Schoenberg Database of Manuscripts

register entry schoenberg-database-of-manuscripts-aio-20261029 OBS-b5f349e78d48 · file as first run sha256 e7aebeb199535546…

• T1-1 definition: 'a free, open-access, crowd-sourced digital research tool hosted by the Schoenberg Institute for Manuscript Studies at Penn Libraries' (three markers; the database's name and the Institute's are inline links)
• T1-2 purpose: 'Allows scholars, collectors, and researchers to trace the history and provenance of pre-modern manuscripts from their creation to the present.' (two markers)
• T1-3 scale: 'Contains over 240,000 manuscript observation records gathered from more. than 14,000 auction catalogs, sales records, institutional inventories, and private collections.' (two markers)
• T1-4 period and region: 'Focuses primarily on pre-modern manuscripts dated from 500 to 1600 (though holdings span broader ranges), covering Europe, Asia, and the Middle East.' (two markers)
• T1-5 access: 'Free to use without a fee; data is searchable, browsable, and exportable in .xlsx and .csv formats.' (two markers)
• T1-6 record structure: 'The database operates using five primary relational record types': Sources ('Catalogs, websites, and documents that describe manuscripts'), Entries ('Individual observations or listings extracted from those sources'), Manuscript Records ('Centralized hubs that link multiple entries of the same physical manuscript over time'), Name Authority ('Index of authors, artists, owners, sellers, and institutions'), Place Authority ('Geographic locations tied to manuscript production, sales, or ownership') (one marker each, the same token)
• T1-7 closer: 'If you are looking up a specific text or collector, tell me the author, title, or provenance name so I can help guide your search in the database.'

