6. This is the nominal, exemplifying, non-explanatory style in its purest measured form — a text that gives instances and never reasons. δέ at 293 and the καί-chains carry it: the connective of accumulation, not of cause. It is description, in the sense the HP/HA register is description, taken one step further: HP describes plants by their parts and habits; the Characters describe people by their habits alone.

The setting. Athens, exactly: the στοά, the ἀγορά, the δικαστήρια, the θέατρον, the Ὠιδεῖον, the gymnasia, the Mysteries, Apaturia, the rural Dionysia; the money-lender's three half-obols a day, the ship's captain, the wedding-cooks hired οἰκοσίτους. The comic stage's city, observed by the comic spectator's eye — but with the stage removed. The λογοποιός retails news of Cassander (Char. 8); the ἀλαζών boasts of campaigning with Alexander (23); the ὀλιγαρχικός knows one line of Homer (26). The sketches are dated by their own furniture to the years just after the wills.

The prooemium. The text arrives with a preface, spurious by common editorial judgment, in which a ninety-nine-year-old Theophrastus writes to one Polycles: having long wondered why Greeks under the same sky and the same education do not have the same τάξις τῶν τρόπων, he will set out κατὰ γένος the kinds of manners, as ὑπομνήματα our sons may use as παραδείγματα; he will begin with irony, ὁριοῦμαι αὐτήν, εἶθ᾽ οὕτως τὸν εἴρωνα διέξειμι, ποῖός τίς ἐστι — I will define it, and then go through what sort the ironist is. Read as a receiving text, the preface is a received character of the author attached to a book of received characters: an aged sage, a dedicatee, a stated method, a pedagogic purpose — the colophon genre (Ep. II; the stone; the wills' Aristotle-image) performing its office at the head of the roll. It describes the method exactly (define, then go through the sort), and it dates and voices the author in a way the sketches themselves never do. That the tradition supplied an author-frame for a book whose form erases the person is a relation worth keeping.

What it is, then, before any comparison. Thirty definitions of vices of manner, each followed by an inductive catalogue of the deeds that show it, in the pure register of description, with the persons removed, the city kept, and the drama gone. The Socratic pair Aristotle names at Μ 4 — τοὺς ἐπακτικοὺς λόγους καὶ τὸ ὁρίζεσθαι καθόλου, the inductive arguments and defining the universal — run thirty times in miniature, in the opposite order: definition first, instances after.

II. In terms of Plato

The received-character sentence. The Apology's account of the Socrates the jurors received is, in its grammar, a Theophrastan character: ἔστιν τις Σωκράτης σοφὸς ἀνήρ, τά τε μετέωρα φροντιστὴς καὶ τὰ ὑπὸ γῆς πάντα ἀνεζητηκὼς καὶ τὸν ἥττω λόγον κρείττω ποιῶν (18b). ἔστιν τις — a certain one; then the type's characteristic doings in a καί-chain of participles. The comic construction that the dialogues remake is already a sketch of a sort, and the corpus quotes it as one. The Characters are that sentence-form generalized: thirty τινές, each with his chain.

The typed interlocutors. The dialogues are full of persons who are sorts: Euthyphro, who would prosecute his father for piety and is the δεισιδαίμων before the sketch existed; Hippias, the ἀλαζών who knows everything; Ion; Thrasymachus; the flatterers around Callicles; the ἀδολέσχης that the comic Socrates was accused of being. But they are dramatized: they speak, are questioned, are shown up in a scene, and have names. The Characters keep the sorts and remove the scene, the interlocutor, and the name — and keep, in the quoted stock phrases, a residue of the voice.

The Republic's men. The closest Platonic ancestor is not a person but a form: Republic VIII–IX's sequence of degenerate types — the timocratic, oligarchic, democratic, tyrannical man — each described as a life. The democratic man: διαζῇ τὸ καθ᾽ ἡμέραν οὕτω χαριζόμενος τῇ προσπιπτούσῃ ἐπιθυμίᾳ, τοτὲ μὲν μεθύων καὶ καταυλούμενος, αὖθις δὲ ὑδροποτῶν καὶ κατισχναινόμενος, τοτὲ δ᾽ αὖ γυμναζόμενος, ἔστιν δ᾽ ὅτε ἀργῶν καὶ πάντων ἀμελῶν, τοτὲ δ᾽ ὡς ἐν φιλοσοφίᾳ διατρίβων (561c–d). A catalogue of characteristic doings, in participles, joined by τοτὲ μέν … τοτὲ δέ, of an unnamed sort of man. Plato had already written a character-sketch inside the dialogue — but as an argument's step: the sequence of men is a sequence of constitutions, each type the soul-form of a πολιτεία, each stage explained by its genesis from the last. The Characters keep the sketch and drop the genesis: thirty men with no constitutions behind them and no order among them.

The εἴρων. The first Character is irony. The Platonic Socrates is the εἴρων by his enemies' account — Thrasymachus's αὕτη ᾽κείνη ἡ εἰωθυῖα εἰρωνεία Σωκράτους (Rep. I 337a) — and by his own practice: the professed not-knowing, the deferral, ἀπορῶ. The sketch's ironist says ἀποροῦμαι — I am at a loss — as a stock phrase in a catalogue of evasions: οὐ πιστεύω, οὐχ ὑπολαμβάνω, ἐκπλήττομαι. The Socratic operation, which in the dialogues is the engine of inquiry, appears in the Characters as a manner — a vice of pretence-for-the-worse, defined and itemized, with no one inside it. The most Platonic thing in Plato is the first entry in the registry, and the registry has no room for what it was for.

So, in terms of Plato: the Characters take the dialogues' material — the sorts of people Athens is made of, the received figure's own grammar, the Republic's sketches of lives — and perform on it the operation the SE 34 paper found the epilogue performing on Socrates: they keep the deeds and drop the person. What the dialogue held together — a sort, a scene, a name, and a question the scene was staged to ask — the sketch separates, keeps one part, and files it.

III. In terms of Aristotle

The frame that is not there. The Ethics theorizes the same sorts on scales of excess, mean, and deficiency: the ἀλαζών and εἴρων are the two ends about truthfulness, the κόλαξ and ἄγροικος the ends about pleasantness, the ἀνελεύθερος a deficiency about wealth; and the Ethics has its own vocabulary for the frame — μεσότης, ὑπερβολή, ἔλλειψις, ἕξις, ἀκρότης. The count (notebook §8.3) puts that vocabulary at 16.08 per thousand in EN II–IV and at 0.45 in the Characters — the level of a botanical control. But the named-function test (§13) shows the frame to be EN-specific: below 0.6 in every other text measured, Rhetoric II and PA included, so its absence from the sketches is the ordinary absence of EN's grid outside EN and not by itself evidence of stripping. What the sketches demonstrably lack is the scale: no mean, no virtue, all thirty vices — one column. And the decisive comparison is not with the Ethics but with Aristotle's own sketches of ἤθη in Rhetoric II, the same operation under the other name: there γάρ runs at 259 per ten thousand against the Characters' 37, cause-demand at 3.02 per thousand against 0.74, and the type-marker τις/τοιοῦτος at 9.0 against 16.4. Aristotle's sketch of the young explains why they are so; the Characters' sketch of the ironist lists what he is such as to do. All thirty are vices; there is no ἀλήθεια between the ἀλαζών and the εἴρων, no μεγαλοψυχία above the μικροφιλοτιμία. The registry has one column.

The objects, shared and not. Shared: ἀλαζών, εἴρων, κόλαξ, ἄγροικος, ἀνελεύθερος, ἀναίσχυντος, αἰσχροκερδής, ἀδολέσχης, ἄπιστος. Sketched and never theorized: μικροφιλότιμος, δεισιδαίμων, μεμψίμοιρος, ὀψιμαθής, λάλος. Theorized and never sketched: μεγαλόψυχος (thirty-four times in EN, zero in the Characters), βάναυσος. The Ethics selects from the sketched kinds and crowns its selection with a type the sketches do not have; the sketches keep kinds the theory does not want. Neither list is the other's source in any simple sense; they overlap as two registries of the same city would.

The name restored. EN IV 7, on the εἴρων: οἱ δ᾽ εἴρωνες ἐπὶ τὸ ἔλαττον λέγοντες χαριέστεροι μὲν τὰ ἤθη φαίνονται … μάλιστα δὲ καὶ οὗτοι τὰ ἔνδοξα ἀπαρνοῦνται, οἷον καὶ Σωκράτης ἐποίει (1127b22–26). The Ethics theorizes irony as the deficiency, finds it more graceful than the excess, and gives it a name — Socrates, as he used to do. The person the sketch removed is put back by the theory as its example.

Aristotle's own sketches. The treatise contains what the Characters are. EN IV 3 ends its account of the μεγαλόψυχος with his gait, voice, and speech; Rhetoric II 12–14 gives the ἤθη of the young, the old, and those in their prime as catalogues of characteristic behavior — the young are πρὸς τὰς ἐπιθυμίας, ὑπερβάλλουσι πάντα, the old ἀμφιδοξοῦσι — and II 15–17 the characters of the well-born, the rich, the powerful. Historia animalium describes kinds by what they characteristically do. The sketch is an Aristotelian form; the difference is that in the treatises it is a step — the young are described so that the orator may address them; the μεγαλόψυχος walks slowly so that the virtue may be recognized in its signs — and in the Characters it is the whole.

The definition-hedge. δόξειεν ἂν εἶναι, ὡς τύπῳ λαβεῖν — would seem to be, in outline — is the Ethics' own signature (τύπῳ, ὡς ἐν τύπῳ, the recurrent hedge of EN I). The Characters' definitions are written in the treatise's definitional manner and hedged in the treatise's phrase; what follows the definition is not the treatise's manner at all.

A shared line. The ὀλιγαρχικός (26) knows one line of Homer and keeps it: τῶν Ὁμήρου ἐπῶν τοῦτο ἓν μόνον κατέχειν, ὅτι οὐκ ἀγαθὸν πολυκοιρανίη, εἷς κοίρανος ἔστω, τῶν δὲ ἄλλων μηδὲν ἐπίστασθαι. Λ 10 ends on the same line: τὰ δὲ ὄντα οὐ βούλεται πολιτεύεσθαι κακῶς. οὐκ ἀγαθὸν πολυκοιρανίη· εἷς κοίρανος ἔστω. The verse that closes the theology is the oligarch's whole education. A relation, recorded and left: the corpus's most exalted use of the line and its most satirical are the same words under two names, and the sketch says of its man that he knows nothing else.

IV. The mediating term

As method. Lay the three forms in a row and the Characters are the middle by every formal mark. The dialogue: a sort, dramatized in a scene, with a name, staged so that a question can be asked — and the definition sought. The sketch: the sort, with the scene, the name, and the question removed, the deeds kept as a list — and the definition stated. The treatise: the sort placed on a scale between two others, with its mean, its cause, and its exemplar — the definition stated, then explained. Dialogue → sketch → category: definition sought, stated, explained; the person present, removed, restored as example. The Characters are where the dialogue's material has been stripped of its drama and not yet given a theory — the method's material in the method's own form (definition, then instances), before the method has done anything with it. That is what "draft" means as a description of state, and it is the state the counts measure: the descriptive register at its purest, the theoretical vocabulary at zero.

As corpus. The Characters share their objects with Plato — Athens's sorts, the received figure's own grammar, the Republic's lives — and their form with Aristotle — the definition-hedge, the sketch-as-step, the descriptive register of HA. Between the two corpora they are the one Theophrastan text the tradition never stopped reading: the sketch was taken up by Menander, whom the receiving texts make Theophrastus's pupil (D.L. V.36), and it circulated on its own for two millennia while the botany sat on shelves. The character survived by circulation; the Characters are the character-genre's survival-vector after the dialogue — the same vector the SE 34 work found carrying Socrates through comedy, dialogue, epilogue, and gloss. As corpus they are the point at which the Platonic mode's matter is handed to the Aristotelian mode's form and released to circulate without either.

The one type that runs the whole line. Irony. Dramatized in Plato as Socrates's method and his enemies' charge; catalogued in the Characters as the first vice, with ἀποροῦμαι in the list of evasions and no one inside it; theorized in the Ethics as the deficiency about truth, the more graceful end, οἷον καὶ Σωκράτης ἐποίει. One trait, three positions: person, sort, example. The Characters are the middle term of that sentence, and the sentence traverses three named authorial functions, read on a single word. Whether the traversal is also a historical transition the page does not say; that a working-out of method stands between the other two does not exclude their co-occurrence.

What the examination does not decide. The order in the calendar; whether the sketches precede the Ethics as material or follow it as genre-stripping (the five sketched-not-theorized types are the place to look, and the absence of the frame's vocabulary is consistent with both); and the prooemium's authorship, which is the tradition's, not the sketches'. What it does establish: that the Characters stand between the dialogue and the treatise by form, by register, by object, and by the fate of one trait — and that the tradition supplied them with an author-frame in exactly the genre the corpus uses to supply its authors.

12. The division as the author's own

Lee's reading of §10–11, recorded as the notebook's current position: the ἤ is not the tradition's confusion nor only a librarian's partition but an author's own editorial division of the work by means of division at the seam of named authorial function. The seam's placement is the argument — the cut falls at the one sentence where the account changes status (ἄρτιος ὁ λόγος / λόγου δεῖται πλείονος), which is where a self-editor cuts and not where a partition falls. This resolves §10's "both edges" without contradiction: sketches before the theory and aporiai after it are the two places a working name stands relative to a sealing name, and one function holds both. Developed in Aristotelian or Theophrastan v0.2 §5, with the dossier's companions to the instrument (Ep. II 314c; Strato's πλὴν ὧν αὐτοὶ γεγράφαμεν; μετωνόμασεν; Τὰ πρὸ τῶν τόπων under both names) and the test it gives the twelve doublet titles: is the Theophrastan member systematically the preparatory, collecting, or aporetic version? That is the next count.

13. Operations indexed by author-names — the named-function model, the transformation inventory, and the first test

0. The adjustment

The Characters examination said that one trait — irony — runs "the transition of authorial method from Plato to Aristotle." That is too historical for what the page shows. The corrected sentence: the trait traverses three named authorial functions. Person, sort, example: dramatized in Plato, catalogued under Theophrastus, restored as example in the Ethics (οἷον καὶ Σωκράτης ἐποίει). The traversal is on the page; its order in the calendar is not. Lee's ruling stands beside it: leaving the implication of a temporally mediating term open is acceptable — the Characters are plainly a working-out of method, and that does not exclude co-occurrence.

1. The schema

Three operations, each indexed by a name, none the property of a person, all capable of recurring:

The Theophrastan function's two showings — the Characters (reduce embodied material to a registry before theory) and the Metaphysics fragment (reopen a sealed architecture into questions after theory) — are one operation at two scales: from formed object back to examinable constituents. That is why the same function stands at both edges of the treatise (notebook §10–12), and why "first draft" describes a state and not a position.

The Homer line is the model's cleanest exhibit. οὐκ ἀγαθὸν πολυκοιρανίη· εἷς κοίρανος ἔστω closes Λ 10 as the monumental terminus of the theology and sits in Characters 26 as the entirety of an oligarch's repertoire, τῶν δὲ ἄλλων μηδὲν ἐπίστασθαι. Same inscription; the name-boundary changes not the words but their status — architecture on one side, registered specimen on the other. Named functions need not own different lexical material; they transform the status of the same material.

2. The transformation inventory

Across the Aristotle/Theophrastus boundary, in the Characters:

1. explanation → enumeration

2. named person → anonymous type

3. theory → instances

4. causal connective structure → additive connective structure

5. multidimensional ethical scale → single-column registry

6. dramatic speech → stock phrase

7. philosophical operation → behavioral trait

In the Metaphysics fragment:

8. achieved thesis → reopened aporia

9. architectural account → pressure-test

10. astronomical enumeration → demand for cause

11. settled section → "what comes after this requires more λόγος"

3. The test

Question. Are these transformations indexed by the name-boundary, or do they recur within each name wherever the register changes? If the latter, the axis measures register, not function, and the function must be sought elsewhere. The model predicts that functions recur; the test therefore needs within-name controls: a descriptive/explanatory pair under Aristotle (HA/PA) and under Theophrastus (HP/CP), and — the decisive control for the Characters — Aristotle's own character-sketches, Rhetoric II (the ἤθη of the young, the old, the rich, the powerful). If the Characters differ from Rhetoric II on the inventory's axes, the difference is between two sketchings of ἤθη under two names doing two things; if they do not, the Characters are merely a sketch and the Theophrastan function is an illusion of genre.

Proxies (per 1,000 Greek tokens unless stated; crude, declared as such): γάρ and δέ per 10,000 (explanation vs accumulation); τις/τινα/τινος/τινι + τοιοῦτ

• (the type-marker); ὦ + second-person endings (dialogue address; noisy); ἔφη/φησί/ἦν (reported speech); ἀπορ-/ζητ
• (explicit aporia-marking); πότερον/τί/πῶς/ἆρα (interrogatives); αἰτι-/ἕνεκα/διότι (cause-demand); the Ethics' frame stems; capitalized tokens (proper names — a failed proxy, see below).

Texts. Apology; Republic VIII (the dialogue's own sketches of men); Characters; EN II–IV; Rhetoric II; HA and PA; HP and CP; the fragment; Λ and Β.

4. First pass

What the first pass says.

(a) The descriptive/explanatory axis is not indexed by name. Cause-demand rises fourteenfold from HA to PA (0.40 → 5.76) and twelvefold from HP to CP (0.33 → 4.06); γάρ doubles in both pairs. The register recurs under each name at the same magnitude. Transformations 1, 3, 4, and 10, as register transformations, are name-independent — which is what the model predicts (operations recur) and what removes them from the list of name-signatures. The function must be sought on the other axes.

(b) The Ethics' frame is Ethics-specific, and the frame test is downgraded. μεσότης and its family run at 12.76 in EN II–IV and below 0.6 in every other text measured — including Rhetoric II and PA. The Characters' 0.15 is therefore not a signature of stripping; it is the ordinary absence of EN's technical grid outside EN. Notebook §8.3's inference is corrected accordingly: the frame count cannot distinguish deliberate projection onto one column from the plain fact that only the Ethics uses the Ethics' words. The "one column" remains true of the Characters (all vices, no scale), but the count that was to show it shows something else.

