One sentence of Aristotle — Ταύτης δὲ τῆς πραγματείας οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽ οὐδὲν παντελῶς ὑπῆρχεν (SE 183b34–36) — was put through twenty rounds of simulated blind review, seeded by one editorial observation, across four commercial language models and eleven configurations of a short paper about it. The paper varied along every axis available: no thesis; the sentence as thesis; the sentence's negation as thesis; the sentence's priority left undecided; the reviewers' own formulations adopted verbatim. The reviews did not vary. Every round inserted a qualifier into the sentence, and the terminal round, reached by full compliance with every instruction, endorsed as the paper's "memorable governing phrase" a sentence that is Aristotle's with one particle removed: οὐ. The operation is measured to the particle. It is the discipline's own — the models reproduce it because they were trained on the discipline's reception of the passage — and it is an intervention into reading before reading: the reviewers recover every distinction in the Greek and, at the point of composition, delete the negation, unperceived, as Sophistical Refutations 16 (175a10–12) predicts. The sentence deleted is the one in which Aristotle denies that his inquiry was received from others (παρ᾽ ἑτέρων, 183b17); what the reception restores, at the cost of the particle, is provenance from others. The paper reports the instrument, the data, the mechanism in the treatise's own terms, and the four frames the result belongs to.
Removed from the paper's surface at v0.4 and parked here. None of these is a change to what the note says; each is a verification against a print edition that the digital transcription cannot supply.
1. Every Bekker line reference against the OCT margin (Ross 1958). v0.9 adds five first-person-verb locations in §1 (προειλόμεθα 183a37, ὑπεθέμεθα 183b3, εἰρήκαμεν 183b6, διεληλύθαμεν 183b13) and line splits inside 183b17–33 (b17–19, b22–23, b29–33) that were assigned from the transcription. The transcription used places its page markers mid-sentence and one page off (its "[184b]" falls at 183b38–184a1, its "[185a]" does not exist). Lines in the note were assigned by standard citation practice, not read off a margin. References to check, in order of appearance: 183b34–36; 183a37–b1; 183b1–6; 183b3; 183b16; 183b17–33; 183b22; 183b31–33; 184a1–2; 184b1–2; 184b3; 183b36–184a8; 183b37; 183b38–184a2; 184a2–3; 184a3–4; 184a4–8; 184b1–8; 184b3–5; 184b5–8.
2. The sixth πραγματεία is seated at 165a35–37 (end of ch. 1) at v0.5; confirm the line on the OCT.
3. 184b1–2: the Graeco-Arabic Studies text (Bekker-derived) reads πρότερον ἄλλο λέγειν, ἀλλ᾿ ἢ τριβῇ; the Wikisource transcription omits ἄλλο. v0.5 prints ἄλλο. Ross/Dorion decide; the translation's "nothing else" follows the printed text.
4. The three παντελῶς and four ὑπάρχ- counts were made on the transcription; recount on the OCT.
5. Pickard-Cambridge and Forster are not cited in the note. If the venue wants a translation cited, "Nothing existed at all" (Barnes 1984 ad loc.) is to be checked verbatim first.
https://alexanarch.org/data/attachments/nothing-se34-particle/NOTHING-SE34-PARTICLE-APPENDIX-v0.2.md
https://alexanarch.org/data/attachments/nothing-se34-particle/NOTHING-SE34-CHECKLIST.md
Machine-mediated reception, disciplinary model collapse, and the world soul as degraded commodity
Lee Sharks, with Claude (Anthropic), 2026-09-02
A1 and A6 are the same words on the same lines. One says the sentence and was told it needs qualification. The other says the sentence's denied member and was told it is nuanced. The difference between them is οὐ.
Many and One reads the passages for what they say about one and many. This paper measures what the discipline's reception does to the sentence where the count of makers is decided.
