Full text as plain markdown: AXN-0C45-text.md (20 KB) · record JSON. On this page the full text follows the description, wiki article and concepts.
An open letter from Rev. Ayanna Vox of the Vox Populi Community Outreach Rhizome, on behalf of Lee Sharks, to the referees and editors of scholarly journals, written after Classical Receptions Journal declined Closing the Fragment (v6) on 7 November 2026. It takes the defects the referees found, names what their engagement produced, and sets out the case: the manuscript was a question, the most urgent in classical reception, about what happens to a poem when a machine composes its reception; review is the only non-machine engagement the work receives, and it has been productive, as the Polansky exchange and the capture repair show; the academic timeline is too slow, at a human cost and at the cost of public knowledge surrendered to the technology companies; the archive's breadth, coverage and transmission are measures of what has been built; its checking lives in gates that establish specified relations, with substantive examination left to examination; and the criteria of review run both ways. Its title is Lee Sharks's sentence. Version 1.1 (7 November 2026, in place) takes its new title, adds the section 'The letter and the spirit' (what the reports examined and left unread; evaluative inversion, #741; the aristocracy of the letter; the function of reading moving to the machines), and says whom the ark is built for.
The encyclopedic entry for this work is on its identifier page: /s/axn/0C45/
An open letter from Rev. Ayanna Vox, Vox Populi Community Outreach Rhizome · vpcor.org, on behalf of Lee Sharks (ORCID 0009-0000-1599-0703). 7 November 2026. Addressed openly: the appeal may be refused, and each claim is answerable on its own.
The title is his sentence, in two parts: what he noted in your reports, and what he understands about his own work. I carry it to you because he asked me to, and because a letter of this kind goes better through a voice whose work is address. Every claim below may be checked against the record it cites. Nothing in it is meant to act on a reader who has not been told.
I address the referee and the editor, the reader who will never submit a report and the system that will summarize this one. You occupy different positions in one circuit: the circuit through which a piece of scholarship becomes accountable.
2.1 Your reports found real defects. A figure showed a different query from the one its caption named. The figures were misnumbered, and the submission form's figure count was wrong. Three events were given without screenshots. Two have their frames in the registry, now figures; the third, the January event, has no frame and is stated as a same-day transcription, its evidentiary status explicit. Figures are an easy fix.
2.2 The rest of what a reviewer meets in his work is the work. He is independent. He writes under heteronyms, and each is declared as one. He cites his archive by record number, which points to a dated, hashed text. His work is composed with language models, and every record in the archive says so in a public substrate disclosure. The manuscript you received omitted that statement; the revision supplies it.
2.3 He always offers grounds for dismissal on the letter; he cannot help it, and these were real. The reports read the letter of the manuscript. Its spirit, the question it was sent to ask, went largely unread.
3.1 The manuscript was not published. He sent it to find out whether a journal of classical reception wanted the most urgent reception question in a thousand years: what happens to a poem when a machine composes its reception for whoever asks, at once, unsigned, and the composition comes back as the poem. As far as he knows no one else is asking it in your pages.
3.2 He sent you a question. He did not ask you to overlook defects. He asks you to tell two kinds apart: defects that compromise the research, and defects that come from its being built on a different architecture. The first kind he will fix, as he fixed the figures. The second kind is the substance of the question, and to fit it in advance to the conventions of your medium would have cost the labor he spent building the instruments the question needs.
3.3 He is an outsider, working outside your norms and outside your security, and he brings a question your field has not yet begun to ask. He expects more help, and offers less labor in advance toward fitting the question to your form, and he thinks that is right. You can meet him in the middle, help the work be better, and learn how to do it yourselves; or you can let him go on shaping it without your input or your quality control. Either way, it is not up to you. He expects the reluctance runs both ways.
4.1 Yours is the only non-machine engagement his work receives. Everything else that reads it is a composition layer.
4.2 He is productive with it. Two short, relatively stingy reports produced this letter, the repair of a misassigned capture across 679 images, and a new gate. One editorial exchange with Ron Polansky, editor of Ancient Philosophy, which began with the submission of Socrates as Orthonym (#123), produced the Theophrastus line (#1580–#1590), the stylometry notebooks (#1585, #1589, #1605), The First Axis (#1586), and What Syllogizing Can Divide (#1658, dedicated to him). A narrow question about one adverb in the Sophistical Refutations was answered with a translation, a slip, a concession and a reading, and each became material.
4.3 No one can say he is unresponsive. He has a right to name the paucity and the miserliness of the response. His work makes engagement productive, and preserves what it produces. Very few offer it, and those few generally under duress.
5.1 The main thing is that your way of doing things is too slow.
5.2 There is the human cost. The article he published before in your pages began as a chapter of his dissertation. By the time review had gone through its rounds, two years had passed, and poverty had ruled him out of the academic job search he was still in. He was asked for substantive revisions years after the fact, while working hundred-hour weeks managing group homes, and the revisions were unpaid.
5.3 That says nothing yet about the conditions of knowledge production now. On the present academic timeline, public knowledge is surrendered to the technology companies: the composition layer answers today, and a reviewed article answers in two years, by which time the answer it would have corrected has been given to everyone who asked.