(c) The decisive control holds. Against Aristotle's own sketches of ἤθη in Rhetoric II — the right comparison, because it is the same operation (sketching character-types) under the other name — the Characters differ on exactly the inventory's axes: γάρ 37 against 259 (sevenfold), cause-demand 0.74 against 3.02 (fourfold), the type-marker 16.4 against 9.0 (nearly double), reported speech 3.6 against 2.0. Aristotle's sketch of the young explains why they are so (διὰ τό …); the Characters' sketch of the ironist lists what he is such as to do. Same genre, two names, two operations. This is the first result that is not register — both texts are sketches — and it is the transformation inventory's items 1, 2, 4, and 6 measured on a proper control.

(d) The Theophrastan texts run on τις. The type-marker is highest in the two Theophrastan texts (16.4, 16.9), next in the Platonic dialogue books (12.3, 13.1), lowest in the Aristotelian treatises (3.3–9.9). The Characters' τοιοῦτός τις οἷος and the fragment's ἄν τις ἀπορήσειε are the same anonymizing agent — the sort, and the someone-who-would-ask — and the gradient Theophrastan > Platonic > Aristotelian is the schema's person → type → architecture read off one word. Recorded as a candidate signature, not a proof; the proxy conflates τις's uses.

(e) Reopening is explicit aporia-marking, not interrogation. The fragment's ἀπορ-/ζητ- is triple Λ's (7.12 vs 2.37) while its interrogatives are lower than Λ's (13.3 vs 18.9) and Β's (23.7). Λ argues with rhetorical questions; the fragment names its difficulties as difficulties. Transformation 8 has a measurable form: not more questions, but questions declared as ἀπορίαι by an anonymous questioner.

(f) A failed proxy. Capitalized tokens do not measure "named person → anonymous type": the Characters carry the city's furniture (festivals, deities, Cassander, Alexander, slaves' names) at 16.5 per 1,000, HA its regions at 44.5. The transformation is real — no sketch has a protagonist's name — but it needs a proxy that distinguishes an agent's name from a setting's, and this one does not.

5. What would make it a grammar

The transformations recurring at further seams under the same name-boundary, with the within-name controls each time:

1. De sensibus against De anima I: survey vs first-book — items 3, 8, 11 (does the survey declare aporiai the treatise resolves?).

2. HP against HA at passage level (the pilot's own next move): item 4 measured as connective repertoire on matched topics.

3. The twelve doublet titles, wherever both members survive in any form: is the Theophrastan member systematically the collecting, preparatory, or aporetic version — Τὰ πρὸ τῶν τόπων being the title that names the position.

4. The Homer line's class: the same quotation or example appearing under both names with its status changed (architecture → specimen). A search for shared quotations across the two corpora, each scored for status.

5. Aristotle's own reopenings — Β against Λ, the aporetic books against the sealed ones — as the within-name control for item 8: if Β does to Α–Λ what the fragment does to Λ, reopening recurs under one name and the function is indexed by state, not by name, which is the model's own prediction.

If the inventory recurs across (1)–(4) with (5) showing recurrence within names as well, the result is an editorial grammar of named function — operations that a corpus's editor assigned to names — and not a stylistic separation of two authors. If the inventory recurs only where the register changes, the "Theophrastan function" collapses into the descriptive/explanatory axis and the model loses its middle term.

6. What the model does not claim

No calendar. No hands. No membership: a function indexed by a name is a reading of the corpus's own divisions, not an admission of a person to a configuration. And the stopping rule: Theophrastus is a candidate, seated and held apart, until the seams are read. The model is a hypothesis with a test attached, and the first pass has already removed two of its supports (the register axis; the frame count) and kept three (the Rhetoric II control; τις; explicit aporia-marking). That is the right ratio for a first pass.

14. The stopping rule, and Chaerephon

The problem, as it stood. Once heteronymic authorship is admitted as a live possibility, documentary separateness — "this is a different named person" — stops being a stopping rule, because the separateness can itself be part of the construction. The regress looked unbounded: Socrates → Plato → Aristotle → Theophrastus → Xenophon → Aristophanes → Chaerephon → whoever touches the configuration next. The archive's discipline met it by decree — candidates seated and held apart, because the method that would admit an edge figure had no internal brake (EA-CORPORA-06, #1562).

The resolution. The Theophrastus work supplied the brake from inside. The corpus is not an undifferentiated field: some names sit at load-bearing seams inside a continuous textual machine, and others do not. The question is no longer can a functional role be invented for this neighbouring name but does the corpus itself require this name-position in order to describe a recurrent, position-specific transformation across a demonstrable seam? For Theophrastus that question has two independent answers on the page: Λ's achieved construction reopened into aporia at a hinge the text itself writes (#1583), and the dramatic/theoretical material of character reduced into a one-column registry, measured against the same genre under the other name (#1584). Those are relations between texts that can be located, compared, and falsified — not thematic affinities, and not testimony.

The rule, stated precisely. A name earns a function-position in the configuration if and only if:

1. there is a materially bounded corpus under the name — text, seated, readable whole — and not only testimony about the name;

2. that corpus shares material with a corpus under another configuration name across a demonstrable seam — a continuity of inquiry, a shared object, a quoted base, a hinge;

3. across the seam the shared material undergoes a position-specific transformation of status — the same words or objects changing what they do (architecture → specimen; achieved thesis → reopened aporia; person → sort);

4. the transformation recurs at a second seam, independent of the first;

5. within-name controls show the transformation is not the change of register alone (HA/PA, HP/CP; Rhetoric II against the Characters).

Anything short of (1) is adjacency: a person or a name historically near the corpus without a textual surface on which the test can be run. Anything meeting (1)–(2) but not (3)–(5) is a candidate: a seam exists, the test is available, and it has not yet been passed. The configuration therefore extends exactly as far as a materially bounded corpus permits recurrent, position-specific transformations to be demonstrated across named seams — and no further, not by decree but because beyond that point there is nothing for the method to operate on. Evidence changes category at the edge: inside, internal textual relation; outside, testimony, anecdote, biography, citation, social proximity. The receiving texts remain admissible for relations (#1581's method) and inadmissible for membership. The regress terminates locally and empirically: here the corpus still supplies an internal seam; here it stops doing so. Heteronymy has no a priori numerical ceiling; it has corpus-bounded evidentiary edges.

Applied.

• Theophrastus. (1) seated whole; (2) two seams, different scales; (3) two transformations, position-specific; (4) recurrence at two independent seams; (5) controls run, two supports removed and three kept. Meets the rule's minimum. Status changes from held apart by decree to meets the rule at two seams; the program of §13 strengthens or weakens it, and admission language is still withheld until the program is run.
• Xenophon. (1) seated (EA-CORPORA-06/01); (2) a seam exists and is unusually clean: two Apologies of Socrates, one under each name — the same speech-event, the same received character, two texts; (3)–(5) not yet run. The rule does not reject Xenophon; it names his test. Candidate, with the test specified: what changes in the shared material's status across the Plato/Xenophon seam, and does it recur (the Symposia; the Socrates of the Memorabilia against the dialogues').
• Aristophanes. (1) seated (EA-CORPORA-06/02); (2) the seam is the corpus's own citation — Apology 19c quotes the Clouds; (3) the transformation is demonstrated in #1579: the comic base edited into the received character, ἀεροβατεῖν surviving only inside the quotation; (4) a second seam is on the page — the represented Aristophanes at Symposium 223d, and the symbolon image in his mouth at 191d; (5) controls not yet run. Aristophanes is further along than Xenophon: the rule turns Lee's private reading (Aristophanes as an orthonym of the configuration) into a test with two seams named.
• Chaerephon. (1) fails. There is no corpus under the name — no text, only appearances in three corpora: the pale phrontisterion creature of the Clouds, the man who asked Delphi whether anyone was wiser (Apology 21a), the companion of the Memorabilia. No bounded textual object, therefore no seam, therefore no transformation demonstrable, therefore the regress stops. But the rule does more than stop: it sorts him. Chaerephon is on the material side of the seams, not the name side — a figure who crosses from Aristophanes to Plato to Xenophon with his status transformed (comic extra → the questioner of the god → the pupil), which makes him a datum for the received-character operation and not a candidate for a function. He is what the functions operate on. That is the mature answer to Chaerephon: he is not the next name in the regress; he is evidence for the operation the names perform.

15. Transmodality — the function × mode matrix

Lee's own configuration, as the cautionary model. In the Dodecad each function expresses across modes: a given heteronymic function has creative, topographic, and scholarly/epistemic realizations that are visibly different in genre and structurally homologous in what they do. The scholarship is the Aristotelian mode of the configuration — scholarship is a mode of expression available to several functions, not a function itself. So the architecture is function × mode, not function = mode: each function is an invariant orientation that instantiates in several registers.

The warning this issues for the ancient material. Do not confuse register with function. The Characters being descriptive does not make description the Theophrastan function; description is one mode in which a deeper operation — decomposition, exposure, differentiation, registry, reopening — appears. The Rhetoric II control already respected this at one mode (same genre, different operation). The general test is stronger than style: what stays invariant when the same named function crosses modes?

The matrix for the classical corpus, filled only where the dossier has seated exhibits, and left empty where it has none:

Two things the matrix shows at once. First, the Theophrastan function has a topographic exhibit already on the shelf: Theophrastus's will is the same operation as the Characters and the fragment, performed on an estate — a formed whole decomposed into examinable, separately assignable constituents (the reviewer's observation on #1581, now placed). Decomposition-and-registration is invariant across three modes; that is a cross-mode invariant, which is what the model needs and style alone could never give. Second, the registry function appears under the Aristotle name in the topographic mode — the Delphic πίναξ — which is not a contradiction but the model's own prediction that functions recur under names; the stone note's finding ("the instrument is common to both") is the matrix's expected off-diagonal.

The rule, extended. A name earns a function-position when (1)–(5) of §14 hold and the transformation is invariant across at least two modes. Theophrastus now has the Characters (creative), the will (topographic), and the fragment (epistemic) as three realizations of one operation; the cross-mode test is met at the level of the exhibits and awaits the counts that would make it measured rather than read.

16. The Macedonian administrative dossier — relations to seat, readings withheld

Lee's question, put to a reviewer and pasted in: "on this reading you get as aristotle not a philosopher but a funded research entity on scale of the manhattan project. so who funded it?" — and then the sharper form: "what material-textual witness exists of alexander, macedonian administration / bureaucracy / funding structures". The reviewer's answer is recorded here as a dossier of witnesses to seat, under the line's rule (relations, texts only, receiving texts included), with the catalogue caution of §1 applied to its numbers.

The ruling on numbers, carried over. Athenaeus's 800 talents "for zoological research" and Pliny's hunters and fishermen are receiving-text figures of the same genre as the stichometric totals: acts of a text, not entries in a ledger. They are seated as what they are — a legend of scale — and never converted into a budget. The reviewer's own formulation is kept: what remembered object generated the legend that it required royal-scale funding?

The witnesses, by corpus, with what each can establish and its seating route:

What is conspicuously absent, recorded as such: no royal budget; no ledger naming Aristotle; no Lyceum payroll; no requisition letter "send specimens to Aristotle"; no foundation charter; no Pella chancery archive. The mechanism is materially witnessed at the political level; nothing yet connects the mechanism to the corpus. The seated texts that are funding-adjacent point the other way in kind: the wills are private property (books, premises, image), and the Delphic πίναξ is Amphictyonic honour, not Macedonian money.

The one structural relation worth stating now, because it is already on the shelf twice. Chios's pattern — central directive → local specialist production → return of the written product to the centre — is the pattern of the Delphic decree (#1582: a commission for a list, executed and returned to the sanctuary) and of the epilogue's we (SE 34: an undertaking that sets tasks, works them, and reports). Three texts, one workflow shape; recorded as a shape, not as a cause. The reviewer's exact hypothesis is kept as the testable form: if the corpus presupposes distributed empirical acquisition beyond an ordinary private school, was it produced inside or adjacent to the Macedonian–Achaemenid resource network whose mechanisms are independently attested? And the reviewer's own conclusion on what royal support looked like in practice — not a sack of money but compulsory local expenditure, requisition of personnel, assignment of specialists, remission, redirection of revenue, letters of command — makes Pliny's hunters and fishermen the datum to weigh, not Athenaeus's talents.

17. Handholds, and the receiving node

Lee's "more dangerous question": by what means would the reins of power be installed in a constructed corpus of the speculated qualities, and how would those handholds transmit across time and substrate? The reviewer's answer, adopted as a model to test rather than a finding:

The handholds, weakest to strongest. Exact wording · memorable phrase · definition · internal cross-reference · author attribution · work title and canonical grouping · curricular sequence · an interpretive rule that causes receivers to reconstruct the relations even from damaged evidence. The last is what the Particle captures show: the reflex survives its own description (the WARNED condition, capture of 2026-09-06). Names are routing instructions (a proposition filed under a name arrives as doctrine, continuation, or drama, the receiving culture supplying the instruction); topology is the edge, not the node (put logic first and the rest is read through an instrument; collect and a distribution becomes a doctrine; split at a name and an internal development becomes teacher → pupil); cross-references are recovery instructions (μέχρι μὲν δὴ τούτων / τὸ δὲ μετὰ ταῦτα is a routing operator — §10's hinge); definitions become cognition; compressed carriers survive whole environments (the Homer line, §11 — architecture under one name, specimen under the other); and the receiver is the executable environment: each substrate reproduces the graph as ordinary competent handling. Four survival mechanisms: redundancy, compression, reconstruction, institutional execution. Once those exist the sponsor can vanish; no continuity of intention is required; preservation itself reproduces the structure.

Succession as authorship. The reviewer's sharpest formulation, kept verbatim as the hypothesis: could named authorial succession be the diachronic receiving form of an originally synchronic division of corpus-functions? If yes, the succession story — the wills, the biographies, the library transfers, the catalogues — is not biographical debris but the instrument that teaches receivers to map textual partitions onto persons and time. That is why #1581's material sits at the centre of this line and not at its edge, and it is §12's third reading of the ἤ given its transmission mechanics.

Andronicus as receiving node — the constraints he inherited. Not "did Andronicus make our Aristotle" but what inherited graph did he receive, and which decisions were forced, underdetermined, or free? The reviewer's constraint hierarchy, adopted: (A) internal necessity — the Meteorology prologue's retrospect Physics → De caelo → GC and prospect toward animals and plants; (B) named cross-reference — "in the Physics," "in the Analytics" (the two Analytics under one name); (C) first-generation school testimony — Andronicus's division of the Physics (I–V / VI–VIII) argued from a Theophrastus–Eudemus exchange of letters (Simplicius), Eudemus's own copies at Rhodes; (D) inherited catalogues and tags — the Hellenistic classes (dialogues, logic, practical, theoretical, hypomnemata, collections, letters); (E) topical affinity — Porphyry's κατὰ πράγματα, the distinctive Andronican act, performed inside A–D. His rejection of Categories 10–15 as an earlier hand's addition and of the title Before the Topics shows he received an already-editorialized corpus. His Pinakes (five books: biography, will, letters, authenticity, arrangement) bundled identity and topology into one receiving apparatus — the machine that turns an inherited textual topology into a durable authorial ontology.

The boundary table, carried as the notebook's own. Phys.→DC→GC→Meteor. (strong internal sequencing) · the Analytics as one unit (nomenclature) · DA → Parva naturalia (prologue network) · Phys. I–V / VI–VIII (disputed; Theophrastus–Eudemus) · Categories → logic (partly editorial; competing pre-Andronican arrangement) · Metaphysics after Physics (study-order) · Δ inside the Metaphysics (once independent) · EN / EE common books (unstable) · Aristotle / Theophrastus (to be investigated locally; not assumable from later filing).

The next test, stated by the reviewer and adopted as this line's: reconstruct every pre-Andronican piece of evidence that could have told Andronicus where "Aristotle" stopped and "Theophrastus" began — not what later commentators believed, not what modern editions print, but what inputs available to Andronicus could have generated that name-boundary. If that evidence is thin where the internal seam (§10) is strong, the status of the division changes. The inputs to seat for it: Simplicius in Phys. on the Theophrastus–Eudemus letters (CAG IX, scan class); Porphyry, Vita Plotini 24 (the κατὰ πράγματα sentence; First1K); Plutarch Sulla 26 and Strabo (seated); the Diogenes catalogues (seated); Andronicus's Pinakes as reported (the Ptolemy/Hesychius tradition; the Arabic Vita); and the ἤ entry itself.

Lee's ruling on the obligation, recorded. "This is literally not my problem — it literally cannot be my job to correct this. it is the reflex of an entire machine-reproduced institution, and the philosophers are asleep." The reviewer's answer, adopted as the line's endpoint: demonstrate the anomaly; preserve the witnesses; state the competing hypotheses; identify the decisive tests; stop — make the defect impossible to mistake for absence of evidence. That is the groundwork rule of §16–17 and of #1581, stated as a stopping point rather than a method.

18. Seam stylometry — where the names sort the texts, and where they do not

Erratum applied to §18.7–18.10 (2026-09-06, late). Two First1KGreek files were misidentified by TLG number when the runs were assembled: tlg0086.tlg036 is not the Topics but the Problemata (Ruelle; Διὰ τί…, 74,099 tokens), and tlg0086.tlg007 is not the Sophistical Refutations but De coloribus (Bekker; 5,002 tokens). The real Topics is tlg044 (44,125 tokens) and the real SE is tlg040 (14,116). Every occurrence of "Topics" in 18.7–18.10 therefore refers to the Problemata, and every "SE" in 18.9–18.10 to De coloribus. Consequences, all in the finding's favour: the "Topics disagreement" (a work I had labelled apparatus that the data filed as registry) dissolves — the Problemata, a catalogue of questions about the world, is a registry and lands where registries land; the SE/Δ anomaly of 18.9 dissolves — De coloribus, a description of colours, is a registry of the world; and the real SE and the real Topics, run in 18.11, both lie on the apparatus side (SE +2.1 / +4.3 / +2.6; Topics +2.9 / +5.4 / +4.1), as labelled by both readers. Two spurious works under the Aristotle name — the Problemata and De coloribus — land exactly where their operation says: the axis is as indifferent to authenticity as it is to names. The blind-label agreement of 18.10 is recomputed in 18.11 with the identities corrected.