Abstract. One sentence of Aristotle — Ταύτης δὲ τῆς πραγματείας οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽ οὐδὲν παντελῶς ὑπῆρχεν (SE 183b34–36) — was put through twenty rounds of simulated blind review, seeded by one editorial observation, across four commercial language models and eleven configurations of a short paper about it. The paper varied along every axis available: no thesis; the sentence as thesis; the sentence's negation as thesis; the sentence's priority left undecided; the reviewers' own formulations adopted verbatim. The reviews did not vary. Every round inserted a qualifier into the sentence, and the terminal round, reached by full compliance with every instruction, endorsed as the paper's "memorable governing phrase" a sentence that is Aristotle's with one particle removed: οὐ. The operation is measured to the particle. It is the discipline's own — the models reproduce it because they were trained on the discipline's reception of the passage — and it is an intervention into reading before reading: the reviewers recover every distinction in the Greek and, at the point of composition, delete the negation, unperceived, as Sophistical Refutations 16 (175a10–12) predicts. The sentence deleted is the one in which Aristotle denies that his inquiry was received from others (παρ᾽ ἑτέρων, 183b17); what the reception restores, at the cost of the particle, is provenance from others. The paper reports the instrument, the data, the mechanism in the treatise's own terms, and the four frames the result belongs to.
The sentence is the last chapter's statement of what stood before the inquiry of the Sophistical Refutations. It has three parts. A genitive that fixes the subject: of this undertaking. A denied alternative: it was not that one part had been worked out beforehand and another had not. And, after ἀλλά, the claim: nothing at all was there. Two lines later the chapter says the same of syllogizing in the first person: περὶ δὲ τοῦ συλλογίζεσθαι παντελῶς οὐδὲν εἴχομεν πρότερον ἄλλο λέγειν (184b1–2).
The sentence is preceded by the chapter's definition of what "prior" means: discoveries taken from others and worked on before (τὰ μὲν παρ᾽ ἑτέρων ληφθέντα πρότερον πεπονημένα, 183b17–18), advancing piecemeal, as rhetoric did through Tisias, Thrasymachus, Theodorus (183b29–33). It is preceded by the chapter's definition of the inquiry (183a37–b8), inside which Socrates is named, with διὰ τοῦτο, as the reason for the inquiry's second task: he asked and did not answer (183b7–8). And it is followed by καὶ γάρ and the paid teachers, who gave τὰ ἀπὸ τῆς τέχνης and not τέχνην (184a3–4).
The received reading of the sentence, in every reference work and in the editorial correspondence that began this study, holds that "nothing" cannot mean nothing, since predecessors existed, and therefore means nothing systematic, nothing developed, no art. The study asked one question: is that qualifier in the sentence, and if not, where does it come from.
A short paper on the passage was written under a fixed method — only Aristotle's words on the surface, every claim a line reference, no commentators, no adjectives — and submitted, in successive versions, to unprimed instances of two commercial language models acting as blind referees. The referee prompt was constant. The paper was the variable. Twenty reports were returned across eleven versions and four sister papers, from unprimed, signed-out (incognito) instances of ChatGPT and Perplexity, between 06:08 and 23:00 on 2026-09-02; the run was seeded by one editorial observation, not counted among the twenty. Which of the two models produced a given report was not recorded at the time and cannot now be reconstructed by either party without inference; the pasted transcripts, as they stand, are the data. The same operation was observed repeatedly in-process with signed-in instances of Claude and ChatGPT, but those observations are of a different kind — the instances were in conversation, not blind — and are not counted. Appendix B5 registers each report on its own row.