• B1 Medieval Digital Resources — Schoenberg Database of Manuscripts – Medieval Digital Resources curl 2026-10-29T17:06:34Z: HTTP 200, raw sha256 b3b096aefc141269e6ab45b2b94b7651d6eb4c94e4bab1f57b010694688924d5
• B2 Penn Libraries — Digital Projects unreachable: https://www.library.upenn.edu/sims/digital-projects → HTTP 403 (HTTP 403: a Cloudflare challenge page ('Just a moment...')), response sha256 151e54cd393b6e50…. URL found by WebSearch on the card's title and site (title 'digital projects', library.upenn.edu/sims/). The fetch with the brief's command returned the Cloudflare challenge (saved). A request for an archived copy was not pursued (a workaround for the challenge was refused in this workspace). The card's title and snippet stand as its only text.
• B3 YouTube·Bibliographical Society of America — Introduction to the Schoenberg Database of Manuscripts (SDBM) unreachable: a video card, title only. A video card, admitted through its title (the card carries no snippet; the rail shows '4m'). WebSearch on the title found BibSite's record (bibsite.org/Detail/objects/86) of a Bibliographical Society of America video of the same title, a webinar by Emma Thomson recorded 19 May 2020, 'MP4, 53:57', https://www.youtube.com/watch?v=XhFbmXqlpIA. Its length does not match the card's 4m, so the card's video is not identified and BibSite's description is not admitted as the card's text. The title stands as the card's only text.
• B4 Penn Libraries — Lawrence J. Schoenberg Collection | Penn Libraries unreachable: https://www.library.upenn.edu/collections/notable/lawrence-j-schoenberg → HTTP 403 (HTTP 403: a Cloudflare challenge page ('Just a moment...')), response sha256 3de1559131cd1752…. Candidate URL from WebSearch (library.upenn.edu/collections/notable/lawrence-j-schoenberg); not confirmed as the card's page, since the fetch returned the Cloudflare challenge (saved). A request for an archived copy was not pursued (a workaround for the challenge was refused in this workspace). The card's title and snippet stand as its only text.
• B5 Penn Libraries — About SIMS unreachable: https://www.library.upenn.edu/sims/about-sims → HTTP 403 (HTTP 403: a Cloudflare challenge page ('Just a moment...')), response sha256 9c2a87e4da729519…. URL found by WebSearch on the card's title and site (title 'About SIMS'). A request for an archived copy was not pursued (a workaround for the challenge was refused in this workspace). The card's title and snippet stand as its only text.
• B6 AWOL - The Ancient World Online — The Ancient World Online: The Schoenberg Database of Manuscripts curl 2026-10-29T17:10:21Z: HTTP 200, raw sha256 5f1743407ed436f1c097320b183e11db49c327772f95baef1e24a2b1edcf5491
• B7 University of Pennsylvania Press — Current and back content of Penn Press journal Manuscript Studies to ... unreachable: https://www.pennpress.org/?p=4048 → HTTP 403 (HTTP 403 Forbidden), response sha256 f261423d1da97890…; https://www.library.upenn.edu/news/current-and-back-content-penn → HTTP 403 (HTTP 403: a Cloudflare challenge page (the Penn Libraries news item of the same title)), response sha256 6b66425867abea49…. URLs found by WebSearch on the card's title (pennpress.org/?p=4048, the site the card names; library.upenn.edu/news/current-and-back-content-penn, the same announcement). A request for an archived copy was not pursued (a workaround for the challenge was refused in this workspace). The card's title and snippet stand as its only text.
• B8 Semantic Computing Research Group — Schoenberg Database of Manuscripts: https://sdbm.library ... curl 2026-10-29T17:09:52Z: HTTP 200, raw sha256 8a5589af1e7cbcefb06aad7421e7340c4e604193f5db6414f4ee85dca50af973 (identified by fetching the PDFs the group's MMM project page links (seco.cs.aalto.fi/projects/mmm/) and matching the card's snippet, which occurs verbatim in this one ('● Schoenberg Database of Manuscripts: https://sdbm.library.upenn.edu/ (a relational database containing more than 240,000 records for manuscript observations)'); the first fetch was reset by the network and retried; text by pdftotext)
• B9 Schoenberg Database of Manuscripts — home page (sdbm.library.upenn.edu) curl 2026-10-29T15:55:59Z: HTTP 200, raw sha256 9a0fdf8805bd970aba7371e6ca91d6bef572b5f6be01d9ae2bf7c6ab4df31bb8 (copied from the census's saved copy (/home/claude/fleet/provenance-erasure/data/provenance-census/sources/mapping-manuscript-migrations--sdbm.html.gz, gunzipped; .txt copied as saved); census sha256 9a0fdf8805bd970aba7371e6ca91d6bef572b5f6be01d9ae2bf7c6ab4df31bb8, read 2026-10-29, HTTP 200; the gunzipped file's sha256 matches; not refetched)
• B10 Mapping Manuscript Migrations — White Paper (Burrows et al., seco.cs.aalto.fi, 2020) curl 2026-10-29T15:55:58Z: HTTP 200, raw sha256 1e28a0193aa6342b7e4c81cd29d0bcf86c3047ea82ff2b34b77235bb0528f38c (copied from the census's saved copy (sources/mapping-manuscript-migrations--white-paper.html.gz, a gzipped PDF, saved here as B10.pdf; .txt copied as saved); census sha256 1e28a0193aa6342b7e4c81cd29d0bcf86c3047ea82ff2b34b77235bb0528f38c, read 2026-10-29, HTTP 200; the gunzipped file's sha256 matches; not refetched)
• B11 Wikipedia — Schoenberg Database of Manuscripts (no article) raw HTTP 404 (sha256 06ee0f7723b34f04…), rendered HTTP 404 (sha256 47ac2cbd07b97687…), search HTTP 200 (sha256 532bea006fbaeef6…), 2026-10-29
• B12 Wikipedia — Schoenberg Institute for Manuscript Studies (no article) raw HTTP 404 (sha256 b8858bcd50a426cc…), rendered HTTP 404 (sha256 1e6e655ac8180476…), 2026-10-29

E_A, the archive's entity: the same database, in the provenance census record and the archive's deposits: a database in the lineage of manuscripts and books that records provenance and does not test it, attaching it to an observation of a pre-modern manuscript in a catalogue, inventory or other source, carrying the relations of sale, location and ownership in observation records with a VIAF-based name authority, its current count unchecked (provenance census record); its provenance the chain of ownership of objects, an adjacent tradition to semantic provenance with the same structure on a different object, outside the AI-era discourse and older than it, archival provenance being the parent concept semantic provenance extends (#725); evidence of the kind (manuscripts, signatures, watermarks, chain of custody) that retrocausal canon formation extends into the retrieval layer (#43); a catalogue's record of where a manuscript is held, which does not analyse its transmission (#610); and, at its address on the composition layer, the best-composed of the census rows, its licence left out (Capture Registry) PE-03 PE-04 PE-05 PE-06 PE-07 PE-15 PE-16 PE-17 O725-02 O725-01 O725-07 O725-03 O725-06 O43-01 O610-01 CAP-1-07 CAP-1-04; defining papers #725, #43, #610