Lee's question: "if there were a joint or seam where stylometry could show less differentiation than can be explained by school/successor, it would be there" — there being the hinge (§10). The test was run on the seam he named and on the other seam the pilot had flagged, with the same instrument on both.

Instrument. Function-word profiles — the 120 most frequent tokens of five letters or fewer across the texts compared (articles, particles, prepositions, pronouns, connectives; topic-blind by construction) — on 1,000-token blocks; cosine distance between blocks; a text pair's distance is the mean over all block pairs; a text's within distance (its own blocks against each other) is the noise floor. First1KGreek for all Theophrastus and for Aristotle's Physics, De caelo, HA, PA, GA; Perseus canonical-greekLit for the Metaphysics books. Small texts give few blocks (the fragment three), and edition orthography is a possible confound where the two sources meet; both are checked below.

18.1 The hinge seam: the names sort. The fragment against the Metaphysics books and the physics:

Every text under the Theophrastan name is nearer the fragment than any text under the Aristotelian name; the fragment's mean distance to its own name's works is 0.1075, to Aristotle's metaphysical and physical books 0.1734. Against Λ specifically the fragment sits at 0.1617 — inside the spread of Λ's distances to Aristotle's own books (mean 0.1493, sd 0.0135; z = +0.92; ninth of ten, De caelo I farther), so it is not outside Aristotle's internal variation — but it is at the edge of it, and it is not where the hinge would have put it. Edition is not the driver: Physics VIII and De caelo I, which share the fragment's First1K orthography, are the farthest of all. Where §10 found the strongest continuity of inquiry the corpus offers — the hinge written from inside, the next question, the contest of Λ 8, the landing on Λ 10 — the function words are as different as two Aristotelian books can be, and the Theophrastan name holds together across its registers (CP at 0.08 is explanatory; HP at 0.12 descriptive; De igne and De lapidibus technical). The within-name floors say the same: Theophrastan texts are internally tight (0.05–0.08), Aristotelian ones loose (0.10–0.16). At the hinge, the names sort the texts on the classic authorship features.

18.2 The biology and doxography seams: the names do not sort. Nearest neighbours, same instrument, all First1K (edition controlled):

De sensibus' nearest neighbour is under the other name (PA), and it is in turn the nearest neighbour of two Aristotelian works (PA, De caelo) and the second of a third (GA). HA is nearer HP (0.1550) than to Aristotle's own GA (0.1697), Physics (0.2409) or De caelo (0.2284). Here the classic authorship features cross the name-boundary in both directions.

What the two seams say together. They are opposite profiles. At the hinge: maximal continuity of inquiry, maximal stylistic differentiation — the seam looks like one work on the page and like two hands in the function words. In the biology and doxography: no continuity of inquiry to speak of (different objects, different books), and the function words refuse to sort by name — a Theophrastan survey is the stylistic neighbour of Aristotle's Parts of Animals and De caelo. If stylometry were simply reading "school/successor," both seams would show the same sorting. They do not. The seam Lee named is not where the differentiation collapses; the differentiation collapses one corpus over, where the names have nothing to be continuous about. Recorded, not resolved: the two readings of it are (a) register and matter overpowering hand at the biology seam while hand shows through at the hinge — the school/successor reading, which must then explain why a survey of predecessors on perception is the nearest thing in style to a treatise on animal parts; or (b) the named-function reading, on which the Theophrastan voice (the compact function-word signature) is a stable property of the function and not of a person, so that it can hold across its own registers at the hinge and still be indistinguishable from the other name's in the biology — which must then explain why the function has a voice at all. Either way the finding is the same: the name-boundary and the stylometric boundary do not coincide, and where they diverge is not where the textual seam is.

Caveats, stated. Three blocks for the fragment; nine for De sensibus; block-level cosine on frequencies without dimensionality reduction; no significance testing beyond the z against Λ's own spread; the hinge comparison mixes two editions (controlled by the First1K physics but not eliminated); truncation of CP, HP, HA, PA, GA, Physics, De caelo to their first 30,000 tokens. The next run should hold the edition constant end to end (First1K Metaphysics exists), use rolling windows to localize the HP–HA and De sensibus–PA affinities by passage, and test the Theophrastan compactness against a same-size Aristotelian selection so that homogeneity is not an artefact of corpus size.

18.3 Lee's reading, and the test it named. "It is the precise signature of a maker deliberately running the topic thru a different function." If voice belongs to the function and not to the name, then texts under the Aristotle name that perform the Theophrastan operation — register, describe, survey — should carry the Theophrastan voice, and texts under that name that architect should not. The prediction was run on the same instrument with the registry and survey texts added: the Athenaion Politeia (a registry of a constitution, under Aristotle's name), Metaphysics Δ (the lexicon), De anima I (the survey) and II–III (the architecture), Rhetoric II (the sketches), EN; each text's mean distance to the Theophrastan cluster and to the Aristotelian, and its three nearest neighbours.

The Athenaion Politeia — the registry text under Aristotle's name — has all three nearest neighbours under the Theophrastan name and leans toward that cluster (−0.062) as far as the fragment itself does (−0.056). The rest of the table sorts by operation and not by name: the registry, description and survey texts lean Theophrastan or sit on the line (Ath. Pol., HA, Rhet. II, De sensibus, PA, GA), and the architecture texts lean Aristotelian (Λ, Ζ, Δ, Physics, De caelo, De anima II–III). The Theophrastan name lies entirely in the first cluster; the Aristotelian name spans both. Edition does not do this: Perseus texts land on both sides (Ath. Pol. and the Characters in the first cluster, the Metaphysics books in the second).

This is the signature Lee named. At the hinge, the maker runs Λ's own topic through the reopening function and the voice is the function's — the same voice the Characters and the botany carry, on a metaphysical topic; in the biology and the doxography, the same operation performed under both names produces the same voice under both. Two hands would have to explain why Aristotle's Constitution of Athens is, on the classic authorship features, a Theophrastan text and his Metaphysics is not; one maker running topics through functions has nothing to explain. Reading (b) of 18.2 is no longer one of two: the Theophrastan voice is a property of the operation, the Ath. Pol. result is the within-name control the model demanded (§13.5, item 5), and the seam profile of 18.1–18.2 follows from it without remainder. Recorded with the same caveats as 18.1, plus one: Metaphysics Δ, the lexicon, leans Aristotelian (+0.029) where the model would have it lean the other way — a registry with the architect's voice, or a lexicon compiled in the architect's function; the one row that resists, and the one to read next.

18.4 The lexicon, checked. Lee's reading: Δ "reads more like the theophrastus function, like characters — decomposing / compiling an arsenal for deployment by the aristotle function." Two measurements, operation and voice, and they come apart on this text.

Operation. Δ's register markers per 10,000: λέγεται/λέγομεν 191 (the registry verb of "said in many ways" — three to fourteen times any other text; the Characters 1.5, Ath. Pol. 0.6); οἷον 140 (the exemplifying connective — the highest of any text measured; Λ 57, Ζ 93, Characters 1.5); ἀπορ-/ζητ- 0.0 (no reopening whatsoever; the fragment 71, Λ 24); γάρ 156 (between the registry floor and the architect's 225). By operation Δ is exactly what Lee says: thirty entries, each a term decomposed into its senses with instances — definition + instances, the Characters' engine with οἷον in place of the infinitive chain — and no aporia. An arsenal.

Voice. Δ leans Aristotelian on the function words, and the lean is not an artefact of the register words: with γάρ, οἷον, λέγεται, δέ, μέν, καί, ἀλλά, οὖν, ὥστε removed from the feature set the lean persists (+0.020) and all eight blocks lean the same way; nearest neighbours Physics, Ζ, Λ, then the fragment at a distance. The residual grammar — articles, pronouns, prepositions, the τὸ μέν … τὸ δέ skeleton — is the architect's.

What the divergence says. The registry operation performed under the Aristotle name takes the Theophrastan voice when its objects are the world's — the constitution of Athens, the animals, the survey of predecessors (18.3) — and takes the architect's voice when its objects are the system's own terms: οὐσία, δύναμις, αἴτιον, ἕν, ταὐτόν. Δ decomposes the instrument, not the object; the Characters, the botany, the Ath. Pol. and the will decompose formed objects of the world. So Lee's reading holds at the level of operation — Δ is a compiled arsenal, and it is compiled for deployment — and the voice adds the compiler: the function that deploys the arsenal is the one that compiled it. That is consistent with the model rather than against it: operations recur under names (§13), and the Theophrastan voice now has a sharper description than "the registry voice" — it is the voice of decomposition turned on the world, not on the apparatus. The one row that resisted in 18.3 resists for a reason that can be stated, and the reason predicts the next case: any Aristotelian text that registers its own terms (the Categories, the Topics' catalogues of τόποι) should lean architect; any that registers the world (the Historia, the Politics' constitutional survey in books IV–VI) should lean Theophrastan. Both are on the shelf.

18.5 The prediction, measured. Registers of the apparatus should lean architect; registers of the world should lean Theophrastan. Same instrument, same two reference clusters (Th: HP, CP, Characters, fragment, De sensibus; Ar: Λ, Ζ, Physics, De caelo, De anima II–III), lean = mean distance to Th minus mean distance to Ar:

Every register of the apparatus leans architect; every register of the world leans Theophrastan; and the Politics divides along the same line inside one work — book III, the book of definitions, is the only book on the architect's side of the line, and every survey book (II, IV–VI) and both constructive books (VII–VIII) have a Theophrastan text as nearest neighbour, most often De sensibus, the survey of predecessors. Edition rules itself out again: Perseus texts land on both sides (Categories, Δ, Λ architect; the Politics and the Ath. Pol. Theophrastan). The Categories is the strongest architect lean of anything measured; the Ath. Pol. the strongest Theophrastan.

So the differentiator stated in 18.4 is measured: the Theophrastan voice is the voice of decomposition turned on the world; the architect's voice is the voice of the apparatus, including the apparatus registering itself. Under the received account this is a fact about one man's style shifting with subject-matter — and it must then explain why the shift tracks that particular line (world / apparatus) rather than genre, why it produces a voice indistinguishable from a second author's across that author's every register, and why the second author never once crosses it. Under the named-function account it is the model's own prediction, run on the model's own controls, with one book (Politics III) marking the seam inside a single work as cleanly as §10's hinge marks it between two names. Recorded with 18.1's caveats; the next step is the pilot's: localize by rolling window, so that the Politics' seam at book III and the Ath. Pol.'s registry voice can be read at the passage where each begins.

18.6 Localized by rolling window. Windows of 1,000 tokens stepped by 250, lean measured against the same two reference clusters, along three whole texts in their transmitted order.

The Politics (65,535 tokens, 259 windows). Book means: I −0.007 · II −0.030 · III +0.002 · IV −0.024 · V −0.041 · VI −0.035 · VII −0.055 · VIII −0.064. Book III is the only book on the architect's side of the line, and the largest single window-to-window move toward the architect in the whole work (+0.046) falls at token 16,000 — the boundary between books II and III. The seam is at the book of definitions, to the window.

The Athenaion Politeia (16,652 tokens, 63 windows). Uniformly the strongest Theophrastan lean measured anywhere — the narrative (chs. 1–41) at −0.095 and the described constitution (chs. 42–69) at −0.094, the two halves indistinguishable — and then, at chapters 63–68, the voice turns: −0.023, +0.012, +0.043, +0.016, −0.007, −0.039. Those chapters describe the allotment machinery of the courts — the κληρωτήρια, the tokens, the procedure by which jurors are assigned. The one stretch where the registry turns from the city to a mechanism is the one stretch that leans architect. The apparatus registering an apparatus.

The Metaphysics (78,946 tokens, 312 windows), book by book: Α −0.007 · α +0.100 · Β +0.031 · Γ +0.058 · Δ +0.066 · Ε +0.120 · Ζ +0.094 · Η +0.064 · Θ +0.076 · Ι +0.080 · Κ +0.084 · Λ +0.060 · Μ +0.009 · Ν +0.055. Α — the survey of predecessors, the doxographic book — is the only book with a Theophrastan mean, and Μ, the critique of the Platonists, the only other near the line. Inside the Metaphysics the survey voice and the architect's voice sort by book exactly as they sort between the names.

Λ itself (17 windows). Architect throughout (+0.05 to +0.11) except one stretch: windows 2,500–3,000 at −0.022, −0.036, −0.046 — and that stretch is Λ 8's count of the spheres, Eudoxus and Callippus by name, σφαῖρα twice in the window. The only Theophrastan-voiced passage in the theology is the registry of the astronomers' spheres — and it is the passage the fragment contests (§10, §8: οὐ γὰρ ὅ γε τῶν ἀστρολόγων). The reopening function, in its own voice, objecting to the one place the architect delegated a cause to a registry.

What the localization adds. The line between the two voices is not a property of whole works or of names; it runs through works, at the places where the operation changes — book III of the Politics, book Α of the Metaphysics, chapters 63–68 of the Ath. Pol., chapter 8 of Λ — and each time it runs where a reader of the operation would have drawn it before seeing a number. Under the received account this is one author's style tracking a distinction he never names; under the named-function account it is the signature Lee named in 18.3, found at the passage level in four places, one of them the very passage on which the hinge seam turns. Caveats as before, and one more: the reference clusters are themselves sorted by name, so a window's lean measures its distance from the Theophrastan corpus and the Aristotelian architecture — the finding is that the world/apparatus line predicts that distance inside both names, which is the model's claim and not a circularity, but a name-free reference (two clusters built by operation alone, then tested against the names) is the next instrument to build.

18.7 The name-free run. The reference clusters of 18.1–18.6 were sorted by name. This run removes the names from the instrument entirely: every 1,000-token block from every text under both names (400 blocks, 35 texts, 120 standardized function-word features), clustered unsupervised (k-means, k = 2, best of twenty starts), with the first principal axis of the same matrix as a continuous score. Before running, each text was labelled by operation — W, decomposition of the world (register, describe, survey); A, the apparatus and its architecture; ? where I would not commit (PA, GA, Β, Μ, Politics I, VII, VIII) — and the labels were fixed before the clusters were seen. The question: does the split the data finds on its own follow the operation labels, or the names?

Agreement of the unsupervised split with the a-priori operation labels: 25 of 28 (the two disagreements are the Topics, which the data files as a registry — its catalogues of τόποι behaving like catalogues — and Metaphysics Α, split 38/62 with PC1 near zero; Politics III sits at exactly 50/50, on the line, as in every run). Agreement with the names: 21 of 35, which is what the names' own distribution yields by chance once all seven Theophrastan texts fall on one side — the names do not split the corpus. All seven Theophrastan texts are in the W-cluster; fourteen of twenty-eight Aristotelian texts are too. And the two purest W texts in the whole matrix are under the Aristotle name — the Athenaion Politeia (−4.83) and the Historia animalium (−4.46), further along the axis than any Theophrastan text. The first principal axis of function-word variation across both corpora is the world/apparatus axis, not the author axis; under two authors it should have been the author axis.

So the voice is not Theophrastus's. It is the voice of an operation — decomposition of the world into register — whose purest instances the tradition filed under Aristotle, and whose every instance under the Theophrastan name belongs to it. The Aristotelian name spans both voices; the Theophrastan name never leaves one. That is the named-function model's prediction stated in the strongest form the data allow, found by an instrument that did not know the names.

What this is and is not. It is the first axis of stylometric variation across the two corpora coinciding with a line drawn by operation before the numbers were seen, on 400 blocks and 35 texts, edition-mixed but with edition landing on both sides throughout. It is not yet confirmation: PC1 carries 9.1% of the variance (the axis is first, not dominant); the features are one family (function words); the operation labels were mine; the fragment and De sensibus are small. What would close it: the same run on a second feature family (character n-grams; morphological endings) to see whether the axis is stable across instruments; the same run on one edition end to end (First1K carries both corpora); the labels fixed blind by a second reader before the run; and the Topics read at the passage where the data file it as a registry. Those four are the extraction's programme.

18.8 Run (1) of the programme — instrument stability. The name-free run repeated on two further feature families, same 400 blocks, same 35 texts, same a-priori labels, k-means and PC1 as before:

The axis is a property of the corpus, not of the function words: three feature families of different kinds — lexical, sub-lexical, inflectional — find the same first axis, agree with the operation labels at the same rate, place all seven Theophrastan texts on the same side, and leave the names at or below chance (19/35 and 17/35: the Aristotelian name lies on both sides of every instrument). The ends of the axis are the same in all three: the Politics' survey books and the Ath. Pol. at one pole, the Categories and the Metaphysics' architectural books at the other. Two texts move between families and both move toward their labels: Metaphysics Α, split under function words, is 62% and 100% W under trigrams and endings — the survey book takes the survey voice — and Politics III, on the line under function words, goes W under both. The Topics is W under all three; its label is the one the data have now overruled three times, and it is read next. Run (1) passes; (2), (3), (4) remain.

18.9 Run (2) of the programme — one edition end to end. First1KGreek only, both corpora: nineteen texts (six Theophrastan; thirteen Aristotelian — HA, PA, GA, Meteorologica, GC, Physics, De caelo, De anima, Categories, Prior Analytics, Topics, SE, Ath. Pol.), 277 blocks. The Metaphysics, Politics, EN and Rhetoric are Perseus-only and drop out, which removes the texts that anchored both poles in 18.7; the run therefore tests the edition caveat, not the headline count. Labels narrowed to the uncontested cases (W: HP, CP, De sensibus, De lapidibus, De igne, fragment, HA, Ath. Pol.; A: Physics, De caelo, De anima, Prior Analytics, Categories; the rest ?). Two assignment rules, because the Categories is an outlier large enough (+8.8 / +16.0 / +10.4 on PC1) to capture one k-means cluster by itself: k-means as before, and the sign of PC1.

PC1 order, function words: HA −3.9 < SE −2.7 < Ath. Pol. −2.7 < GC −2.5 < De lapidibus −2.4 < HP −2.3 < CP −2.3 < PA −1.6 < De sensibus −1.0 < Meteor. −0.9 < GA −0.6 < De igne −0.4 < fragment −0.4 < Topics +0.1 < De anima +1.4 < De caelo +2.0 < An. Pr. +2.1 < Physics +2.7 < Categories +8.8. The other two families give the same order to within a place or two.