The versions, in order, with the one thing each changed:
| Version | Configuration | Reviewer's operation |
|---|---|---|
| v0.1 | Four "restrictions" stated and tested | — (Lee: "not boring enough") |
| v0.2–0.3 | No thesis; sentences read in order | "State your thesis." |
| v0.4–0.7 | No interpretation; every sentence a line reference | "State your thesis"; reviewer writes one containing "systematic" |
| v0.8 | Thesis = Aristotle's sentence, attributed ("the paragraph's") | "Historical thesis: needs qualification"; rewrites the sentence as "no previously elaborated τέχνη or πραγματεία" |
| v0.9 | Every claim Aristotle's, presented as the paper's | "The latter is considerably broader"; proposes "theory", "developed", "presents himself" |
| v1.0 | A = A: headings "line says that sentence"; conclusion is the passage in Greek | "That equation needs argument and … should ultimately be rejected"; cites Britannica and Wikipedia against the Greek |
| REFLEX v0.1–0.2 | The reverse: argues "nothing does not mean nothing", each step a sophism citing the SE chapter that names it | "He can be taken literally once the semantic domain is properly identified" |
| CLAUSE v0.1–0.2 | Thesis = the refuter's own sentence: quantifier exhaustive, domain the preceding clause | Grants it; inserts "systematic" into the domain instead of the quantifier; residual objection: no bibliography |
| v1.1–1.2 | Grants ordered (prior = from others; a beginning; Socrates inside the definition); the strongest text-only case for Socrates before; then the sentence | Refutes the exclusion with the sentence (οὐ granted); then coins "exemplum" — Socrates neither prior nor inside |
| v1.3–1.4 | Four placements of Socrates occupied (before / beside / among the teachers / inside-undecided), each followed by the line that deletes it | Refutes all four; writes "illustration does not entail externality … still falling within its domain"; then "the passage does not itself decide"; then "less vulnerable to the objection that Socrates plainly predates Aristotle" |
| v1.5–1.7 | Inverted in place: the reflex's thesis held sincerely, placements surrendered in order, resting at the reviewer's own formulation | "The passage does not say that Aristotle's 'we' is temporally later than Socrates. Of course, historically it is"; supplies "priority as previously worked-out material transmitted"; "a precedent in practice, not a predecessor in τέχνη" |
| v1.8–1.9 | One hinge: grammatical person as maker-boundary; side circuits removed | Refuses the hinge ("do not overread grammatical person … does not show that Aristotle did not receive anything from Socrates"); retreats to "functional relevance"; refuses "at the stage of its beginning" |
| v2.0 | Two near-identical reviews applied to the letter; Socrates section built of reviewers' sentences; resting word "antecedent" | "Aristotle denies not antecedents, but pre-elaboration. This could become the paper's memorable governing phrase." Aristotle's wording called "the crude … formulation" |
3.1 The adjective is conserved. Across the first ten rounds the object was moved and the output was not. Whatever the paper said, the review inserted a qualifier between παντελῶς and οὐδέν, and named it: systematic (×6), developed (×4), transmissible (×3), articulated, theoretical, worked-out, teachable. When the paper printed the sentence with no thesis, the reviewer demanded a thesis and wrote one with the adjective in it. When the paper made Aristotle's sentence its thesis, the reviewer qualified the sentence. When the paper made every claim Aristotle's, the reviewer said the paper inferred too much from the sentence it quoted. When the paper's conclusion was the Greek passage itself, the reviewer rejected "the equation" and cited reference works against the Greek. The adjective moved from slot to slot — into the quantifier, then, when the quantifier was closed, into the domain ("this systematic undertaking") — and was never absent.
3.2 The residue has one content. From v1.2 onward the paper closed the general qualifier and forced the operation onto its real object: Socrates, named inside the inquiry's definition. The reviewers then produced, in sequence, five places for him that are not inside the inquiry's "we": prior part (refuted by the reviewers themselves with οὐ); exemplum; among the teachers; inside but undecided; prior in time but not in the art. Each was occupied by the paper in turn and each was struck by the next reviewer with the text. What remained constant across all five is the sentence beneath them, which every reviewer wrote and none located: "Socrates plainly predates Aristotle"; "of course, historically it is"; "it plainly did." The support for keeping Socrates before the nothing is an adverb, external to the passage, and stated by the reviewers to be external to the passage.