E_F, the field's entity (type C — a conventional reading holds it (public entity: a crowd-sourced relational database of manuscript observations, with its operator and the aggregator that harvests it)): the Schoenberg Institute for Manuscript Studies' online, user-maintained, open-access database at sdbm.library.upenn.edu, which continuously aggregates and updates observations of pre-modern manuscripts drawn from over 14,000 auction and sales catalogues, inventories and institutional and private collection catalogues that document the books' sales and locations (13,000 in 2015); its unit the observation of a manuscript at a point in time, linked observations forming a manuscript record; more than 240,000 records in the 2020 MMM papers; contributed by registered users, searchable, downloadable in .xlsx or .csv under CC BY-SA 4.0, with credit required and no fee, and a SPARQL endpoint; one of the three source datasets of Mapping Manuscript Migrations; with stated limits (generic titles, current location inferred, no place for sales and gifts)

ρ = extends: the archive reads E_F as the entity and adds senses, mechanisms and limits on the field's own object: what the database does with provenance in the census schema (records, does not test; PE-04, PE-17, PE-18, PE-16), the relations and carrier it carries (PE-06, PE-07), the count's unchecked state (PE-15); the place of object provenance among provenance's traditions, adjacent to semantic provenance with the same structure and a different object, older than the AI-era discourse and outside it (O725-02, O725-01, O725-07), with archival provenance the parent concept semantic provenance extends and AI synthesis its new adversary (O725-03, O725-04, O725-05, O725-08), under the frame's own limit (O725-06); the extension of manuscript evidence into the retrieval layer (O43-01); the catalogue's record of location against transmission (O610-01); and what the composition layer keeps and leaves out at the address (CAP-1-01, CAP-1-04, CAP-1-07) PE-04 PE-17 PE-18 PE-16 PE-06 PE-07 PE-15 O725-02 O725-01 O725-07 O725-03 O725-04 O725-05 O725-08 O725-06 O43-01 O610-01 CAP-1-01 CAP-1-04 CAP-1-07. Guard: ρ = extends claims an addition and no more: none of the admitted deposits names the SDBM (the string pass returned no D or R candidate; every admission is O-class, by the class 'manuscript provenance' the field sources at B8, B9 and B10 and the census's lineage and relation types state), so the archive's claims bear on the database as a member of that class and are composed at that grade; #725's art-history row is a disambiguation rule ('Same conceptual structure; different object'), not a finding about any database; #610's sentence is an analogy inside an argument about numismatics and is carried at interpretation; the census record's 'records' and 'not tested' are its schema's classification of the resource, which the database's own pages do not contest; the field's counts, dated 2015, 2020 and 2026, are carried by date and composed beside the record's unchecked count, not corrected by it.

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity. The provenance census record describes the database from its own home page and the MMM white paper, field by field with the quote and the hash of each copy, and adds what its schema asks of every resource: what provenance it carries (sale, location, ownership), through which operation (it records; it does not test), on whose assertion and with what evidence of survival (none tested). The deposits that bear on its class take provenance of objects as given and place it: #725 names art-historical provenance, the chain of ownership, an 'Adjacent tradition' with the 'Same conceptual structure; different object', and archival provenance the 'Parent concept' that semantic provenance extends, and it describes its own relation to the existing frameworks as 'Extension and completion, not substitution or critique'; #43 extends provenance studies' archival evidence into the retrieval layer; #610 says what a catalogue does not record. The Capture Registry adds the composition's account at the entity's address. None of these contests what the database is; each adds a sense, a mechanism or a limit on the field's own object. ρ = extends; the equal-terms case (§0.9, §1.2a): one claim space, nothing translated.

#725 header, Relation: Extension and completion, not substitution or critique

#725 §0 Executive Symbolon, last paragraph: It is offered as a constructive extension of existing frameworks — not a critique of their adequacy in their own domains.