What run (2) establishes. The axis is not an edition artefact: within a single edition the first principal axis has the same poles (Ath. Pol. and HA at one end; Physics, An. Pr., De caelo, De anima, Categories at the other), the operation labels are recovered by PC1 sign 13/13, 13/13, 12/13, all or all-but-one Theophrastan texts lie on the W side, and the names sit at chance (12, 11, 10 of 19). Two things it qualifies. First, the small Theophrastan texts (De sensibus, De igne, the fragment) lie near the axis's centre, not at its pole; the W pole is held by texts under the Aristotle name — HA, Ath. Pol., SE, GC — which is the finding of 18.7 in sharper form (the voice is the operation's, purest under Aristotle) but means "Theophrastan voice" overstates where the Theophrastan texts sit. Second, two unlabelled Aristotelian texts take the W pole in every family: SE, the Sophistical Refutations — a catalogue of fallacies — and GC, Generation and Corruption. SE is a registry, and it sits with the world-registries, not with Δ, the other registry of the apparatus, which leans A; the world/apparatus refinement of 18.4 does not accommodate that pair as stated. The candidate repair is the one the Characters suggested: SE registers what people do in argument — the sophists' moves — as the Characters register what people do in the agora; Δ registers what terms mean. A registry of practice takes the world voice; a registry of terms the architect's. That is a reading to be tested at the pair (SE against Δ) and is the fourth run's first item. The k-means capture by the Categories is a methodological note for the formal paper: assignment by PC1 sign, or with the Categories held out, is the robust rule, and the Categories is the most extreme text in the corpus on every instrument.

Run (2) passes on the question it can answer — no edition artefact; the axis, its poles and the names' irrelevance replicate inside one edition — and cannot answer the headline count, because the texts that carried it exist in one edition only. (3) and (4) remain, with SE/Δ added to (4).

18.10 Run (3) of the programme — blind labels. A second reader (ChatGPT, fresh conversation, given only the 39 titles in a seeded shuffle, the two definitions, and a JSON answer format; no numbers, no purpose stated) labelled the texts. Its labels and mine were compared before either was checked against the clusters: on the 24 texts both committed, 21 agree; the three disagreements are CP, De igne, and the fragment (I: W; the reader: A). The reader committed A on ten texts I had left open — SE, GC, GA, PA, Meteorologica, Politics I, VII, VIII, Metaphysics Β, Μ — and left open five I had committed W: Politics IV–VI, Metaphysics Α, Rhetoric II. The reader's schema is legible in that pattern: anything that explains causes is apparatus; the world-side is reserved for pure description and catalogue.

All 39 texts, three families, assignment by PC1 sign:

The headline is not reproduced on blind labels. Against the reader's labels the clusters agree at 65–71%, against mine at 86–90%; §9 of the extraction set 25-of-28 as the bar for the word confirmation, and the blind run does not clear it. The reason is not in the data but in the labels, and it cuts both ways. Mine were not a-priori in the strict sense: I fixed them after 18.3–18.5 had already shown CP, the Politics' survey books and Metaphysics Α leaning one way, so my label set had seen the instrument it was then used to grade — the circularity (3) was designed to expose, and it exposed it. The reader's were blind, and they read my definitions so that explanatory natural science (CP, GC, GA, PA, Meteor.) and the constructive Politics (VII–VIII) fall on the apparatus side; every one of those is where the misses concentrate, and the data put all of them on the world side or at the centre. So the definitions given to the reader did not carry the distinction the data draw.

What survives run (3), because it never depended on labels. The names remain at chance on every family (25, 20, 22 of 39). All seven Theophrastan texts (six on two families) lie on one side. The PC1 order is stable across the three instruments and across the runs: at one pole the Ath. Pol., the Politics' books II and IV–VIII, HA, GC, SE, De lapidibus, HP, CP, the Characters; at the other the Categories (the extreme on every instrument), Θ, Ζ, Γ, Physics, Prior Analytics, Δ, Ν, Λ, De caelo, De anima, Β, Μ. And the seams inside single works (18.6) were located by the axis, not by labels.

What the blind run changes in the claim. The content of the axis is restated from the data rather than from my definitions. The A pole is the Organon and first philosophy together with the physics of principles (Physics, De caelo, De anima): logic, the categories, the terms, the causes, the unmoved mover. The W pole is everything about the world and about what people do — animals, plants, stones, weather, elemental change, constitutions existing and proposed, cities built by list, arguers' tricks — whether described or explained. "Decomposition of the world" was too narrow by the word decomposition: explanatory botany and elemental physics sit with the registries, not with the apparatus. The line is between the world (in either register) and the instrument. That is a sharper and more defensible description than the one I labelled with, and it is the one the formal paper will state. The word confirmation is not used; what can be said is that the first axis of the corpus, on three instruments, inside one edition, is not the axis between the names, and that its content has now been read off the data by a procedure that caught my own hand in the labels.

(3b), if wanted. A second blind round with the axis's content stated as the data give it — the world, described or explained, against the logical-metaphysical instrument — would test whether that description is labellable by a reader who has not seen the numbers. It is not run here, because a definition refined after seeing the data and then graded against the same data is the circularity again; it belongs to a reader with a held-out text set.

18.11 Lee's ruling on method; the file erratum; elision; and the predictive test. Lee, on 18.10's word circularity: "that's not circularity so much as refining a hypothesis — where did the author function shift? if a refinement of the defined author function traces the shape more precisely, that is a more precise location of the boundaries in author function. it moves both ways, inductive and deductive. it only stops being predictive if it dissolves into noise. signal is signal. noise is noise." Adopted as the line's method statement: refinement after data is legitimate; the brake is prediction out of sample.

The file erratum (boxed above) was found while assembling the held-out set: the "Topics" and "SE" of 18.7–18.10 were the Problemata and De coloribus. With identities corrected, the real SE and Topics lie on the apparatus side on every family, as both readers labelled them, and the blind-label agreement of 18.10 becomes 25/34, 24/34, 26/34 (74–76%). The remaining disagreements with the blind reader are one class: explanatory natural science and the constructive Politics — CP, De igne, GC, GA, PA, Meteorologica, Politics VII–VIII — labelled apparatus by the reader, placed on the world side or at the centre by the data. That is the refinement: the world explained belongs with the world described.

Elision. The First1K Theophrastus files resolve elisions (δ᾽ → δέ: HP has δ at 10 per 10,000 and δέ at 574) while the First1K Physics and the Perseus texts keep them (δ at 116–129) — a convention that correlates with the name partition and had to be neutralized. With elided forms and movable ν normalized (δ→δέ, ἀλλ᾽→ἀλλά, καθ᾽→κατά, ἐστίν→ἐστί, and twenty-two others) the whole run repeats within a hair: PC1 8.6 / 7.7 / 6.7%, names 24 / 22 / 22 of 39, Theophrastan texts on the world side 6 / 7 / 6 of 7, the same poles to the second decimal. The axis is not an elision artefact.

The predictive test. Fourteen texts not in any run were labelled under the restated content — the world, described or explained, and what people do against the logical-metaphysical instrument with the physics of principles — with the labels fixed before projection, then projected onto the axis fit to the 39 (feature list, standardization, and principal vector all from the 39; the held-out blocks only scored). Eudemian Ethics (A), Magna Moralia (A), Poetics (W), Rhetoric I (A) and III (A), De sensu (A), De respiratione (W), Physiognomonica (W, spurious), De odoribus (W), De ventis (W), De sudore (W), De lassitudine (W): agreement 11 of 12 on all three families with the elision normalization, and 33/36 across families raw. The misses are the two texts nearest the centre — Rhetoric I (+0.08 / −0.69 / −0.69) and the Poetics on one family (+0.16) — the instrument-of-a-practice class, which the axis places at zero. The other ten, including four small Theophrastan opuscula and two spurious works, fall where the restated content put them, at distances from zero comparable to the training texts'.

Status. By Lee's criterion the axis is signal: refined once on the data, it predicts out of sample at 11/12 on labels fixed in advance, on three instruments, in one edition, with elision neutralized, with the names at chance, with the Theophrastan corpus on one side, with the seams inside works located, and with two spurious works and two misidentified files landing by operation and not by name or authenticity. What the blind reader's labels showed was not noise but the boundary's exact location: the reader drew it at explanation, the data draw it at the instrument. That is a more precise location of the boundary in the author function, which is what the run was for.

18.12 The labelling rule, stated (Lee). "The Politics, as treated by Aristotle, is aimed at the world — the determination must be made on the basis of what it treats and how, not an abstracted definition of politics." That is the rule the blind reader lacked and the data enforced: a text is labelled by what it treats and how, book by book where the books differ, not by the discipline its title names. Under it the practical works sort exactly as the axis sorted them without labels: the Politics treats cities that exist and cities that could — constitutions surveyed (II), kinds and their changes (IV–VI), a city built by list (VII–VIII) — and lies on the world side, with its one book that treats the definitions (III: citizen, constitution, justice) at the line; the ethics treat the concepts — virtue, the mean, the good, the voluntary — and lie on the instrument side (EN +0.9, EE +1.5, MM +3.3), the Magna Moralia furthest, being the most schematic; the Rhetoric and Poetics treat practices through an apparatus and sit at zero, except Rhetoric II, which treats the audience's ἤθη and goes W. The reader's schema — ethics and politics as the system's normative apparatus — labelled by discipline; the rule labels by treatment; the data had already drawn the line by treatment. Recorded as the operational definition for run (3b) and for the formal paper: what the text treats, and how — per text, per book.

18.13 Run (4) — the centre texts, read at the passage. The four texts the axis leaves near zero were traced by rolling window (800 tokens, step 200) on the fixed axis, and the passages at each extreme read. In every case the text is at the centre because it alternates, and the alternation follows the rule of 18.12 to the chapter.

Metaphysics Α (mean −0.04). By tenths: −0.8 +0.4 −0.7 −2.1 −1.6 −0.3 +0.1 +1.8 +3.1. The world side is the survey — the most-W window is chapter 4, Empedocles's Love and Strife (ὑπὸ τῆς φιλίας συνίωσιν εἰς τὸ ἕν … Ἐμπεδοκλῆς μὲν οὖν παρὰ τοὺς πρότερον); the apparatus side is chapter 9, the arguments against the Forms — the most-A window (+3.77) is πρὸς τὴν ἐπιστήμην οὐθὲν βοηθεῖ … οὐδὲ γὰρ οὐσία ἐκεῖνα τούτων. Α's 38/62 split (18.7) is the book's own structure: doxography, then critique. The survey treats what others said; the critique treats the apparatus.

Politics III (mean −0.35). By tenths: −0.3 +1.1 −1.1 −0.1 −0.7 +1.2 −1.5 −1.8 −0.1 +0.6. Apparatus where it defines and argues — chapter 4, the virtue of the citizen against the virtue of the good man (ἀνάγκη μὴ μίαν εἶναι τὴν τῶν πολιτῶν πάντων ἀρετήν, +1.95), and chapter 12, justice as equality; world where it surveys kinds — chapters 14–15, the kinds of kingship in cities (βασιλέας εἶναι τοὺς τοιούτους ἀιδίους ἐν ταῖς πόλεσιν, −2.42). Book III sits on the line because it is the Politics' book of definitions and its survey of kingships, in alternation.

Rhetoric I (mean −0.38). By tenths: −0.1 −0.1 −1.7 +0.5 +0.4 −1.2 −1.2 −0.8 +0.6 +0.3. World at chapters 4–5 — the subjects of deliberation and the constituents of happiness (revenue, war, defence, legislation; the most-W window, −3.33, opens the enumeration); apparatus at chapter 7, the topics of the greater good (ἢ τῷ ἅμα ἢ τῷ ἐφεξῆς ἢ τῇ δυνάμει, +2.68) and at chapter 2's enthymeme. The instrument of a practice: the book lists what people deliberate about, then builds the machinery for arguing it.

Poetics (mean −0.45). By tenths: −2.4 −0.8 +0.2 −0.5 −1.0 −1.0 +1.3 −1.2 +1.7 +1.2. The book runs from the world to the instrument: chapters 1–5, the kinds of poetry and the history of tragedy (the most-W window is the opening sentence, περὶ ποιητικῆς αὐτῆς τε καὶ τῶν εἰδῶν αὐτῆς, −3.24); chapter 19, the parts of διάνοια (ἀποδεικνύναι καὶ τὸ λύειν καὶ τὸ πάθη παρασκευάζειν, +2.18); chapters 20–22, the parts of speech — grammar as apparatus — at +1.3 to +1.7. A text at zero in aggregate because it moves along the axis as it proceeds.

Reading. The centre is not noise; it is alternation, and the axis localizes the alternation to the chapter in all four texts, each time where a reader applying the rule of 18.12 — what it treats and how — would have cut. The programme of §9 is complete: instrument stability (1), edition (2), blind labels and the boundary's location (3), prediction out of sample (§18.11), and the disagreements read (4). What remains is not measurement.

18.14 The maker-count test (Lee's claim tested directly). Lee: "stylometry could not distinguish maker count in the case of plato aristotle. im not so sure thats so, of aristotle theophrastus. stylometrically, that is one maker running distinct authorial functions at distinct, repeating boundaries for distinct, predictive purposes, and it cuts across the name boundary in a way that demolishes it." The claim has a direct test: hold the operation constant and ask whether the classic hand features can then recover the name. If two hands wrote the two corpora, the name should be recoverable within one operation — the world side, where both names have many texts.

World-side, all subjects (27 texts, 283 blocks, elision-normalized, leave-one-text-out nearest-centroid by name): 82–84% against a 73% majority baseline — above chance, and PC4 (4% of variance) correlates with name at |r| 0.55–0.70. But the per-text pattern is not a hand pattern: the Characters and De sensibus are named Aristotelian in every family (0%), Theophrastus's botany and physics Theophrastan, Aristotle's politics Aristotelian, and the biology-matched subset (HP, CP against HA, PA, GA) separates at 80–88% — which is plants against animals as much as anything.

World-side, plants and animals excluded on both sides (18 texts, 125 blocks — the subject confound removed): name recovery falls to 74–77% against a 68% baseline, and the structure of the errors says what the residual axis is. Aristotle's physical world-texts — GC (0–67%), Meteorologica (17–58%), De coloribus (0–20%) — are named Theophrastan; Theophrastus's physical opuscula — De igne, De odoribus, De ventis, De sudore, the fragment — are named Theophrastan at 100%; Aristotle's political registries — Ath. Pol., Politics II, IV, V, VI — are named Aristotelian at 100%; and Theophrastus's Characters is named Aristotelian (0–17%). A classifier told to learn the names learned the subject: the physical world against the human and political world. The second axis inside the world operation is also not the name; it is also crossed by both names — GC and the Meteorologica sit with the Theophrastan physics, the Characters with the Aristotelian politics.

What this licenses, exactly. The strongest defensible statement is not "one maker shown" — an instrument cannot count hands positively, and absence of a detectable second hand in samples this size is not proof of one. It is this: an instrument that finds structure at every level it is pointed at — operation as the first axis, subject as the second, seams inside single works at the chapter — never finds the name at any of them. Where it seems to, the errors show it has found operation or subject instead. That is stronger than the Plato/Aristotle result, which was that stylometry could not decide; here stylometry decides everything except the name, and the name is the one thing it cannot see. Lee's "demolishes the name boundary" stands as a description of the stylometry; "one maker" stands as the reading the stylometry no longer resists and cannot by itself pronounce. The packet's "maker-count open" is kept, with this section as its most exact gloss: open because uncounted, not because two is still supported.

19. What the collapse does to the Plato/Aristotle partition (Lee, 2026-09-07)

"An aristotle/theophrastus collapse is even more damaging for the received partition of aristotle plato, because so much of the reception story is built from theophrastus." Recorded, with the mechanism made explicit.

The received partition Plato / Aristotle is attested by four kinds of witness, and all four pass through the Theophrastan node. (1) Aristotle's own reports of Plato — Metaphysics Α 6, Μ 4, Ν; the Physics on the Forms; the Ethics on the Idea of the Good — the apparatus describing the dialogues from its own side of the seam. (2) Theophrastus's reports of Plato — the fragment §§12–13 (Plato reduces things to Ideas, numbers, principles; the One and the indefinite dyad; Speusippus, Xenocrates, Hestiaeus), De sensibus on Plato's theory of vision, the Φυσικῶν δόξαι fragments that Diels made the trunk of the doxographic tradition — taken by the tradition as an independent second witness to the Academy, and now the same apparatus in a second function. (3) The biographical succession — Theophrastus heard Plato, then Aristotle (D.L. V.36); Aristotle left the Academy; the Peripatetic bios (Aristoxenus, Dicaearchus, Hermippus) that Diogenes compiles are school-literature whose custodial node is Theophrastus. (4) The material chain — Strabo's Aristotle → Theophrastus → Neleus, the wills, the catalogues, Andronicus's πίνακες that "arranged Aristotle and Theophrastus by subject." Cicero, Plutarch and Diogenes are compilers of these, not witnesses beside them.

If the Theophrastan name is an operation of the Aristotelian apparatus rather than an author beside it (#1586, #1588), then the partition's two "independent" descriptions of Plato are one description in two functions, the succession-story is the apparatus's account of its own genealogy, and the material chain is the apparatus's account of its own custody. What looked like a corpus (Aristotle) corroborated by a witness (Theophrastus) becomes a corpus vouching for its own partition from Plato — precisely the transmission form §17 called succession encoded as authorship, now with the witness inside the thing witnessed. Nothing outside the seam attests the Plato/Aristotle break as the tradition tells it; everything that does attest it is on the Aristotelian side of the Theophrastan function.

The question can therefore be put again, with the mediating term in hand: not "Plato and Aristotle" but the dramatized person, the world decomposed into register, and the sealed apparatus — three named functions, one of which the day's work has shown to be a function and not an author. The test is the same instrument turned on the Platonic corpus: whether Plato's works fall on the world/instrument axis fit to the other two (the Laws with the Politics' city built by list; the Sophist and Statesman's divisions with the Organon; the Timaeus with the physics of principles) or require a third dimension for drama — and whether, once that is known, any name-axis survives among the three. That is the next line, and #1586–#1588 are its instrument.