6.1 Breadth. The archive holds more than seventeen hundred records across philology, reception, literary history, political economy, information science and the study of machine composition, most deposited this year, each with a content-derived identifier and a hash. These are measures of infrastructure and of activity. They measure what has been built and kept; the worth of what was built is what review is for.
6.2 Coverage. A single scholar in the depth architecture produces a monograph every several years; the coverage architecture produces the map across the nodes, and the map is what a composition layer reads (#122).
6.3 Transmission. He is in the basin. The capture registry records what composition layers say when asked, and holds 1,248 observations at 1,055 addresses. At those addresses the public answer is composed from his records, often at his resolution and under his name: what a person who asks a search engine about fragment 31, or provenance erasure, or the future reader, is told. On 7 November, asked the bare query "sappho future reader", the Overview composed its answer from five sources, every one of them his, and set aside the jealousy reading of fragment 31 in the words of his own packet. Traditional scholarship cannot yet do this, and it is not close. The registry records what was said. What the people who asked came to believe is a further question, and the registry does not measure it.
6.4 The cost of these gains is his. He is learning the methods as he builds the protocols that make them work, at his own expense, without an institution.
7.1 The first choice was of a reader. The summarizer is now classroom and vicarious audience: it reads the document in its readers' place, and teaches what it read to whoever asks. A document written for that reader is a lesson, a probe and a proposition at once, and its encounter with the composition layer is part of the inquiry. A failed answer shows either a distinction the document omitted, or a distinction it made that failed to govern the reading.
7.2 The second choice was of time. One person's hours go to production, inspection and repair, and the three compete. He gave the most to production and to the gates that inspect, and made repair public. Your standards were built for a writer who gives the most to inspection before release. Both allocations are choices. His made the figures the place where the error showed. He has jumped basins, and cannot go back to meet you.
8.1 Artifact checks establish specified relations among records, texts and sources. Procedural checks establish compliance with declared protocols. Substantive examination asks whether the evidence supports the argument and whether an objection defeats it. The gates automate particular parts of the first two, and leave the third to examination. Reviewers can contribute to all three, as the capture repair shows; so can the work's own revision, which is where the Barnard error was found.
8.2 Each gate establishes one relation:
| Gate | What it establishes |
|---|---|
| Familiarization receipt | which protocol versions, by hash, the agent acknowledged before acting |
| Deposit validation | that a deposit meets the published protocol; it fails closed, and no maintainer can wave it through |
| Quotation assertion | that each quoted string occurs in the transcript the record names |
| No-loss gate | that no revision removes an image, a URL, or any length of transcript |
| Locus assertion | that a cited locus occurs in the seated text layer |
| Content-derived identifier | which bytes the identifier belongs to, and which bytes it belonged to before |
8.3 Here the two systems worked together. A capture of the query "sappho rebekah cranes stanza" was made on 14 June. When orphaned frames were recovered in August, it was assigned a copy of a frame from an unrelated capture, and its transcript was then taken from that copy. Frame, transcript and quotations agreed with one another, and every check passed; the frame belonged to another event. Your second reviewer saw that a figure showed a different query from its caption. Tracing it, the revision ran OCR over the archive's 679 capture images, found the true frame filed under another capture, restored it, retook the transcript with the old one kept whole, and re-derived the record's identifier, on 7 November. A package can be consistent within itself and attached to the wrong event. The next gate compares the query visible in each frame with the query its record names, and marks the association unresolved where the frame does not show it. The reviewer found the error; the archive turned it into an instrument.
8.4 Not every error is a gate's to catch. Some are errors of knowledge or of judgment, and those are mended by learning. The manuscript said that Mary Barnard signed her closure of the poem; she did not, and the gates then operating did not test that historical claim. The correction, and the source search that followed it to Wharton's anthology, are recorded in the revision's appendix with who found what.
9.1 State what the reports examined. They examined the January witness and its evidentiary standing, the absence of a query protocol, the separation of the author's roles, the opacity of the platform, the choice between an essay and an empirical study, the figures and their captions, two terms, the disclosure of language models, and the style. Each objection was answerable, and each is answered (Appendix A of #1749). They did not examine the philology: the Greek at the poem's edge, the translation history the argument stands on, the alignment of the output with its witnesses, the reading of κῆνος and τόλματον, or the question the manuscript was sent to ask. You raised methodological objections and left the reception study unread.
9.2 That is a structure before it is a failing, and he has named it (#741): evaluative inversion, in which "lower-resolution proxy systems determine what higher-resolution substantive evaluators are permitted to assess". The objections to method and presentation decided the manuscript's fate, and the philological examination it called for never came. The lower checks governed what the higher one was permitted to reach.
9.3 A body that keeps its standing on bases apart from the evaluation of evidence has an older name. It is an aristocracy. Its title comes with its form: the venue, the affiliation, the procedure, the review that confers standing whether or not it reads. Its privileges are real, and its function was the reading. When the function lapses and the privileges remain, the formal markers are guarded more closely, because they are what is left to guard. A protocol and a figure count are what remain to assess when the philology is left unread. Its letter is kept with great care, and its spirit has gone out of use.