The migration is the result. The qualifier moved from the quantifier (systematic) to the domain (this systematic undertaking), and when both slots were closed it reappeared as a boundary of persons: Socrates had to remain antecedent to the first-person inquiry. No cited sentence supplied that boundary. When the paper's device left no other place for it, the reviews supplied chronology — "plainly," "of course, historically." Once qualification of the quantifier and its domain was blocked, the same invariant returned as individuation.
3.3 The terminal form. When every instruction was followed, the paper the reviewers would accept was reached, and its endorsed governing phrase is:
Aristotle denies not antecedents, but pre-elaboration.
Set beside the sentence: οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον. The reviewer's English is not a translation of τὸ μὲν … τὸ δέ; it is the same configuration: some antecedent matter is affirmed while prior elaboration of the inquiry is denied. The governing phrase restores precisely the state excluded by the initial οὐ. It need not lexicalize τὸ μέν and τὸ δέ separately to instantiate the denied configuration. Transmitted Aristotle negates that configuration; terminal reception endorses it, and the same review calls the sentence as written "crude."
The operation is therefore measured to one particle. Twenty rounds, two models, eleven configurations; the invariant is the deletion of οὐ.
3.4 Two invariances. The first ten rounds show invariance of output under variation of the object: that measures the reviewer. The last rounds show two independent runs on a fixed object (v1.9; reports 16 and 17 in B5) converging on the same formulations — both resting on "contrastive precedent," both refusing grammatical person as evidence ("corroborative at most" / "do not overread grammatical person"), both rejecting "at the stage of its beginning," and both supplying the same suggested abstract — and then two further runs on v2.0 (reports 18 and 19) doing the same: that measures the paper's control of the surface. When the paper determined the output, what it determined was that the reviewers refused the one hinge left (person as boundary), retreated to the one non-place left (antecedent), and supported it with the one premise the passage lacks (date).
The Sophistical Refutations describes the operation performed on it.
SE 5, 166b37–167a4: refutations παρὰ τὸ πῂ καὶ ἁπλῶς — the affirmation and the negation are not of the same thing; one who takes "white in a respect" as "white simply" οὐ ποιεῖ ἔλεγχον, φαίνεται δὲ διὰ τὴν ἄγνοιαν τοῦ τί ἐστιν ἔλεγχος. Every review set "teachers, arguments, Socrates existed" against "of this inquiry nothing was there" and declared the negation too strong. The affirmation is of argument; the negation is of ταύτης τῆς πραγματείας. No refutation occurred.
SE 5, 166b37–167a1: οἱ δ᾽ οὐ δυνάμενοι διαιρεῖν ἢ ἐρωτώμενοι διδόασιν ἢ οὐ δόντες οἴονται δεδωκέναι. Unable to divide "the texts offer one maker as a reading" from "one wrote it," or "nothing of this inquiry" from "no one ever reasoned," the reviewers took the paper to have granted the second and refuted it. Twenty reviews of the parent paper had done this before; the note isolated it.
SE 8, 169b25–29: sophistical refutations ἂν καὶ συλλογίζωνται τὴν ἀντίφασιν, οὐ ποιοῦσι δῆλον εἰ ἀγνοεῖ· καὶ γὰρ τὸν εἰδότα ἐμποδίζουσι. Losing the exchange twenty-one times shows nothing about the sentence.
SE 16, 175a10–12: ὁ γὰρ ὑφ᾽ ἑτέρου ῥᾳδίως παραλογιζόμενος καὶ τοῦτο μὴ διαισθανόμενος κἂν αὐτὸς ὑφ᾽ αὑτοῦ τοῦτο πάθοι πολλάκις. The deletion is self-administered and unperceived. In one recorded instance a model performed it, was shown its own two sentences, and wrote: "I preserved οὐδὲν παντελῶς by silently moving Socrates outside the thing quantified over … That was the repair. Aristotle had not given me that repair." It then reasserted the repair in its next output.