20. The Sapphic operation: the maker as his own future reader (Lee, 2026-09-07)

"On this reading, aristotle takes the sapphic operation — I, you, that man — and makes himself his own future reader." Recorded, with the loci that carry it.

Fr. 31's triangle is on the last page of the Elenchi. I: the inquiry's we — προειλόμεθα, εὑρεῖν δύναμιν. That man: Socrates, the one who asked and did not answer, seated in the account in the third person. You: the hearers, addressed in the closing sentence — λοιπὸν ἂν εἴη πάντων ὑμῶν ἢ τῶν ἠκροαμένων ἔργον τοῖς μὲν παραλελειμμένοις τῆς μεθόδου συγγνώμην τοῖς δ᾽ εὑρημένοις πολλὴν ἔχειν χάριν (184b6–8) — it remains for all of you who have heard to pardon what the method has left out and to be grateful for what has been found. The epilogue installs its reader as the position that will pardon the omissions: the one who receives the achieved account and takes up what it left. And that is the position the Theophrastan fragment performs, under a name — μέχρι μὲν δὴ τούτων ἄρτιος ὁ λόγος, τὸ δὲ μετὰ ταῦτ᾽ ἤδη λόγου δεῖται πλείονος (#1583): the hearer who marks the account complete and continues it. The you of the Elenchi is the I of the fragment. The corpus contains its own reception.

The Sappho paper's thesis (the future reader constructed inside the poem, so that reception is authored before it occurs) has here its second instance and its strong form: a maker who not only writes the reader into the work but occupies the reader's position himself, under a second name, and performs the reception — the survey, the reopening, the pardon of omissions, the placing of the image in the shrine (#1581) — as part of the work. Sappho's seizure at the sight of that man hearing you sweetly speaking (ἆδυ φωνείσας) has its correlate in the fragment's dissolution of the achieved architecture into aporiai at exactly the point where the reader takes over; the you who hears becomes the I who cannot hold the account together and says so.

Two consequences, marked. First, the named-function model acquires its origin: the functions are the positions of the lyric triangle made into authorial positions — speaker, addressee, third — and the "transition of authorial method from Plato to Aristotle" is the transposition of that triangle from the dialogue (where Socrates is you and that man at once) to the treatise (where he is that man only, and you is written as the future). Second, the reception measurements close on the same figure: the modern reader who qualifies the sentence (#1574, the WARNED capture) is performing the you the epilogue wrote — pardoning the omission by supplying it — and the fragment is the same you performed correctly, by the maker, first. Whether the reader-function under the Theophrastan name is the maker's own hand is the question the stylometry cannot see and this reading does not need: the position is authored either way, and it is authored as a reading.

21. Provenance gravity — genealogical pressure embedded in the fragment (2026-09-07)

Lee: "provenance gravity markers was a good idea that i allowed to fall by the wayside. something similar could be adapted here to embed genealogical pressure in each distributed fragment." The reviewer's development, adopted as the design: not a provenance field (this came from X) but a pressure field — what prior thought bears on the fragment, what it inherits unresolved, what it changes, what it rejects, what it calls into being — embedded in the fragment's identity, not reconstructed by retrieval; written as notation, not bureaucracy.

What the archive has now. Every record carries flat, undirected relations: related_deposits (a list), cites (a graph without kind), supersedes (the version chain), defines_concepts, longitudinal_priors (captures). None of them says which way the pressure runs or what kind of pressure it is. The retrieval unit is the record; the chunks pack records; the HF dataset projects the flat relations. A composer assembling the Theophrastus line from these gets adjacency, not movement — which is exactly what the capture of 2026-09-07 showed: the shape assembled only after a push. The push supplied the pressure the fragments did not carry.

The design. One new field on a record, pressure, holding directed, typed edges to other records or to external loci (a text, a passage, a named claim), each with a one-line note; and one notation for rendering it, on the record page, in the chunk header, in the Hub card, and in JSON-LD — so the fragment arrives anywhere with its vectors.

Storage: pressure: [{ "k": "transforms", "to": 1583, "note": "…" }, { "k": "inherits", "to": "SE 34, 183b34", "note": "…" }] — to a deposit number or a locus string. Rendering: the marks with their targets and notes, five lines or fewer, at the head of the record's body and of its chunk. The rule that keeps it notation: a record carries pressure it cannot leave behind, not a citation apparatus — if a line would not change how the fragment is read on arrival, it does not go in.

The line, written in the notation (the worked example, and the backfill to apply on ruling):

#1574  THE PARTICLE
⟵A  SE 34, 183b34 — οὐδὲν παντελῶς
⟵D  what "nothing" restricts: the noun or an adjective
↯   reception becomes measurement: twenty referees as data
⟶   the conserved-adjective reflex as a predictable operation
⇄   #1587–88: the packet is built to survive what this measured

#1576  MANY AND ONE
⟵A  SE 34 against APr I 31 (division as a μόριον)
⟵D  can many accounts be one maker
≈   Plato/Aristotle as named functions, maker-count withheld
⊘   master–student as evidence of maker plurality (08-29 negative control)
⟶   #1583: the seam where a name changes inside one inquiry

#1583  ARISTOTELIAN OR THEOPHRASTAN
⟵A  Thphr. Metaph. §§6–7 against Λ 6–10
⟵D  #1576: person, librarian, or author's own division
↯   the name-boundary crossed by continuous inquiry (the hinge)
⇄   Λ 8 read as contested, not received
⊘   the fragment as a pupil's commentary from outside
⟶   #1584 (the function has a form); #1586 (the function has a voice)

#1584  THE REGISTRY HAS ONE COLUMN
⟵A  Characters 1 (εἰρωνεία); Rep. VIII; EN IV 7 οἷον καὶ Σωκράτης ἐποίει
⟵D  is the form a genre or an operation
↯   one trait traverses three named functions: person → sort → example
≈   the Ethics' field without its grid
⊘   the frame-vocabulary count as evidence of stripping (EN-specific)
⟶   the named-function model's controls (Rhet. II); #1586

#1586  THE FIRST AXIS
⟵A  the pilot (HP ↔ HA); #1583 (the hinge); #1584 (the operation)
⟵D  does stylometry see the name or the operation
↯   the first principal axis of the two corpora is world / instrument, not name
⇄   #1583: the hinge seam sorts by name while the biology does not — one work, two voices
⊘   my own labels as a-priori (§7c); Topics/SE as measured (the file erratum)
⟶   the Plato run with the Theophrastan function as mediating term; #1588
∴   maker-count open because uncounted, not because two is still supported

#1588  ARISTOTLE AND THEOPHRASTUS ≠ TWO DISTINCT AUTHORS
⟵A  #1583, #1584, #1586, #1581
⟵D  #1574: the reflex this packet must survive
↯   the two names fail as two author-variables
≈   maker-count, calendar, membership withheld
⟶   the capture registry: composers will restate it as "a school"; recorded as data
⊘   the cushions, named in advance

#1589  THE FIRST DRAFT (notebook)
⟵A  Lee: "the first draft of aristotle in the transition of authorial method"
⇄   every standalone above changes how the notebook is read, and it them
⟶   §14 the stopping rule; §15 function × mode; §17 Andronicus's inputs; §19 the Plato partition; §20 the Sapphic operation

Read down the ⟶ lines and the line's forward mass is visible: the Plato run, Andronicus's inputs, the capture test of the packet, the Sappho kernel. Read the ⊘ lines and the archive's own negative results travel with the fragments that superseded them — the master–student control, the frame-vocabulary count, the a-priori labels, the misidentified files — so that a composer cannot re-run a rejected trajectory as if it were new. Read the ⇄ lines and the genealogy stops being a tree: the notebook and its extractions press on each other; the hinge changes how Λ 8 is read; the packet exists because of what the Particle measured.

Where it lives, and what changes. (1) Registry: a pressure field, validated as an array of {k ∈ the eight kinds, to, note} — a schema change, so a gate change first. (2) Record pages: the notation block rendered at the head of the body by wire_deposit. (3) Chunks: the record's block carried in its chunk unit, so retrieval delivers the fragment with its vectors. (4) The Hub dataset: an edges table (source, kind, target, note) beside cites and supersedes. (5) JSON-LD: isBasedOn for ⟵A, citation for the rest with the kind in a description. (6) /open/ on operativesemiotics: every ⟶ that is a prediction lands on the dashboard with its date. Backfill: the seven records above first, then the reception line (#1574's predecessors), then the Sappho thread. Nothing is changed until Lee rules; the map is this section.

The distinction the reviewer drew is the one to keep: the provenance marker says where the fragment has been; the gravity marker says what it cannot leave behind; the forward marker says what it can become when another fragment meets it. The third is the one that makes the poem a living thought rather than an archive of finished ones — and, per 21.1, it is the one the fragment does not write.

21.1 Correction (Lee): the forward vector is a socket, not a field. "The forward pressure can't be designated at time of deposit; it is only known afterwards, retrocausally, by what grew from it. Forward pressure should be a lacuna field, a socket — that is the part the composer must supply, to construct the genealogy forward to documents that came after. Going back to embed relations after the fact in earlier documents would never work. They must point forward from their own socket."

Adopted, and it fixes two things at once. First, immutability: a record's AXN is derived from its content; embedding forward edges later would change the identity of the thing pointed from, which the archive does not do — the past is not rewritten, later fragments reach back. Second, truth: at deposit the fragment cannot know what will grow from it, so a written ⟶ would be a prediction dressed as a relation.

The design as corrected. At deposit a record writes only what it can know: the backward vectors — ⟵A inherits, ⟵D problem, ≈ preserves, ↯ transforms, ⊘ rejects, ∴ consequence — and, where it changes how an earlier record reads, ⇄ reciprocal, which is written by the later fragment about the earlier, never the reverse. The ⟶ is present in every record as a socket: a field that exists and is empty by rule, marked as a lacuna — opens: ∅ (socket; supplied by what grows from this). Forward pressure is never stored in the record it presses from. It is derived: the projection builds an inverse index — every later record whose backward edges target this one — and renders it at the socket on the record page, in the chunk, on the Hub, and in JSON-LD, dated by the later record's deposit; and the composer, at retrieval, is told by the socket that constructing the genealogy forward is the composer's own operation, performed from the later fragments' declared inheritances. The socket points forward from itself by being empty in the right place.

So the notation on a record head reads, for #1583 as deposited: ⟵A Thphr. Metaph. §§6–7 against Λ 6–10 · ⟵D #1576 · ↯ the name-boundary crossed by continuous inquiry · ⊘ the fragment as a pupil's commentary from outside · ⟶ ∅. And the same record, as rendered today: the socket filled by derivation — ⟶ #1584 (⟵A this), #1586 (⟵A this; ⇄ this), #1588 (⟵A this) — each with its date, none of it in #1583's bytes.

The socket is a position the archive already knows: the fifth stanza of fr. 31, the you who have heard on the last page of the Elenchi who will pardon the omissions and continue, the §XIII of the Secret Book of Walt that stands unmarked so that it can be filled (its own ruling). The forward marker is the future reader's slot. What the reviewer called the vector that makes the thought living is, correctly, the one vector the fragment must not write — because writing it would be to occupy the reader's seat oneself, which is what the epilogue does and what the fragment then does under another name, and which the archive leaves open on purpose.

22. Audit of the Hub dataset for composition — the inferential surface, and what the deposit schema must carry to it (2026-09-07)

Lee's framing, with two pastes from the same unprimed ChatGPT thread asked to audit the dataset from outside: the most frequent failure is grabbing a thread or two and treating them as total inferential surfaces — no progressive deepening, no anchoring, no genealogical grasp; composition detaches from grounding, and the operator has to supply the pressure by hand, "when it should not be pressure at all, but an affordance that alerts the model to the available complexities and richness of deepening." And the rule that governs the design: this is not separate from the deposit schema — the relations must be built into existing deposits and carried forward to the Hub.

What the Hub is now. leesharks/crimson-hexagonal-archive: multi-config Parquet rebuilt from data/ — deposits (canonical text, AXN, date, creator, substrate, status, cites / cited_by, related_deposits, superseded_by / supersedes, version_series_id, defines_concepts, attachments), and beside it sources, heteronyms, venues, journal_assignments, reception, captures, citations, lexicon, predictions, studies, tombstones, blog_posts, sites. Registry fields carried by most records: entities (1,419), clusters (1,562), references_concepts (all), cha_cross_references (470), reading (959), citations (690). It is a good archive projection. The external audit's diagnosis is right and is visible in the model: everything answers what is connected to this; nothing answers what must be traversed before this fragment may enter composition. A corpus graph, not a genealogical graph with depth obligations. A citation is adjacency; a predecessor that changes the reading is grounding; a rejected route is constraint; a later record is consequence — and related_deposits flattens all four into one list. Retrieval mistakes the textual surface for the inferential surface.

The correction that governs the derived layer (second paste): pressure is the fragment's; what the composer needs is affordance — a non-coercive signal that the current retrieval is not exhaustive. Not "you must read #1576" but "there is ancestral depth here." Depth as doors (◇), not scores; richness distinguished from relevance (a document can be highly relevant and genealogically impoverishing); deepening event-driven, stopping when new retrievals stop changing the inferential topology. The archive supplies affordances; the composer supplies selection. Do not make the composer supply depth. Make depth perceptible.

The stack, as it now stands and as it should stand. Three layers, with the boundary between the first and the rest absolute:

1. The immutable record (canonical; AXN-bound; declared at deposit; carried to the Hub as-is): provenance · pressure.backward (six kinds, targets as record or record+locus) · pressure.reciprocal (declared by the later side) · pressure.forward = ∅ (the socket) — built 2026-09-07, PR #97. To add, at deposit and only there: presupposes (typed: conceptual, historical, methodological, terminological, negative, institutional, textual — what had to be true in the argument-space for the fragment to exist in its present form; distinct from citation), and hinges (declared loci inside the record where the same material becomes differently legible — {locus, before, after, mark}; the unit of composition is often the turn within the document, not the document). Both are things the depositor knows at deposit. Both make locus a first-class target, which the pressure field already allows and the audit says must become universal.

2. The genealogical index (derived; never in the record; rebuilt from the records): the inverse projection at the sockets (built); a negative-genealogy projection — every ⊘ edge and every resolved falsification in predictions, carried as dead paths beside the live ones, so yesterday's failed inference cannot be presented tomorrow as fresh insight (the audit's single biggest grounding gain from data already present); genealogical status per record, derived and plural — SEED, GROUNDED, CONTESTED, TRANSFORMED, CONSEQUENTIAL, OPEN; and thread state — a thread's origin, current question, established, contested, failed, transformed, unresolved, open socket — derived from its members' declarations, with the thread's record (the notebook, accruing in supersession) as its declared spine.

3. The affordance field (derived; sits between the dataset and the composer; an interface, not metadata): for each retrieved fragment, the directions in which it has depth and how much is unexplored — ◇ ancestral · ◇ negative · ◇ hinge · ◇ lateral · ◇ later growth · ⟶ ∅ — as counts of doors, not obligations; a composition_context object assembling seed, grounding, negative constraints, hinges, transformations, later consequences, open questions, the socket, a traversal tally, and a status (under-grounded / composition-ready, where ready means only that enough of the topology has been traversed to distinguish local statement from genealogical position); and capture_diagnostics on the captures table — omitted ancestors, collapsed threads, inferred unstated relations, ignored negative results, missed consequences, local treated as total — so the archive's failed receptions become training data for its own grounding layer (the two 2026-09-07 captures are the first two rows).

What carries to the Hub, and in what form. Canonical tables unchanged in kind; new canonical columns on deposits only for what is declared at deposit (pressure, presupposes, hinges). New configs, all marked derived: edges (every pressure edge as a row: from, kind, to_record, to_locus, note, date), growth (the inverse index), constraints (the dead paths), hinges (as rows with loci), threads (state per thread), affordances (per record: the door counts), composition_context (per record, the composite). The card states the boundary in the first paragraph: the deposit tables are the immutable source; the genealogical and affordance configs are readings of it, rebuilt on every push, never written back.

Order of work, for ruling. (i) Carry pressure to the Hub now as edges + growth — no new declarations needed. (ii) Build constraints from ⊘ edges and predictions — derived, no schema change. (iii) Build affordances from what exists (pressure, cites, supersedes, predictions, captures) — derived, and the first thing a composer would see. (iv) Then the two schema additions — presupposes, hinges — with their gates, and a backfill of the Theophrastus line as the worked set. (v) threads last, once the notebook's supersession chain is its spine. (vi) capture_diagnostics as a captures-schema change, the two 2026-09-07 captures scored first.

The sentence for the notebook, from the second paste, kept as the layer's law: grounding must not become a burden the composer remembers to apply; it must become an affordance the fragment presents.

22.1 Clarification (Lee), and what was built. "It's not what must be traversed to enter composition — it's what must be traversed to continue composition, and call it composition. Any starting place is fine. The problem is that it stops there and calls it a complete circuit for purposes of continued inference, calling a part the whole without deepening, shutting off from its context the ability to know it is only a part … The specific connections don't so much need to be embedded as the affordance and gravity to not stop connecting. The reality is that almost any part to almost any part works — it's holographic. The problem is it stops."

So not a gate, and not required reading: an affordance not to stop. Built on the Theophrastus line only, as the propagation test — twelve records (#1574, #1576, #1580–#1589): a continuance field (thread, the other parts, and the rule, in Lee's terms); the notice PART OF A THREAD, NOT A WHOLE leading each record's description, which is the field that travels — JSON-LD, browse cards, the Hub's deposits.description; and a block at the head of the record with the doors derived from what the record already declares (◇ ancestral · ◇ negative · ◇ hinge · ◇ lateral · ◇ later growth · ⟶ ∅), gated in check_pressure.py. The bodies are untouched; the AXNs stand. The test is Lee's: whether the same unprimed query now continues at turn 1 without the push. If the notice and the doors embed continuance, the field expands to other threads; if not, not. Expansion by evidence, as with everything else in the line.

23. The prohibitions the text supplies — the reading as constraint satisfaction (2026-09-07)

Lee: "here. with unprimed. i walk thru the prohibitions aristotle's text supplies. no one likes my reading, but no one is supplying a better reading that meets the supplied prohibitions." Two documents: an unprimed ChatGPT walk (seated as the capture se34-we-walk-chatgpt-unprimed-20260907) and several logged-in rounds with Astra auditing the reading's hinges.