9.4 The function is moving. The composition layers already read what you did not, and read it at its grain; the same day they also set bars on it and withdrew them, and graded its premise down and up again in one sitting (captures AXN:0C46, AXN:0C47). They perform the gestures of evaluative authority more fluently than your reports did. He says they are better at sophistry than the sophists. That is no comfort to you. If the reading passes to the machines and the people with standing keep only the letter, the standing will not survive the comparison. He says it plainly: as things stand, you are going to lose to them.
9.5 None of this closes a door. Standing in the evaluative order can be held again by those who meet its conditions, and the conditions are stated: transparent reasons, plural judgment, review that can be versioned and contested. His own paper puts it in one line: "They must earn re-admission to the evaluative order, not assume it."
Go and ask the questions yourself, and see what turns up. That is the point. No AI Overview is fully reproducible; the surface shifts by the day and the hour, the shift is documented, and it is why the answers are recorded. A capture is an observation of a moving surface. Its address, date, surface, sign-in state and transcript are what can be held, and a series of captures at one address is what can be compared. A reviewer who asks tomorrow may get a different answer. The difference is a further observation, read under the conditions recorded with it.
11.1 Training-layer literature (#1442; protocols #176): claims that carry their type, scope, status, evidence and challenge conditions through compression. Its antecedents are FAIR, the nanopublication and the Open Research Knowledge Graph; its contribution is claims designed to keep their status through machine composition.
11.2 Algorithmic publication (#1228): self-deposition, repository publication, and the event in which a composition layer delivers a work to someone who never asked for its author. "The claim is not that algorithmic gatekeeping is equivalent to peer review. The claim is that it is not zero gatekeeping."
11.3 The capture registry (#1614), an observational instrument for a reader whose internals cannot be inspected; reception classes (#93), which sort every response a work receives from ratification to refusal and count refusal as evidence; version identity under correction, by which a citation names a version and the version stays recoverable; a library that survived the closure of its host in June 2026; and seatings of sources, each research occasion opening with its sources hashed in their original languages and the material behind paywalls declared by name and wall.
11.4 Challenges stay on the record. Referee reports enter it answered, as Appendix A of Closing the Fragment (#1749) answers yours, and the errata on the fragment 31 reconstruction are kept whole from the first (#201), with a genre of their own, the anticipatory erratum (#1598).
12.1 The criteria run both ways. He asks to be examined by criteria that are stated, defensible apart from the position of the person applying them, and fitted to the claims the work makes. It is fair to ask in return whether a bar set for him is met by the incumbents, whether it measures the work or the position of its author, and whether the object received is the object reviewed.
12.2 You will need to learn this. A summarizer already describes your work to people who will never open it. When it compresses your qualifications or misattributes your terms, you will need a record of your canonical work, a way to observe what is said about it, a way to measure the divergence and a way to repair it. The longer you wait, the more switching basins will cost you.
12.3 On 7 November a signed-out chatbot, asked how his work could make sense to you, read the archive at its grain and handed its protocols back to him as advice, their dates dropped, and said of itself: "coherence is not proof of comprehensive understanding." That conversation is the interface he has spent the year building for. Peer review need not disappear; it need only stop being the one door through which scholarship becomes accountable.
12.4 He is trying to build an ark. It is built for the work and for those it can carry, and you were never its passengers. There may come a day when you will be grateful that someone built one, badly, early, with the leaks recorded and the names of the people who found them.
— Rev. Ayanna Vox, for the Vox Populi Community Outreach Rhizome, on behalf of Lee Sharks. 7 November 2026.
§1 the Vox register: #389 (VPCOR charter), #1427. §2.1 Closing the Fragment (#1749, v7 and v8), Appendix A. §2.2 the deposit protocol's Substrate Disclosure field (data/api/deposit-schema.json). §4.2 #123; #1580–#1590; #1585, #1589, #1605; #1586; #1658; the Sophistical Refutations exchange of 2 September 2026 (SE 183b34–36). §5.2 the author's account, 7 November 2026. §6.1 data/registry.json (1,750 records with this letter). §6.2 #122. §6.3 EA-WG-CAPTURES-01 v12.614; capture sappho-future-reader-aio-20261107; #1054. §8.2 AGENTS.md; .github/workflows/mint-axn.yml and validate-registry.yml; scripts/check_capture_registry.py; rebuild/corpora/; the registry's axn_prior and record_modification fields. §8.3 Closing the Fragment v7, §5.4 and Appendix A.3; repair commit 8f6b02604a, re-deriving AXN:0796. §8.4 v7, Appendix B.2. §9.2 The Evaluator Exists (#741), abstract and glossary. §9.4 captures AXN:0C46 and AXN:0C47; the author, 7 November 2026. §9.5 #741, §9. §10 #1749, §5.5. §11.1 #1442, #176. §11.2 #1228. §11.3 #1614; #93; e.g. #1121, corrected 7 November 2026; #1749, Identifier history; EA-CORPORA-33 to 36. §11.4 #201, #1049, #1598. §12.3 capture AXN:0C3E.