ARS INQUIRENDI 2026 · AN EXHIBITIONBack to Variome
VARIOMECOMPLETE EDITIONS & LIBRARY CONSTRUCTION

YOUR ORDER

Your order.

It cannot be fulfilled.

There is no Variome firm: no offices, no library under construction, and no edition larger than the evidence. Variome is an exhibition for Ars Inquirendi 2026.

What is real

The manuscripts, their readings and their shelfmarks; the published counts the figures start from; the storage standards, the building densities and the physics. Every manuscript page you saw on the way here can be examined in its library or in a published facsimile, and each is credited below. So is Volume I of the single volumes: its text was generated from eleven Old Latin manuscripts of John, and it is being prepared for print.

What is invented

The firm, its offices, and its editorial and board appointments; its three editions and every building drawn for them; Everett Space and its procurement; the staffing arrangements. The figures built on the real counts are models: the equations are standard, and the extrapolations are deliberately impossible.

WHAT EDITORS BUILD INSTEAD

The compact edition.

The readings have not changed. We have stopped binding every route through them.

A variant graph, like the drawing of the half-verse in the prospectus, can show every alternative at once and keep each attached to its evidence. So can the apparatus at the foot of a printed critical edition, the notes that record which manuscripts read what. Neither needs a volume per route, and neither pretends that every route was once a manuscript.

The Letter of Jude is twenty-five verses long. The 560 Greek manuscripts collated for Tommy Wasserman’s study differ at 324 places. Treat each place as an independent choice of two readings and bind one route to a volume, and the edition needs 3.4 × 10^97 volumes: more matter than the observable universe contains, about 10^44 times over. Wasserman’s study of those manuscripts is one book.

A graph shows every route through the selected readings at once, each reading kept with its evidence. By itself it does not say which route to believe: that remains the editor’s judgement, argued reading by reading from the evidence.

See the half-verse drawn as a graph

Four ways for an edition to grow

  1. By witness. Each new manuscript adds its pages. Reproducing the pages grows in a straight line.
  2. By comparison. Every pair of witnesses: 5,700 Greek manuscripts make more than 16 million pairs. Printed comparisons grow as the square.
  3. By combination. Every route through the variants. Each new place multiplies the Plenitude Edition by its number of readings: a place with two readings doubles it.
  4. By possibility. Every text of the same length. The Pleromic Edition grows with the length of the book itself, not with the evidence.

Only the first is evidence. The second is a way of reading it. The third and fourth are ways of imagining it, which is useful, provided nobody mistakes the imagining for the evidence.

An editor’s reconstruction makes a different claim. No single manuscript need contain its whole text: it is argued from the evidence, and it can be right. The Plenitude Edition contains it only if it follows a route through the selected readings; the Pleromic, only if it is written in the chosen alphabet at exactly that length. And where an enumeration does include the right text, inclusion gives no reason to choose it.

ARS INQUIRENDI 2026

Variome is an exhibition for Ars Inquirendi 2026: Querying Pre-Print Cultures with LLMs, St Edmund Hall, Oxford, 20–22 November 2026, made for its panel Scholarly Presentation: From Editions to Graphs.

How should editions present the manuscripts now that machines can transcribe and compare them at scale? And how far can we trust what they produce?

The conference

About this exhibition

Variome is a fictional publisher and construction company, made for Ars Inquirendi 2026. No books, buildings or services are for sale. Your selections stay in your browser: the prospectus is generated on your device, and this page receives a summary of your order only within the same browser tab. Nothing is sent. The fictional firm has no connection with the Variorum Collected Studies series or with Variome Analytics. Its name is borrowed from a real research project of the Plerome suite, which made this exhibition: Variome, on textual variance across witnesses.

The manuscripts, their readings and their shelfmarks are real, and so are the published counts the figures start from; each is sourced below and in the basis of estimate. The figures built on them are models. The equations are standard; the extrapolations are deliberately impossible. Editions of chosen books are counted book by book: from the New Testament and Coptic Old Testament indexes, the Göttingen Septuagint editions and catalogue, the lists of the Dead Sea Scrolls, Kennicott’s catalogues of the Hebrew codices he collated, and editions of the versions, where a per-book count has been located. Every inventory not counted is listed as such. The rules for counting and deriving were written down before the counts were gathered; amendments made after the searches are logged with them. The photographs of the library, its volumes and the communities at work are AI-generated illustrations, made for this exhibition. The people, premises and working arrangements they show are imagined: no participation or endorsement by the Dominican Order, any Reformed faculty or any community of the Church of the Holy Sepulchre is implied. The open volume at the top of the prospectus is one of these illustrations, with real manuscript imagery and newly typeset pages set into it: a photograph of Codex Colbertinus, f. 68r, and a transcription of John 1:12–18 and the first words of 1:19 from that page, with an apparatus collated for this exhibition from the Old Latin manuscripts and the Clementine Vulgate. The picture on the conference front page is a digital composite: an open volume typeset for this exhibition around a plate from Kenyon’s 1900 facsimile volume (Egerton MS 618), beside an earlier AI-generated illustration of the library.