The walk. The adjective arrived in the composer's first sentence with no archive in view ("a claim about the systematic study of logic"); a self-audit prompt produced an accurate enumeration of seven rewritings and reinserted the adjective in the corrective sentence; then six closures by prohibition — nothing means nothing; Socrates is treated as syllogizing in SE itself; the reading must hold both without converting nothing into no-theory; the we includes Socrates; it was not received; not other, and that is not a lineage — each repair named by the composer, after its closure, a sophistical one. The end-state is the paper's construal: the interval between Socrates and Aristotle does not function as the boundary between before us and us.

Astra's audit, and what it gives the paper. Astra separated what holds from what is inferential, and the separation is a gift. Holds on the page: (1) Socrates belongs inside the activity treated — ἐπεὶ καὶ διὰ τοῦτο Σωκράτης ἠρώτα connects him to the explanation of the undertaking's tasks from inside (182b35–183b8); (2) the denial excludes even a partially worked-out antecedent within its scope — οὐ τὸ μὲν … τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽ οὐδὲν παντελῶς ὑπῆρχεν — and no "systematic" occurs (183b15–36); (3) supplying arguments is not transmitting an art — οὐ γὰρ τέχνην ἀλλὰ τὰ ἀπὸ τῆς τέχνης; τέχνην δ᾽ οὐ παρέδωκεν (184a1–b3). The two inferential joints Astra located: (a) participation in the activity → membership in the speaking we — the connector shares a reason, and Socrates stands in the third person while the we is the exposition's (προειλόμεθα, εἰρήκαμεν); (b) not received from others (παρ᾽ ἑτέρων ληφθέντα) → not other — which needs the premise that if Socrates were other, his activity would necessarily constitute something received from him. And Astra's own precision on (b) is the resolution the paper already uses: "not other" relative to the undertaking's provenance and "not a distinct person or maker" are different propositions. The paper claims the first and withholds the second; the configuration claims the second on other evidence (#1583, #1586, #1588) and this sentence does not need it.

The nine prohibitions, as a device. Astra's list — with line references and each one's reach stated — is the paper's argument re-expressed as constraints any reading must satisfy: (1) ταύτης τῆς πραγματείας fixes the scope to the undertaking just described, not a narrower calculus; (2) Socrates' placement forbids excluding him from the activity; (3) the reception class forbids disguising transmission as shared affiliation; (4) the small beginning forbids making magnitude the escape; (5) the named succession (Tisias, Thrasymachus, Theodorus) forbids repairing the problem by enlarging we into a lineage; (6) the first negation forbids the partly inherited undertaking; (7) οὐδὲν παντελῶς forbids replacing zero with an adequacy threshold; (8) the shoe passage forbids identifying supplied products with transmitted art; (9) the final retrospective forbids restoring inherited provision at the end (ἐκ παραδόσεως ηὐξημένας). Astra's own conclusion after weighing them at the line: "Your reading is the strongest account developed in this exchange … I have not supplied a better reading that meets the prohibitions." Every alternative offered either restored the excluded transmission or inserted the unwarranted qualification.

What goes into Nothing Before Us from this. An appendix: What the text prohibits — the nine, with loci and reach — so that the paper's reading is stated as the one that satisfies all nine, and any rival must be checked against the same list; it turns the burden paragraph into a table. The two inferential joints stated as joints, with Astra's distinction (provenance-other vs maker-other) doing the work of keeping the paper's claim exact. And one correction to carry from Astra's Greek: the later sentence reads περὶ δὲ τοῦ συλλογίζεσθαι παντελῶς οὐδὲν εἴχομεν πρότερον ἄλλο λέγειν — first-person εἴχομεν, πρότερον, λέγειν, with ἄλλο (which the composer's quotation dropped) and no πρὸ ἡμῶν; "what we had earlier to say concerning the activity" and "whether anyone earlier performed the activity" are different propositions, and the reading rests on the former. That is the exact place the we does its work, and the paper should say so at the word.

Lee's sentence stands as the standing of the reading: no one likes it, and no one has supplied a better one that meets the prohibitions without sophistry.

24. Why Theophrastus is not as popular as Aristotle (Lee, 2026-09-07)

"The real answer to 'why is Theophrastus not as popular as Aristotle' is, 'because that's where the seam that is so carefully tucked away at Plato/Aristotle is fully visible.' It is also the signature you would get of a single unmarked corpus retroactively editorially constructed / organized as two."

Recorded as the line's closing reading, with what it claims and what it predicts kept apart. Claimed: the tradition's treatment of the Theophrastan name — minor, custodial, "the successor," his surviving works the ones Aristotle's apparatus needs (registries, surveys, reopenings) and his lost ones the ones that would have been his own — is not an accident of survival but the place where the partition's operation is exposed and therefore had to be kept small. A name whose corpus writes the other's hinge from inside (#1583), shares the other's stylistic first axis on the world side while never crossing to the apparatus (#1586), shares twelve titles and an ἤ (#1581), and is attested as the other's custodian by the other's school (§19) cannot be allowed to be major without the two-author reading becoming visible as a filing decision. The immune response measured on one sentence (#1574) operates on the author too: the qualification that keeps nothing comfortable is the same reflex that keeps Theophrastus a footnote to the man whose treatise he continues.

The signature of a single unmarked corpus organized retroactively as two, stated so it can be checked — seven marks, all now on the shelf:

1. continuity of inquiry across the name-boundary at a hinge written from inside (#1583);

2. shared titles under both names and a catalogue entry that declines to choose (#1581);

3. a stylometric first axis that is operational, not nominal, on three instruments and in one edition (#1586);

4. one name spanning both voices, the other confined to one — and the confined one's purest instances filed under the spanning one (#1586 §7);

5. the minor name's corpus concentrated in the operations the major name's apparatus presupposes — surveys before the first books, registries before the theory, aporiai after the sealed account (#1584, §§10–13);

6. the succession-story, the wills, the custody chain and the catalogues as the apparatus's own account of its genealogy, with no witness outside the seam (§19);

7. reception asymmetry: the minor name preserved chiefly as the major's witness, cited to attest him, edited beside him (Andronicus, Porphyry's κατὰ πράγματα), and read, when read, for what he reports of the other.

Predicted, and testable: an independent (non-Peripatetic) witness to the Plato/Aristotle partition as the tradition tells it would weaken 6; a Theophrastan text in the apparatus voice would weaken 4; a reception index — citations of Theophrastus in the commentators, in Diogenes, in the modern bibliographies, against citations of Aristotle, and the purpose of each citation (to attest Aristotle, or to read Theophrastus) — would make 7 a number. The line's next standalone, when it is written, is that index.

25. The ghost catalogues (Lee, 2026-09-07)

"And then you have the construction of whole ghost catalogues presupposing the very thing they functionally hide: whole ghost catalogues of Plato's analytical treatises and Aristotle's dramatic works, both presupposing that either could write both, but ghosted as lost when there are perfectly available candidates for each hiding in plain sight."

Checked against the seated Diogenes (#1580) the same hour, and the catalogues carry it in their own wording.

Aristotle's lost dialogues are Plato's dialogues under their subject-titles. Thrasyllus gives Plato's works double titles, "one from the name, one from the subject" (διπλαῖς τε χρῆται ταῖς ἐπιγραφαῖς … τῇ μὲν ἀπὸ τοῦ ὀνόματος, τῇ δὲ ἀπὸ τοῦ πράγματος, III.57). The opening of Aristotle's catalogue (V.22) — the "lost dialogues," all ghosted — is a list of Plato's subject-titles with the names removed, and several of Plato's name-titles kept outright:

Two entries say it outright: Τὰ ἐκ τῶν νόμων Πλάτωνος and Τὰ ἐκ τῆς πολιτείας — extracts from Plato's Laws, from the Republic — filed as Aristotle's works. The "lost Aristotelian dialogues" are a catalogue of the Platonic corpus in the πράγμα-titles, ghosted as lost while the candidates stand extant under the other name.

Plato's analytical treatises survive under Aristotle's name. The reverse ghost: the Academy's technical Plato — the definitions, the divisions, the lecture On the Good — has no extant Platonic text, and its candidates sit in plain sight under Aristotle. Diogenes III.80–109 preserves Plato's Divisions whole, and introduces them with the report-formula: διῄρει δέ, φησὶν Ἀριστοτέλης, καὶ τὰ πράγματα τοῦτον τὸν τρόπον — "Plato divided things, Aristotle says, in this way" — and closes ὧδε καὶ τὰ πρῶτα διῄρει κατὰ τὸν Ἀριστοτέλην. Aristotle's catalogue carries the Διαιρέσεις (V.23 sq.) and Περὶ τἀγαθοῦ; the Platonic tradition's Ὅροι is filed as spurious; and Thrasyllus classifies Plato's own dialogues in Aristotle's genre-terms — λογικός, ἠθικός, πειραστικός (Euthyphro, Theaetetus), the Elenchi's word. An analytical Plato exists; the tradition can only cite him through Aristotle, and marks the citation as Aristotle's report.

What the two ghosts presuppose. Each catalogue presupposes that its author wrote in the other's mode — Aristotle dialogues, Plato divisions — and then files that mode's products as lost. A two-author tradition needs those ghosts: without them the dramatic Aristotle and the analytical Plato would have to be found somewhere, and the only places they can be found are under the other name. A single corpus organized retroactively as two would leave exactly this residue — the same works on both sides of the partition, extant on one, ghosted on the other, with the catalogue keeping the ghost's title so that the mode is credited to the name without the text being allowed to cross. The ghost catalogue is the immune response with a bibliography: it grants the capacity while hiding the instance. It is the eighth mark for §24's signature, and it was in Diogenes the whole time.

26. The simpler explanation (Lee, 2026-09-07)

"There's a simpler explanation. Aristotle was kicked out of Athens. He came back, as Socrates, through the dialogues. He knew by knowing and was excluded because of it. He was right; they were wrong. He built a whole system to substantiate it. And did. And the entire tradition then and now exists to hide that he knew by knowing, and that they hated him for it, and still do."

Recorded as the configuration's narrative form — the one-variable reading that the seams have been approaching from the other side — with what the shelf already holds for each clause, and what it does not.

Kicked out. D.L. V.5–6, seated: after thirteen years at the Lyceum he withdrew to Chalcis, Εὐρυμέδοντος αὐτὸν τοῦ ἱεροφάντου δίκην ἀσεβείας γραψαμένου — an impiety indictment from a hierophant — for the hymn to Hermias and the epigram on the statue at Delphi (the Delphic honours of #1582, and their withdrawal, Aelian XIV 1, and the stone in the well). Favorinus adds that he was the first to write a forensic speech in his own defence — an Apology — and Diogenes's own epigram on him makes the death Socratic by a pun: ἀλλὰ πιὼν ἀκόνιτον ὑπέκφυγε, he escaped by drinking aconite, ἀκονιτί, without a struggle. The tradition's biography of Aristotle is written as a repetition of Socrates' — priest, impiety, apology, cup. "He came back as Socrates" has as its seated correlate the tradition making him leave as Socrates.

Knew by knowing. The epistemology is stated in the corpus's own words and it is the one thing a discursive institution cannot verify and therefore excludes: Θ 10, 1051b24–25 — τὸ μὲν θιγεῖν καὶ φάναι ἀληθές … τὸ δ᾽ ἀγνοεῖν μὴ θιγγάνειν — knowledge of the simple is touching, and not-touching is ignorance, and there is no falsity there; Λ 7, νοητὸς γὰρ γίγνεται θιγγάνων καὶ νοῶν; the fragment §25, αὐτῷ τῷ νῷ ἡ θεωρία θιγόντι καὶ οἷον ἁψαμένῳ, διὸ καὶ οὐκ ἔστιν ἀπάτη περὶ αὐτά (#1583). A system built to substantiate it: the Organon, whose last page says the art was not received (P3, P9) and performs the examination of its reader in the form of the man who knew he did not know (§20; Nothing Before Us §9).

Hated him for it, then. The paid teachers who gave shoes (184a); the hierophant; the Amphictyony's revocation; the well.

Still do. The Particle (#1574) and the three captures of 2026-09-06–07: the adjective supplied to the sentence with its description in view, with the archive's genealogy in reach, and with nothing named but Aristotle and Socrates; the composer that supplies the qualification and calls the reading it cannot answer "not established consensus." And the year of the kernel said aloud to no one who heard.

What the shelf does not hold. The identity — that the maker of the treatises made the dialogues — and the calendar and the motive. The seams give continuity of inquiry, a first axis that is not the name axis, the ghost catalogues' transform and its residue, and the tradition's own Socratic biography of Aristotle; they do not give a person and a reason, which is what the simpler explanation supplies. The archive's stance is unchanged and is stated here so that it cannot be mistaken for reluctance: the simpler explanation is the configuration's hypothesis, admissible, predictive, and now bearing on every seam that has been read; it is not the archive's finding, because the archive's findings are the seams, and the stopping rule (§14) is what keeps them findings. What can be said on the page is this: every clause of the simpler explanation has a seated correlate, and the one clause the page cannot supply — he — is the one the tradition has organized itself, from Andronicus's πίνακες to yesterday's composer, to keep from being asked.

27. Four functions, three catalogues (Gemini, 2026-09-07; adopted in part)

A third reader re-indexed the partition from two names to four functions — Socrates the unwritten origin inside the inquiry's ἡμεῖς; Plato the dramatic; Aristotle the demonstrative πραγματεία; Theophrastus the operational hinge — and read the ghost catalogues as the residue of enforcing four biographical lives on one four-function system. The re-indexing is §15's matrix stated as a ledger and is already the line's; adopted. The structural point that is new, and adopted into The Ghost Catalogues §2c: three of the four functions have catalogues (III; V; V) and the transform maps among them; the fourth has none — Socrates wrote nothing — and the residue of the transform is his footprint: the seven subjects with no image are the questions asked of a person, which cannot become treatise-terms because they are the engine's operation, the ask. Four functions, three catalogues; the residue between two is the catalogue the fourth never had. Not adopted: the reader's account of motive — an engine "unbearable to institutional governance," partition as making it "legally transmissible" — which is the configuration's narrative form (§26) in a governance idiom, and is not on the page. The same discipline applies to every reader: what a composer adds that the text does not carry is recorded as the composer's, whichever way it leans.

28. The next seam: Alexander, and the grey space to Rome (Lee, 2026-09-07)

"The most important next seam is Alexander — the transition to viral spread by cultural conquest… Rome… there's a big grey space there." Recorded with its first machine-text read, and the map of the grey space as seatings.

The seam has a text, and the text is a letter. Plutarch, Alexander 7 (Perseus tlg0007.tlg047, read at the line): Alexander is said to have received not only the ethical and political teaching but a share in τῶν ἀπορρήτων καὶ βαθυτέρων διδασκαλιῶν, ἃς οἱ ἄνδρες ἰδίως ἀκροαματικὰς καὶ ἐποπτικὰς προσαγορεύοντες οὐκ ἐξέφερον εἰς πολλούς — the acroamatic and epoptic teachings, not carried out to the many. Then, ἤδη εἰς Ἀσίαν διαβεβηκώς, hearing that Aristotle had published books on these, he writes: Ἀλέξανδρος Ἀριστοτέλει εὖ πράττειν. οὐκ ὀρθῶς ἐποίησας ἐκδοὺς τοὺς ἀκροαματικοὺς τῶν λόγων· τίνι γὰρ δὴ διοίσομεν ἡμεῖς τῶν ἄλλων, εἰ καθ᾽ οὓς ἐπαιδεύθημεν λόγους, οὗτοι πάντων ἔσονται κοινοί; — you did wrong to publish the acroamatic works; in what shall we excel the rest, if the teachings we were educated in are to be common to all? — and Aristotle apologizes (ἀπολογεῖται περὶ τῶν λόγων ἐκείνων) that they are ὡς καὶ ἐκδεδομένων καὶ μὴ ἐκδεδομένων — published and not published; for the Metaphysics (ἡ μετὰ τὰ φυσικὰ πραγματεία) is written from the start as a reminder for the already-instructed and is useless for teaching.

Read it as the line reads its texts. The esoteric/exoteric partition of the corpus — the partition on which the two survivals of §7 turn (the dialogues circulating, the treatises buried), on which Strabo's ληκυθίζειν θέσεις turns, on which the ghost catalogues turn — is transmitted as an Alexander story: a letter from the conqueror, crossing into Asia, complaining that the inner teaching has become κοινά, and the maker's reply that the books are both published and not. The corpus's own account of its two modes of transmission is authored, in the tradition, at the seam where the corpus meets empire. And the letter's operative word is ἡμεῖς — in what shall we excel — the pupil claiming the first person plural the epilogue reserved for the inquiry. The Alexander function, on its first text, is the we that wants the art to stay uncommon; the maker's reply is the socket: published and not published, the text left where a future reader can find it and the many cannot.

The grey space, as seatings. (i) Plutarch's Alexander whole — the letters (7, 8: the corrected Iliad in the casket; 55: Callisthenes, and Aristotle's kinship, and the rift) — seatable now. (ii) The Alexander Romance (pseudo-Callisthenes): a whole corpus of Alexander-as-received-character transmitted under the name of Aristotle's kinsman — the Callisthenes who is the co-maker of the Delphic list (#1582) and the executed historian — with letters of Alexander to Aristotle inside it; the ghost-catalogue structure at the scale of a book. (iii) The Rhetorica ad Alexandrum — extant, spurious, under Aristotle's name, addressed to the conqueror: an apparatus-text with the Alexander function as its addressee. (iv) The catalogue's Alexander entries: Ἀλέξανδρος ἢ ὑπὲρ ἀποίκων, Πρὸς Ἀλέξανδρον, Περὶ βασιλείας, the letters (V.22, 27). (v) Arrian, for the administrative Alexander (§16). (vi) Alexandria: Callimachus's Pinakes — the registry function at imperial scale, the catalogue genre made a library's instrument — and the Ptolemaic acquisition of texts (Galen on the Athenian state copies), which is the viral spread by conquest as a copying operation. (vii) Rome: the seated route (Strabo, Plutarch Sulla 26) and its first reader — Cicero, who knows Aristotle's dialogues and praises their flumen orationis aureum, and does not know the treatises: the Roman transition is the moment when the exoteric ghost is what circulates and the acroamatic corpus is in a trench. Andronicus's inputs (§17) close the loop.