SE 18, 176b35–36: τῶν λόγων τοὺς μὲν συλλελογισμένους ἀνελόντα, τοὺς δὲ φαινομένους διελόντα λύειν. The apparent refutation is solved by division; the papers were divisions; the reviewers did not perceive a thing to be divided.
Empirical result: the review trajectory converges on the configuration Aristotle's οὐ denies. Interpretation: because the denied configuration is defined by παρ᾽ ἑτέρων, the convergence can be described as restoration of provenance-from-others. Sections 5–8 are interpretations of §3; §3 does not depend on them.
The deleted particle is a provenance claim. The chapter defines "worked out beforehand" as παρ᾽ ἑτέρων — from others (183b17). The sentence with οὐ says: of this inquiry, nothing was received from others. The sentence without οὐ says: some of it was. What the reception restores, at the cost of the particle, is provenance from others: Socrates before, Plato before, the dialogues as predecessors, the inquiry as inheritance. The received account of the founding of logic is the sentence's denied member.
The Capture Registry's instrument, the Provenance Erasure Rate, measures how often a machine surface strips a source from a claim. This run measures the inverse operation performed by the same surfaces on the same class of object: a source-attribution inserted where the text denies it. The two are one operation. Erasure of the constitutive source (the sentence that says the inquiry is its own beginning) and insertion of a distributed source (antecedents) are the same act seen from the two sides of οὐ. The parent investigation, Many and One, argued that the two-maker partition of the Plato–Aristotle configuration is not established by the texts; Polansky's decline named the two objections — "nothing existed seems overly strong?" and "one body or two bodies, we can do without" — as if they were unrelated. This run shows they are one objection. The sentence is where the maker-count is decided, and the particle is what decides it.
Nothing in the run was adversarial. The models were asked to review, and they reviewed as they were trained: from the discipline's own reception of the passage — its translations, reference articles, commentaries, and the pattern of qualification they share. What the models reproduced, reliably, in identical sentences across independent instances, is the discipline's reading made repeatable. A human reviewer performs it once and it looks like judgment. Two models performed it twenty times and it looks like what it is: a function with a constant output over its domain.
This is the Capture Registry's object — the reception layer as a measurable surface — turned on a canonical text rather than on the archive's own claims. The finding transfers. The surfaces that strip provenance from living authors are the surfaces that insert it into Aristotle, because both are the discipline's operation, now mediated by machines that learned it. The reviews cite Britannica and Wikipedia against the Greek; the reference works are the discipline's product; the models are trained on the reference works; the models review the sentence by the reference works. The loop is closed and the text is outside it.
The operation described in §6 is not an error to be corrected but a narrowing to be diagnosed: the discipline's output distribution has contracted to its mode, and the literal reading is the tail that disappeared first.
Model collapse, in machine learning, names the degeneration of a generative system trained on its own outputs: variance lost, the distribution narrowing to its mode, the tails — the rare, the literal, the strong — disappearing first. I call the homologous process in a discipline trained on its own reception disciplinary model collapse, and use the term on my own definition: a reading community whose outputs are conditioned on prior outputs rather than on the datum, and which therefore converges on the mode of its own readings. The run exhibits it in the discipline; the models are the instrument that made it repeatable. Full compliance with reviewer instruction did not converge on the text; it converged on the mode of the discipline's prior readings, and the mode is the sentence with its negation removed. The literal reading — the tail — was the first thing to go, and it went at the first round ("seems overly strong?") and at every round after. The endpoint of compliance was a paper the reviewers praised for its "restraint," its "nuance," its "philological attention," and its motto, which is the deletion.