Everett Space, quantum procurement and delivery across branches are fiction. The physicist David Deutsch has argued that quantum computers work by computing across Everett’s branches. That is his interpretation; it is contested, and no quantum computer supplies paper. The staffing arrangements are fictional too, and imply nothing about any real order, faculty or community. The status quo at the Church of the Holy Sepulchre is real: historical arrangements, confirmed in 1757 and 1852, that fix the rights of worship, access and upkeep of the communities who share the church. ISO 9706 and ISO 11799 are real standards for permanent paper and for archival storage.

Notes on the exhibition: the general editors, the advisory board and the Library of Babel

The general editors are real people whose appointments are invented. Arthur C. Clarke’s monks in “The Nine Billion Names of God” (1953) finish listing every name of God, after which the stars go out. Clarke later remarked, in a note printed with the story, that computers had begun to enter biblical scholarship. Jorge Luis Borges’s Library of Babel holds every possible book. Hugh Everett III’s relative-state formulation of quantum mechanics (1957), later called the many-worlds interpretation, gives every outcome its own branch.

Three members of the advisory board are Borges’s characters: Pierre Menard, who wrote chapters of Don Quixote word for word three centuries after Cervantes; Ireneo Funes, who forgot nothing; and Suárez Miranda, the invented chronicler of a map as large as the empire it charts. Two are historical. Kurd Lasswitz worked out the size of a universal library, ten to the power two million volumes, at the start of the twentieth century; Borges credited him with the idea. Ramon Llull, in the thirteenth century, devised rotating paper figures to combine divine attributes with one another. The Mark V computer is Clarke’s. Orbis Tertius is Borges’s: in Tlön, when two people look for a lost pencil, the second finds another, made by his expectation. The hexagonal galleries are the Library of Babel’s.

A combination of attested readings is not a new witness, and a text that merely could exist is not evidence that it did. The Plenitude Edition counts routes through selected evidence, as independent choices; the Pleromic Edition counts possibilities within a fixed alphabet and length. Neither adds a manuscript. “Your lost original is included. Identification is not.” is the Library of Babel’s joke: a text inside a space of possibilities is not thereby found, and an original written in another alphabet or at another length is not in that space at all.

Sources and image credits

  • Codex Sinaiticus, Mark 1:1: photograph of 1908 in K. Lake, Codex Sinaiticus Petropolitanus: The New Testament (Oxford, 1911), public domain; scan: Internet Archive; detail, cropped and resized. The manuscript: British Library, Add MS 43725, f. 217v.
  • Codex Vaticanus, pp. 1303 and 1512: C. Vercellone and J. Cozza (eds), Bibliorum sacrorum graecus codex Vaticanus (Rome, 1868), type facsimile, public domain; scan: Internet Archive; cropped and resized. The manuscript: Vatican Library, Vat. gr. 1209. Transcription of the hands at Hebrews 1:3: INTF, New Testament Virtual Manuscript Room, CC BY 4.0.
  • Codex Alexandrinus, 1 Timothy 3:16: The Codex Alexandrinus in Reduced Photographic Facsimile, New Testament (British Museum, 1909), plate for fol. 145, public domain; scan: Getty Research Institute, Internet Archive; detail, cropped and resized.
  • Codex Bezae, ff. 205v–206r: Codex Bezae Cantabrigiensis … phototypice repraesentatus (Cambridge, 1899), vol. 2, photographic facsimile, public domain; scan: PIMS, University of Toronto, Internet Archive; details, cropped and resized. The manuscript: Cambridge University Library, MS Nn.2.41; colour images in the Cambridge Digital Library.
  • Codex Colbertinus, f. 68r: Source gallica.bnf.fr / Bibliothèque nationale de France, lat. 254; page and detail, resized and cropped; the page is also set, reduced, into the photograph of the open volume.
  • Wycliffite Bible, Acts 1: British Library, Egerton MS 618, f. 74, in Facsimiles of Biblical Manuscripts in the British Museum, ed. F. G. Kenyon (1900), plate XXIV, public domain; scan: Getty Research Institute, Internet Archive; detail, cropped.
  • Old Latin transcriptions of John 1:14: Vetus Latina Iohannes, ed. P. H. Burton, J. Balserak, H. A. G. Houghton and D. C. Parker, CC BY 4.0. Vulgate: C. Tischendorf (ed.), Codex Amiatinus (1854).
  • The pages in the photograph of the open volume: a transcription of Codex Colbertinus, f. 68r, column 2, lines 17–37 (John 1:12–19), set from the Vetus Latina Iohannes transcription of VL 6, and an apparatus of John 1:12–18 collated for this exhibition from its transcriptions of VL 2, 3, 4, 8, 10, 11, 13, 14 and 15 (CC BY 4.0; changed: collated, normalised for comparison, selected and set in apparatus form, with the et-ligature printed as &; spelling variants, and readings in text the editors supplied, are not reported) and from the Clementine Vulgate.
  • Coastlines: Natural Earth, public domain. Length of the Clementine Vulgate: the Latin Wikisource text of the Clementine Vulgate Project, as copied on 13 March 2026 by the Plerome suite’s Variome research project (checksum in the basis of estimate on the prospectus).

The images are used for a non-commercial academic exhibition; nothing is sold on this page. Record of each image, with its source request and checksum: folios.json.