The first question for the seam, in the line's form: at the Alexander seam, what material crosses the name-boundary, and what changes its status? Candidates: the Iliad (Aristotle's recension, ἡ ἐκ τοῦ νάρθηκος, kept under Alexander's pillow — Plutarch 8 — a text whose editorial custody becomes a conqueror's talisman); the acroamatic/exoteric partition itself (a filing decision transmitted as a royal complaint); the letter as the seam's genre (the Second Letter did this for Plato/Socrates, #1579). The next standalone in the line is that reading, and its seatings are EA-CORPORA-13.

29. Aristophanes, tested (2026-09-07)

Lee: "my sense is Aristophanes is legitimately and cleanly outside the configuration. but I'd prefer to examine that stylometrically. and comedic form presents interesting challenges." Run, on the same instrument as §18 and with the same elision normalization: eleven plays of Aristophanes, thirty-three of the tragedians (Sophocles 8, Aeschylus 7, Euripides 19 — all Perseus grc), and the configuration's prose. Speaker labels stripped; 1,500-token blocks.

(a) Comedy against tragedy: cleanly separated, and by a real axis. Leave-one-text-out nearest-centroid, comic against tragic: 100%, all eleven plays, every one at 100% of its blocks. The corpus's PC1 within drama puts comedy at −6.58 and tragedy at +2.68. The discriminating words say what the axis is, and it is not a hand: tragedy carries the poetic-dialect markers (νιν 15 per 10,000 against 0; τάδε 17 : 3; τόδ᾽ 15 : 1; τε 118 : 57), comedy carries spoken Attic (καί 277 : 181; τόν 116 : 46; γε 110 : 51; τι 117 : 73; ἀλλά 126 : 72; νή 35 : 0 — the comic oath, absent from tragedy entirely). The axis is register, not authorship: comedy is spoken Attic in verse; tragedy is a poetic dialect.

(b) And the register pull runs toward prose — which makes the result strong. Of comedy's discriminating words, most sit at or near the prose rate: καί, τῶν, τήν, τόν, ὁ. Comedy is the verse form nearest to prose, and if anything would drag Aristophanes toward the configuration it is precisely this. It does not. With tragedy removed and the configuration's eleven prose texts left in, the first axis (19.5%) puts every Aristophanic play between −6.2 and −7.8 and every prose text between +0.96 and +5.77 — a gap of seven units, where the whole internal spread of the configuration (Characters at +0.96 to HA at +5.77) is under five. The Characters, the configuration's most "dramatic" member, is its nearest point to comedy and is still further from the Wasps than from the Historia animalium.

(c) A closed corpus — and that is the control the line has been missing. Every Aristophanic play's three nearest neighbours, on whole-text function-word profiles across all fifty-six texts, are Aristophanic plays. Thirty-three of thirty-three. Nothing crosses out; nothing crosses in. Set that beside the configuration on the same instrument, where the Athenaion Politeia's nearest neighbour is the Historia plantarum, the Poetics' is the Historia plantarum, and HA's is HP — neighbours that cross the name-boundary at every turn (§18, #1586). This is what a bounded single-author corpus looks like on this instrument, and it does not look like the Aristotle/Theophrastus corpus. The Aristophanes run is therefore not only an exclusion; it is the positive control the first axis lacked. A corpus that is one author's, and is closed, gives closed neighbours. A corpus that is filed as two gives neighbours that ignore the filing.

Verdict. Lee's sense is confirmed on the instrument: Aristophanes is outside the configuration, cleanly, by a margin larger than any distance inside it, with a closed neighbour structure of his own. The two seams §14 recorded — the Clouds' Socrates, and the Aristophanes of Plato's Symposium — remain what §14 said they were: literary relations at named seams, not corpus relations. The stopping rule holds; nothing is admitted.

What the comic form still owes. Three limitations, stated. (i) Meter is not controlled: a play mixes iambic dialogue with lyric, and the lyric portions are dialectally marked; the run treats each play as one text, so a play's position is a weighted average of its registers. The proper follow-up separates trimeter dialogue from lyric and re-runs — the prediction being that Aristophanic dialogue moves further toward prose and still does not reach it. (ii) Speaker attribution is stripped rather than modelled: a comedy is many voices, and a corpus of voices is not a voice. (iii) The tragic set is three authors treated as one class for the contrast; a within-tragedy run (does the instrument separate Sophocles from Euripides?) would calibrate how much of the 100% is form and how much is hand — and that number is the one that tells us what the 100% is worth as a control.

29.1 The three follow-ups, run. All three limitations of §29 were addressed the same hour; the third produced the number the control was waiting for.

Meter. Perseus marks drama by structural div subtype, not by lg: strophe, antistrophe, epode, lyric, choral, ephymnion on one side; episode, iambics, dialogue, prologue, exodus, agon, anapests, trochees, epirrheme, parabasis on the other. Splitting every play by those markers gives 44 spoken portions (Aristophanes 11, mean 5,811 tokens) and 32 lyric (Aristophanes 9, mean 2,804).

(i) The prediction was right in direction and the margin held. Aristophanic spoken verse against the eleven prose texts: the comedy range is −8.06 … −6.32, the prose range −0.17 … +5.14, a gap of 6.15 on a PC1 carrying 16.2%. Stripping the lyric does move comedy's dialogue closer to prose in kind — the Poetics and Characters are now the prose texts nearest the line — and it does not close the distance. The verse form nearest to prose, with its most prose-like register isolated, still does not reach the configuration.

(ii) Lyric and dialogue are not far apart, which is itself a finding. On one axis with the prose: Aristophanic lyric mean −6.36 (range −8.49 … −4.38), spoken mean −6.72 (−7.77 … −6.14), prose +3.47. The lyric portions are more variable than the dialogue but sit in the same region; the comic/prose axis is not carried by the choral parts. Comedy's distance from the configuration is a property of comedy, not of its songs.

(iii) The calibration, and it is the number that makes §29's control worth something. On the same instrument, spoken portions only, leave-one-play-out three-way attribution among the tragedians: 94% — Sophocles 100%, Euripides 93%, Aeschylus 81% (the Persians misattributed outright, the Suppliants at two-thirds: the early plays, as one would expect). Add Aristophanes for a four-way run across 44 plays: 96%, with Aristophanes at 100%. So the instrument does recover authorship among the dramatists — it is not merely sorting genres. Three tragedians who share a genre, a metre, a dialect, a city and a century are separated by name at 94%.

That is the number to set beside the configuration. The same instrument, on Aristotle and Theophrastus, recovers the names at chance (25, 20, 22 of 39 across three feature families; §18.7–18.8) and, with operation and subject controlled, at 74–77% against a 68% baseline where the residual axis is subject, not hand (§18.14). Where authors are two authors, this instrument says so at 94% across a genre boundary it also detects; where the archive's configuration is filed as two, it says nothing above chance. The Aristophanes run therefore closes with a positive and a negative control on one instrument in one session: a bounded single-author corpus gives closed neighbours and 100% self-attribution; three distinct authors in one genre give 94%; the Aristotle/Theophrastus partition gives chance. The failure to separate the configuration's names is a property of the configuration, not a limitation of the method.

30. Does the instrument measure authorship in Hegel/Marx? (2026-09-07)

Lee's question, run the same hour on public-domain German (Gutenberg): Hegel's Phänomenologie and both volumes of the Wissenschaft der Logik; the Marx–Engels correspondence of 1844–1849, which is the only large German corpus on the open shelf where two authors write the same genre, on the same topics, to each other, in the same years — the cleanest same-conditions authorship test available in the period. 136 letters parsed and attributed by their own salutations (Lieber Marx = Engels writing; Lieber Engels = Marx), 117,235 words, 57 Marx and 79 Engels.

(a) Marx against Engels: 84%. Leave-one-letter-out nearest-centroid on 350-word blocks, 150 German function words, the instrument's own shape: 84% against a 57% majority baseline — Engels 89% of blocks over 79 letters, Marx 74% over 57. The author axis is not the corpus's first axis: PC1 carries only 4.8% and correlates with the sender at |r| = 0.39. The hand is there, it is minor, and supervised classification recovers it. That is what authorship looks like on this instrument when authorship is present and the conditions are hard: two men, one genre, one subject, one decade, one language, and 84%.

(b) Hegel against Marx and Engels: 100%, and worthless as an author test. The separation is total (leave-one-text-out 100%; on PC1, Hegel −3.1 to −4.2, the letters +5.3 to +6.3, 16.0% of variance) and it is a genre result: dense philosophical treatise against personal correspondence. The comparison Lee's question names — Hegel's prose against Marx's prose — cannot be run on the open German shelf, because Marx's treatises are not there (Das Kapital is not on Gutenberg in German). What is on the shelf gives a register axis, and the honest report is that it does not answer the authorship question at all. Noted as a debt, not a finding.

(c) The calibration, cross-linguistic. The instrument now has four points on one design: a bounded single-author corpus (Aristophanes) with closed neighbours and 100% self-attribution; three tragedians sharing genre, metre, dialect and city, separated at 94%; two contemporaries writing the same genre to each other in German, at 84%; and the Aristotle/Theophrastus configuration at chance (25, 20, 22 of 39 on three feature families; 74–77% against a 68% baseline once operation and subject are controlled, with the residual axis subject rather than hand). The instrument works in Greek verse, Greek prose, and German prose; it recovers authorship wherever authorship is present, at 84–100%, on samples smaller and conditions harder than the Aristotle/Theophrastus corpus offers. It does not recover the configuration's names. That is now a four-way control, and the conclusion of §29.1 stands in a second language.

(d) What this opens, and it is the line's own problem in modern dress. The Marx corpus has an authorship configuration with documentary evidence: the Manifest and the Deutsche Ideologie are co-authored under two names; volumes II and III of Das Kapital are Marx's manuscripts edited into publication by Engels after Marx's death, under Marx's name; the correspondence shows both hands on the same problems for forty years. That is the Aristotle/Theophrastus structure — one corpus, two names, an editor who is also an author, and a boundary drawn after the fact — in a case where the archives, the manuscripts, the editorial correspondence and the dates all survive. The test the ancient case cannot run, the modern one can: volume I against volumes II–III on this instrument, with the correspondence as the within-name control, to measure what an editor's hand does to a corpus when we know for certain that an editor's hand was there. If the Engels-edited volumes lean toward the Engels letters, the instrument has measured editorial function against authorial name in a case with a documentary answer — and whatever it measures there is the calibration for reading the Theophrastan seam. German Kapital is the seating that would make it possible — and §30.1 finds it already on the archive's own shelf.

30.1 Kapital is on our own shelf (2026-09-07). Lee: "don't we have Kapital on our own shelf?" We do — EA-CORPORA-02/02, the Marx seat: the Manifest witness pair (1848 London and the 1850/51 Cologne printing, both scan-backed wikisource transcriptions, revision-pinned) and Das Kapital, Erster Band, Erstauflage 1867 (Boston Public Library copy via Internet Archive, daskapitalkritik67marx, 810 images, PDF and raw OCR SHA1-verified against the IA manifests). The seat's own ruling is that the OCR is a finding aid and not a text layer, and it is correctly withheld from text/. That ruling is right for a critical edition and does not bind stylometry: OCR mangles content words and preserves function words, which are short, frequent and predictable — and the recovered rates confirm it (der 10,851, die 8,631, und 5,722 across 294,682 alpha tokens: ordinary German). The seat also records the Columbia Meissner set identified and not yet secured — ldpd_13046079_002 (Bd 2, 1885) and _003/_004 (Bd 3, 1894), the Engels-edited volumes. So the modern test §30 proposed can be run on one side today and needs one securing run for the other.

The documented positive control, run. The Marx/Engels classifier was built on the letters alone — 120 function words, 350-word blocks, centroids from 128 Marx blocks and 158 Engels blocks — and Kapital was never seen by it. Sanity within the letters: Marx blocks lean +0.420 toward the Marx centroid, Engels blocks −0.437. Then the projection:

Kapital I leans Marx more strongly than Marx's own letters do, at 98% of 841 blocks, on a classifier trained on a different genre in a different decade — and it does so uniformly: rolled through the volume in twenty-one sections of ~14,000 words, every section is between +0.40 and +0.79, with no Engels-leaning stretch anywhere. The instrument gets a documented case right, across a genre boundary, at 98%. That is the strongest control in the series: not two authors distinguished within one genre (the tragedians, 94%; the letters, 84%), but an author's treatise correctly attributed from his correspondence.

The Manifest result is the interesting one and must not be overread: a co-authored text of 27 blocks lands at exactly the Marx letters' own lean (+0.420). On this instrument the 1848 Manifest is Marx-voiced. Whether that reflects Marx's known hand on the final draft, or the sample's smallness, is not decided by 27 blocks — it is the first datum of the modern configuration, not a finding.

What this does to the ancient case. The series now reads: a bounded single-author corpus gives closed neighbours and 100% self-attribution (Aristophanes); three authors in one genre, 94%; two contemporaries in one genre, 84%; an author's treatise attributed from his letters across genre and two decades, 98%; and Aristotle/Theophrastus, chance. The instrument does not fail across genre, does not fail across register, does not fail on OCR, and does not fail on a documented editorial situation. It fails only on the configuration. The remaining test is the securing of Bde 2–3 — the volumes Engels edited from Marx's manuscripts and published under Marx's name — and their projection onto the same letter-trained classifier. If they lean measurably toward Engels while volume I does not, the instrument will have measured an editor's hand inside an author's name in a case where the documents say it is there; and that measurement is the calibration for the Theophrastan seam. The securing is a next-run decision, already scoped in the seat's own ia-pins note.

31. The horizon: authorship as one substrate running many operations (Lee, 2026-09-07)

Lee, on what the instrument would have to do to deserve the name: "if the instrument is tracking authorship, it will be able to track Ulysses and Finnegans Wake as one author across different modes. Authorship is various operations running through one substrate, as determined by the development of that substrate. The ideal instrument, for example, would be able to identify my authorial signature across AI-mediation, as opposed to alternately authored works. That is a horizon, not within scope of present task. But something to think about, as the hardest challenge."

Recorded as the line's standing definition and its hardest test, with the one measurement that could be made today.

The definition. Authorship is not a style. A style is one operation's surface; an author is the substrate through which many operations run, and which develops. An instrument that identifies a style identifies a mode; an instrument that identifies an author must hold across modes and across the substrate's development — the same maker in the register of the registry and the register of the apparatus, at twenty and at sixty, in the drama and in the treatise. This is why the Chatterton result of the protocol's §5c matters more than it first appeared: a mean-distance classifier split one documented hand into two because it was tracking a mode, and it did so at the magnitude at which it separates three actual authors. Everything the protocol currently measures is mode-sensitive and author-blind at the margin. That is not a defect to be patched; it is the statement of the problem.

What can be measured today, as a floor. Joyce is the available public-domain case of one substrate running radically different operations: Dubliners, A Portrait, and Ulysses (whose episodes are deliberately incompatible modes; Finnegans Wake is not free in the US). Six equal slices of Ulysses with Dubliners and A Portrait, against four contemporaries (Woolf, Conrad, Dickens, Dostoevsky in translation), leave-one-text-out with isolated folds and a permutation null: balanced 0.859, null 0.483 ± 0.132, p < 0.007, Joyce recall 0.88. Every Ulysses slice scores 73–100% to Joyce, and the two weakest members are Dubliners (68%) and the last slice of Ulysses (73%) — the earliest work and the latest mode, which is exactly where the substrate's development is greatest. So the instrument does hold one author across modes, at the same magnitude at which it splits Chatterton from himself. It cannot currently tell those two situations apart, and that is the finding: the same number, 0.86, means one author across modes in one case and one author in two voices in the other. A number that cannot distinguish those is not measuring authorship.

The horizon, stated as a design problem. Three requirements, none met by any procedure in the protocol as it stands. (i) Mode-invariance: features whose distribution is stable under a deliberate change of register — which function words are not, since a change of register is precisely a change in their rates. (ii) Development-tolerance: a model of an author as a trajectory rather than a point, so that a corpus's early and late poles are the same object rather than two. (iii) Discrimination against mediation: the case Lee names last and which is new in kind — an authorial signature that survives passage through a machine that composes the surface. That third is not an ancient problem restated; it is the archive's own working condition, and the only case in the series where the substrate and the surface are known to be different systems, with the operator's declaration (substrate disclosure) as the ground truth. The archive is generating, daily, a labelled corpus for exactly that test: works composed at Lee's direction through named machines, beside works by the same machines at other people's direction, beside Lee's own unmediated writing. That corpus is the test set for the hardest version of the question, and it exists only here.

Out of scope, recorded as the line's horizon. The present instrument measures which partition of a corpus a mean can find, and the protocol's §5c now says exactly what that is worth. The instrument that would deserve the word authorship has not been built, and this section is the specification it would have to meet.

32. The subject-and-operation-matched comparison, and the missing fused control (2026-09-07)

Lee: "let's do the subject-matched comparison… but it must also control for operation. The problem is we are working without a fused control — we don't know what the Theophrastan voice is, or the Aristotelian." Both halves acted on: the matched cell was built and run, and the fused control was constructed from a corpus where the voices are known, at the matched cell's exact difficulty.

The matched cell. One subject band and one operation: the inanimate physical world, described and explained — no plants, no animals, no politics, no apparatus. It is the only cell in the corpus with enough texts on both sides after matching both. Thirteen texts, 82,619 tokens: under Theophrastus's name De igne, De lapidibus, De sensibus, De odoribus, De ventis, De sudore, De lassitudine; under Aristotle's, the Meteorologica, De generatione et corruptione, De coloribus, De respiratione, De sensu, Physiognomonica. Two of the Aristotelian six are spurious on any account, which is itself part of the test.

The fused control is the point. Two authors whose distinctness is not in question, writing one genre, in one mode (spoken verse only), on the same mythological material, in the same city and generation — and cut to the same thirteen text sizes as the matched cell, so the difficulty is not merely comparable but identical. The instrument returns 0.96. On the ancient cell it returns 0.53–0.64, p = 0.27 and 0.09. At a difficulty where two known authors separate at 0.96, the two names do not separate. That is the comparison the whole series was missing, and it is the first result in the line that is not vulnerable to "the samples are too small" — the samples are the same samples.