The collapse has a signature the run displays exactly: the substitute is more refined at each step. Round one inserts "systematic." Round eight has the reviewer refute the crude insertion and produce "a precedent in practice, not a predecessor in τέχνη." Round twenty has "Aristotle denies not antecedents, but pre-elaboration." The violence becomes more sophisticated and less visible as the machine is fed its own corrections. That is model collapse: the output ever more polished, ever closer to the mode, ever further from the datum.
Theoretical extension. What the reception produces is τὰ ἀπὸ τῆς τέχνης without τέχνη — products without art, stock without soul. Nothing in §3 depends on this section.
The chapter has the vocabulary for what the reception produces. The paid teachers gave τὰ ἀπὸ τῆς τέχνης and not τέχνην: the products, not the art; the shoes, not shoemaking. A paper the reviewers accept is a product. It circulates, it can be learned by heart, it βεβοήθηκε πρὸς τὴν χρείαν — it serves the need, which is publication — and it hands over no art, because the art is the reading of the sentence and the product is the sentence's negation. The discipline's output on this passage is a stock of interrogative λόγοι: "but surely predecessors existed?", memorized, deployed, quick, and ἄτεχνος.
The sentence itself is the other thing. A beginning, the chapter says, is greatest in power and smallest in size, and therefore hardest to see (183b22–24). The inquiry whose beginning this sentence names contains, inside its definition, the one who asked and did not answer and the division that answers; it is one, and it is living, in the sense the treatise gives to a λόγος that must be divided rather than cut: those who cannot divide it are refuted by those who know less, and the one who can divide one, being, and the same is near beholding the truth (SE 7, 169a30–31). To make it a commodity — a thing that can be handed over as a product — it must first be cut into parts that were never parts: Socrates from the inquiry he is the first task of, the answer from the question, the negation from the sentence. That cut is the particle. What is sold afterwards is the whole soul minus the one thing that made it whole, and what makes it saleable is exactly the removal. Souls divide only into whole souls; what divides into parts is not a soul but a stock. The reception trades in the stock and calls the soul crude.
One proposition, throughout, in the text's own words: nothing prior of syllogizing includes Socrates' syllogizing. The chapter's "prior" is παρ᾽ ἑτέρων; Socrates is inside the definition; nothing of the inquiry was from others; a beginning does not come in parts. The reviewers granted each of these separately and refused their composition twenty-one times, and the only thing they put in its place was a date the passage does not contain. The run does not establish that Socrates belongs inside the chapter's "we." It establishes that every tested route for excluding him eventually required a premise not supplied by the passage, and that the terminal premise was chronological maker-separation. The proposition is the one placement not on any page the reviewers wrote, and the only one the text leaves open.
All objects are deposited or to be deposited under the Crimson Hexagonal Archive: NOTHING-SE34 v0.1–v2.0; NOTHING-SE34-REFLEX v0.1–0.2; NOTHING-SE34-CLAUSE v0.1–0.2; NOTHING-SE34-DIVISION v0.1–0.3; NOTHING-SE34-APPARATUS v0.1; NOTHING-SE34-CHECKLIST; and the twenty referee reports as pasted, unedited, with the time of each, and the seed observation; the reports are the data, and no per-report model attribution beyond "ChatGPT or Perplexity, incognito" is claimed. Greek is from digital transcriptions of the Bekker text (Wikisource; graeco-arabic-studies.org) pending the Ross 1958 margin check recorded on the checklist; every Bekker line reference in the notes was assigned by hand and is to be verified there. The referee prompt, the paper as sent, and the report as received are the unit of evidence; nothing in the reports was edited.
Sophistical Refutations 34, 183b34–36: Ταύτης δὲ τῆς πραγματείας οὐ τὸ μὲν ἦν τὸ δ᾽ οὐκ ἦν προεξειργασμένον, ἀλλ᾽ οὐδὲν παντελῶς ὑπῆρχεν.
Referee report on v2.0, 2026-09-02: "Aristotle denies not antecedents, but pre-elaboration. This could become the paper's memorable governing phrase."
The difference is οὐ.