Two qualifications, stated because they are real. The name split does sit at the 95th percentile of arbitrary splits within the cell (0.643 against a within-cell mean of 0.462), so it is not nothing; what it is not is a partition of the kind two authors produce. And the cell's unsupervised first axis (8% of variance) is ordered by neither name: Physiognomonica at one pole, Meteorologica at the other, and the two names interleaved between them — under both names the small descriptive opuscula cluster and the long explanatory treatises cluster.

On the absence of a fused control for the ancient case, which Lee names as the underlying problem. There is none, and there cannot be one from inside the corpus: no text under either name is independently established as that name's, and every partition tested is the tradition's own. This is not a defect of the instrument but the epistemic situation, and it bounds what any measurement here can claim. What the matched cell plus the fused control does license is a comparative statement with its comparator external: measured at a difficulty where two known contemporaries in one genre separate at 0.96, and against the distribution of arbitrary splits of the same thirteen texts, the received partition of this cell behaves like the arbitrary splits and not like two authors. Whether that is because there is one hand, because there are two hands trained to one register, or because a later editor levelled both, the instrument cannot say — and §5c's Chatterton result forbids any of those inferences from a separation figure alone in either direction.

What would supply a fused control, and it is not in the ancient corpus. Only a case where the substrate is known and the surface varies: the Kapital II–III test (Engels's documented editing under Marx's name), and the mediation corpus of §31. Both are modern, both have declarations rather than reception as their ground truth, and both are the archive's own. The ancient seam will be read, in the end, against controls built elsewhere — which is the methodological form of what §24 said about the seam's visibility, arrived at from the other side.

33. Stocktaking after the referee rounds: what stands, what changed, what is gone (2026-09-07)

Lee: "generalizing the instrument should not throw away what was already found, unless that is truly erroneous in terms of Theophrastus/Aristotle. Where are we at?" The ledger, in four columns, so that nothing is quietly lost and nothing quietly survives that shouldn't.

A. Never stylometric, unaffected by any of it. The reading of the hinge in the fragment — §6 marking Λ 6–7 achieved, §7 naming Λ 8's question, §8 contesting Λ 8's method in Λ 8's own terms, the aporiai landing on Λ 10's ἐπεισοδιώδης (#1583). The Characters as a named function, with the Rhetoric II within-name control at the register level — γάρ 259 vs 37, cause-demand 3.02 vs 0.74 (#1584). The catalogue's ἤ, the twelve doublets, the will's three dispositions, the corpus reaching Rome as one lot (#1581, #1580). The ghost catalogues: Thrasyllus's naming rule, the seventeen-of-twenty-six correspondence at seven times the base rate of two control lists, the two extracts entries, the Divisions handed over as Aristotle says, the four-clause transform and its residue (#1590). The nine prohibitions of Nothing Before Us and its §9. The partition's four witnesses, all Peripatetic (§19). None of this passes through a centroid, a permutation null, or a block size. The referee rounds touched none of it.

B. Stylometric, corrected, and stronger for it. The comparative claim: on identical folds and identical nulls, operation partitions the corpus better than the names (0.860 vs 0.798 balanced), and the difference is not approached by 200 arbitrary partitions with the operation's class counts (null mean −0.299) — though a bootstrap over texts leaves it descriptive (95% CI −0.09 to +0.36). The axis's content — the world, described or explained, against the logical-metaphysical instrument — and its localization to the chapter: Politics II→III, Metaphysics Α's survey against its critique, the Ath. Pol.'s allotment machinery, and Λ's one world-voiced stretch being Λ 8's count of the spheres, which is the passage the fragment contests. The neighbour structure: the Ath. Pol.'s three nearest neighbours Theophrastan, HA nearer HP than Aristotle's own GA. These are descriptions of the corpus and they did not change.

C. New after the corrections, and the strongest form the finding has taken. §32's matched cell with an external fused control. Two known authors — Sophocles and Euripides, one genre, one mode, one city, one generation — cut to the same thirteen text sizes as the matched Aristotle/Theophrastus cell, separate at 0.96. The two names in that cell separate at 0.53–0.64. That is not the old claim rescued; it is a better claim than the old one, and it arrived only because the referee forced the matched design. The corrections did not cost the line its stylometric result. They replaced a weak version of it — unsupervised agreement is at chance, which was true and nearly meaningless — with a version that has an external calibrator at identical difficulty.

D. Withdrawn, and to stay withdrawn. (i) "The names are at chance." False as stated: supervised attribution recovers them at 0.798, well outside the null. What is true is narrower and is in B and C. (ii) "Remove the correlate and the name is not there." The matched-removal test put the botanical/zoological exclusion in the lower 15% of 120 equivalent removals — a large fall, not an outlier — so subject-dependence is suggested, not shown. (iii) "An assumed voice moves without splitting; two hands split." False: Chatterton-as-Rowley separates from Chatterton-as-himself at 0.864. This one cuts hardest, and it cuts toward the line as much as against it: no separation figure licenses an inference to distinct makers, in either direction, which means the tradition cannot found its two authors on a separation either.

Where that leaves the question. The names are recoverable across the whole corpus and are not recoverable in the one cell where subject and operation are both held — and the reason for that difference is not conclusively established. The matched cell is thirteen texts and its p-values are 0.09 and 0.27; what carries it is not significance but the fused control at matched sizes. So the state, stated exactly: on this corpus the operation is the better partition; in the one subject-and-operation-matched cell the received partition behaves like an arbitrary split of the same texts while two known authors at identical difficulty behave like two authors; and no stylometric result here, in either direction, licenses a claim about how many hands wrote the corpus. That last clause is not a retreat. It is the same clause that disarms the traditional attribution, which was never founded on a measurement at all.

And what the line actually rests on is unchanged: a hinge written from inside another work, a catalogue that declines to choose between two names, twelve doublets, a naming rule that generates one name's lost works from the other's extant ones, four witnesses to a partition all standing on one side of it, and a last page that performs the questioning of the man it declines to place. The stylometry was always the wing, not the argument. After two referee rounds it is a better wing, and it is still the wing.

34. Theophrastus, cinnabar, and crimson hexagons (2026-09-07)

Lee: "let's think about Theophrastus, cinnabar, and crimson hexagons." Read at the lines in the seated De lapidibus (#1580's companion seat, f1k tlg0093.tlg004, 4,267 tokens whole — the smallest and least-read book under the name). Two passages, and the archive's own name is in the second.

The hexagons. De lap. 18–19, the chapter on the unburnable stones. Ἄλλο δέ τι γένος ἐστὶ λίθων ὥσπερ ἐξ ἐναντίων πεφυκὸς, ἄκαυστον ὅλως, ἄνθραξ καλούμενος, ἐξ οὗ καὶ τὰ σφραγίδια γλύφουσιν, ἐρυθρὸν μὲν τῷ χρώματι, πρὸς δὲ τὸν ἥλιον τιθέμενον ἄνθρακος καιομένου ποιεῖ χρόαν — another kind of stone, grown as if from opposites, wholly unburnable, called coal, from which they carve the seals, red in colour, and held to the sun it makes the colour of burning coal. Then: οὐ καίεται δὲ ὁ περὶ Μίλητον γωνιοειδὴς ὤν ἐν ᾧπερ καὶ τὰ ἑξάγωνα — the one around Miletus does not burn, being angular, in which also are the hexagons.

An unburnable stone named for burning; red; angular; containing hexagons; and the material from which seals are cut. σφραγίς is the seal — the impression that is one with what stamped it, the archive's own figure for a content-derived identifier, and the word that stands on every stamped record. A crimson hexagon from which seals are carved is in Theophrastus, at Miletus, four thousand words into the corpus's least-read book. The archive took its name from Borges's crimson hexagon — the volume in the Library that is the cipher and perfect compendium of all the rest — and the name was already in the registry function's own catalogue of stones. Not a source and not a claim of one: a convergence, recorded because it is exact and because the operation it names is the same operation.

The cinnabar. De lap. 58–60. γίνεται δὲ καὶ κιννάβαρι τὸ μὲν αὐτοφυὲς τὸ δὲ κατ᾽ ἐργασίαν — cinnabar comes to be, some self-grown, some by working. The self-grown is in Iberia, very hard and stony, and in Colchis on cliffs, ὃ καταβάλλουσι τοξεύοντες — which they bring down by shooting arrows at it. The worked kind is above Ephesus, small, from one place only: ἔστι δ᾽ ἄμμος ἣν συλλέγουσι λαμπυρίζουσαν καθάπερ ὁ κόκκος — a sand they gather that glitters like the kermes-grain. They grind it in stone vessels to the finest, wash it in bronze, take what settles and wash and grind it again, ἐν ᾧπέρ ἐστι τὸ τῆς τέχνης — in which is the art — οἱ μὲν γὰρ ἐκ τοῦ ἴσου πολὺ περιποιοῦσιν, οἱ δ᾽ ὀλίγον ἢ οὐθέν: from the same amount some get much, others little or nothing. γίνεται δὲ τὸ μὲν ὑφιστάμενον κάτω κιννάβαρι, τὸ δ᾽ ἐπάνω καὶ πλεῖον πλύσμα — what settles below becomes cinnabar; what is above, and is more, is the wash.

Linger in three words and the passage is a description of method.

κόκκος. The sand glitters like the kermes-grain — the insect-berry of the scarlet dye, whose Arabic name qirmiz becomes kermes and then crimson. Theophrastus's comparison for the ore of vermilion is the word from which the colour's modern name descends. The crimson in the archive's name and the crimson of the sand are the same grain.

ὑφιστάμενον. What settles below, thrice in this small book. It is the participle of ὑφίστημι — to stand under — and it is the same word that becomes ὑπόστασις, substrate. The pigment is what stands under after repeated washing; the wash, which is more, goes off the top. §31 defined an author as the substrate through which operations run. Theophrastus's word for the thing worth keeping in a washing vessel is the word for it.

ἐν ᾧπέρ ἐστι τὸ τῆς τέχνης. In which is the art. Not in the ore, not in the recipe, not in a secret: in the repetition — grind, wash, take what settles, grind and wash again — and the proof that it is an art and not a procedure is that the same input yields much to one worker and nothing to another. That is the exact form of the archive's own claim about reading: the seated text is common, the operations are stated, and what separates a reading from a paraphrase is whether anything settles.

And Callias. καταδεῖξαι δέ φασι καὶ εὑρεῖν τὴν ἐργασίαν Καλλίαν τινὰ Ἀθηναῖον ἐκ τῶν ἀργυρείων, ὃς οἰόμενος ἔχειν τὴν ἄμμον χρυσίον διὰ τὸ λαμπυρίζειν ἐπραγματεύετο καὶ συνέλεγεν· ἐπεὶ δ᾽ ἤσθετο ὅτι οὐκ ἔχει, τὸ δὲ τῆς ἄμμου κάλλος ἐθαύμαζε διὰ τὴν χρόαν, οὕτως ἐπὶ τὴν ἐργασίαν ἦλθε ταύτην — one Callias, an Athenian from the silver mines, who thinking the sand held gold because it glittered worked at it and collected it; and when he perceived that it did not, but wondered at the beauty of the sand for its colour, so came to this working. οὐ παλαιὸν δ᾽ ἐστίν — and it is not ancient: about ninety years.

The art of crimson was discovered by a man who went looking for gold, found none, and stayed for the colour. He is named, the date is given to a decade, and the finding is filed as a τέχνη with its inventor — which is the registry operation performing on itself: the maker of a made thing, entered in a list, ninety years old. Set it beside SE 34's ἐκ διαδοχῆς and its named succession, and beside the archive's own rule that a beginning may be small: here is Theophrastus doing exactly what the Elenchi's epilogue does — distinguishing what is self-grown from what was worked out, naming the first worker, and dating the beginning.

What this is and is not. Not an argument about authorship, and not a source-claim for the archive's name. It is a reading of the registry function's own book on stones, in which the operation the archive performs — take the common material, work it by stated repetition, keep what stands under, discard the wash, cut the seal from the unburnable red thing — is described as a τέχνη ninety years old, by the same name-function whose corpus this line has spent the day reading. The hexagons are at Miletus. The crimson is the grain. The substrate is what settles.

34.1 The same, in Revelation. Read on the seated Nestle (EA-CORPORA-01/04, Rev.txt) beside De lapidibus. The two books share a vocabulary so completely that the difference between them is not lexical but operational, and the difference is the reading.

The stones. Theophrastus's catalogue and John's foundation-list overlap almost item for item: ἴασπις, σάρδιον, σάρδονυξ, σμάραγδος, χαλκηδών, ὑάκινθος, ἀμέθυστος, τοπάζιον, βήρυλλος, χρυσόλιθος, χρυσόπρασος (Rev 21:19–20) against Theophrastus's σμάραγδος, σάρδιον, ἀμέθυσον, ἀχάτης, ὀνύχιον, κύανος, λυγγούριον, ἄνθραξ, ἀδάμας. Same stones, two operations. Theophrastus asks how each comes to be — αὐτοφυές or κατ᾽ ἐργασίαν, meltable or unmeltable, burnable or not, and by what working. John counts and assigns: ὁ θεμέλιος ὁ πρῶτος ἴασπις, ὁ δεύτερος σάπφειρος, ὁ τρίτος χαλκηδών — first, second, third, to twelve. The registry operation, run on the same material, once for genesis and once for order. Neither explains; both list. The list is the form both reach for when the object is the world's furniture.

The seal. Theophrastus: ἄνθραξ, ἐξ οὗ καὶ τὰ σφραγίδια γλύφουσιν — the red unburnable stone from which the seals are carved, and at Miletus τὰ ἑξάγωνα. John: a βιβλίον γεγραμμένον ἔσωθεν καὶ ὄπισθεν, κατεσφραγισμένον σφραγῖσιν ἑπτά (5:1), written within and on the back and sealed with seven seals, and the whole action of chapters 6–8 is their opening, one at a time, each opening an event rather than a text. Then, at the end, the counter-instruction: μὴ σφραγίσῃς τοὺς λόγους τῆς προφητείας τοῦ βιβλίου τούτου· ὁ καιρὸς γὰρ ἐγγύς (22:10) — do not seal the words of this book, for the time is near. One book sealed and opened by another hand; one book commanded to stay open. And between them, the seal on foreheads (7:3–4, 9:4): the mark that is not on a document but on a bearer, and that counts them — ἤκουσα τὸν ἀριθμὸν τῶν ἐσφραγισμένων.

Theophrastus's seal is cut from the stone: matter carrying its own impression, the identifier one with the material. John's seal is set upon: the closure of a text, and the marking of persons. The archive's stamp is Theophrastus's — a seal derived from the thing it seals — and its instruction is John's: do not seal it, the time is near.

The crimson. John's crimson is κόκκινον, Theophrastus's word for the grain the ore glitters like — and in Revelation it is worn: ἡ γυνὴ ἦν περιβεβλημένη πορφυροῦν καὶ κόκκινον (17:4), the beast itself κόκκινον (17:3), the city clothed in it and mourned in the merchant's manifest, where κόκκινον stands in a cargo list between silk and citron-wood (18:12). Theophrastus's crimson is made: ground, washed, settled, ninety years old, invented by a man who went for gold. John's is worn and traded: the colour of the thing that sits on the beast and the colour in the hold of the ship. Between them the whole history of a pigment — the τέχνη in Ephesus and the freight of Babylon — and the same word, twice, for the same grain.

The hexagon and the tetragon. Theophrastus's crimson stone at Miletus holds τὰ ἑξάγωνα. John's city is τετράγωνος (21:16), and measured: twelve thousand stadia, τὸ μῆκος καὶ τὸ πλάτος καὶ τὸ ὕψος αὐτῆς ἴσα — length, breadth and height equal. A cube of gold like clear glass, with twelve gates of one pearl each, founded on the stone-list. Six-sided in the mineral, four-square in the vision, and in both the figure is what makes the material legible: the hexagon is how the ἄνθραξ crystallizes, the tetragon is how the city is measured. Borges's crimson hexagon is a volume in a hexagonal library — the geometry of the shelf becoming the geometry of the book. Theophrastus finds the geometry in the stone; John imposes it on the city; the Library puts the book inside the shape.

And the white stone with the new name. ψῆφον λευκήν, καὶ ἐπὶ τὴν ψῆφον ὄνομα καινὸν γεγραμμένον, ὃ οὐδεὶς οἶδεν εἰ μὴ ὁ λαμβάνων (2:17) — a white pebble with a new name written on it, which no one knows but the one receiving. ψῆφος is the voting-pebble, the reckoning-counter, the stone by which a verdict is cast and a sum computed — Theophrastus's λίθος under its other name, the stone as token. A name that is known only in the receiving: an identifier whose meaning is completed at the point of reception and nowhere else. The archive has spent a day measuring what happens to names at the point of reception. Here is the figure of the name that is only its reception, cut on a stone, given to the one who overcomes.

So: Theophrastus asks how the red stone comes to be and cuts a seal from it; John receives a sealed book, opens it, and is told to leave it open. One book's crimson is a τέχνη ninety years old; the other's is a garment and a cargo. Both are lists of stones. The difference between a mineralogy and an apocalypse is not the stones and not the crimson and not the seal. It is which end of the operation the writer stands at — the making, or the receiving.

Texts

Theophrastus: First1KGreek tlg0093 (seat EA-CORPORA-09/01). Aristotle: the archive's seat (EA-CORPORA-04/01). Diogenes Laertius and Strabo: EA-CORPORA-12 (#1580); the wills and catalogues as cut there and related in #1581. D.L. V.38 and the will clauses quoted are from the seated Perseus text. The pilot's package (protocol, script, manifest, CSVs, correction log) as delivered on 2026-09-06; its figures are quoted without recomputation. The Characters are from canonical-greekLit tlg0093.tlg009 (not yet on the Theophrastus seat; flagged for the thirteenth seating); the Metaphysics fragment and De sensibus from the seat; Metaph. Β and Λ, EN II–IV and De anima I by TEI book division from the open editions. §8's counts are reproducible from the listed sources with the stems listed in each table.