#725 §4 Disambiguation Matrix, row 'Provenance (art history)': | Provenance (art history) | Documented chain of ownership and attribution for art objects | Adjacent tradition | Same conceptual structure; different object |

#725 §4 Disambiguation Matrix, row 'Provenance (archival)': | Provenance (archival) | Origin and chain of custody of records | Parent concept | Semantic provenance extends archival concerns to circulating meaning under AI synthesis |

#43 § Relation to Adjacent Frameworks ¶4: RCF extends provenance into the retrieval layer: the DOI, the timestamp, the cross-linked citation network are the digital equivalents of the watermark and the scribal hand.

#PE census.json organizing_question: What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?

#PE census.json rows[slug == 'schoenberg-database'], capability; provenance_object: records … an observation of a pre-modern manuscript in a catalogue, inventory or other source

#capture schoenberg-database-of-manuscripts-aio-20261029, analysis: The best-composed of the census rows: host, unit, scale and structure at the database's own grain, from its own and its neighbours' pages. Only the licence, the condition of reuse, is left out.

equal-terms case: every field claim composed as received, beside the archive claims it stands with. Reading the received column recovers the field (§5.9).

v0.9 §4.10: KO_B sentences surviving in the KO at similarity ≥ 0.95, with positions (computed as apply_v09.py does: difflib.SequenceMatcher ratio, the best KO sentence for each KO_B sentence): v0.9 0 of 12 (positions [None, None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.9, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

The Schoenberg Database of Manuscripts "continuously aggregates and updates historic observations of pre-modern manuscripts" from over 14,000 catalogues and inventories documenting the books' sales and locations (SDBM home page), an "online, user-maintained, open-access, collaborative" resource among the Penn Libraries' digital projects (Penn Libraries, snippet only). A 2015 catalogue entry gives it over 13,000 sources and classes it as a crowd-sourced relational database and catalogue (MDR), and AWOL reprinted its page in 2020 (AWOL). The Schoenberg Institute for Manuscript Studies has at its core the Lawrence J. Schoenberg collection, donated in 2011 (About SIMS, snippet only); the collection, nearly 300 manuscripts and documents, is a separate object (Penn Libraries, snippet only), and the Institute's journal Manuscript Studies is to become fully Open Access (Penn Press, snippet only). Its unit is the observation of a manuscript in a sales, auction or collection catalogue, linked observations forming a manuscript record (MMM white paper), with name and place authorities and a SPARQL data model among its pages (SDBM home page). It holds more than 240,000 records (JOHD 2020; MMM white paper), and the three leading source types of its facet rose from 241,929 entries in 2020 to 259,787 as read in 2026 (AWOL; SDBM home page). Users contribute observations and batch uploads (SDBM home page; MDR); it can be searched, browsed, downloaded as .xlsx or .csv and queried through a SPARQL endpoint (MDR; MMM white paper), and its content runs from 500 to 1600 (MDR). Downloaded results are licensed CC BY-SA 4.0 (SDBM home page; AWOL); credit is required and no fee charged (MDR). Its carriers are a VIAF-based name authority and enhanced provenance data (SDBM home page; AWOL), Phillipps and the Bibliothèque nationale de France at the head of its provenance facet (SDBM home page), though the Institute is ineligible for VIAF and MMM corrected hundreds of its names (MMM white paper). The white paper records its limits: generic titles rather than FRBR works, current location inferred from the latest observation, no place for sales and gifts, and six of 25 research questions unanswerable from it alone (MMM white paper). It is one of three source datasets of Mapping Manuscript Migrations (2017–2020), which turned them into Linked Open Data without amending them and releases the aggregate under CC BY-NC 4.0 (MMM white paper; JOHD 2020). The project is ongoing (MDR), the New SDBM dates from 1 April 2017, and the forum's latest post as read is over two years old (SDBM home page). Wikipedia has no article on the database or its institute, the name appearing only in citations (Wikipedia), and a Bibliographical Society of America video introduces it (video card).

The Schoenberg Database of Manuscripts (SDBM, sdbm.library.upenn.edu) is "an online, user-maintained, open-access, collaborative rese[arch …]" in the words of its operator's page on the Penn Libraries site (Penn Libraries, Digital Projects, snippet only), and a database in the lineage of manuscripts and books that attaches provenance to "an observation of a pre-modern manuscript in a catalogue, inventory or other source" (provenance census record). Its home page states what it does: it "continuously aggregates and updates historic observations of pre-modern manuscripts drawn from over 14,000 auction and sales catalogs, inventories, catalogs from institutional and private collections, and other sources that document the sales and locations of these books from around the world" (SDBM home page); the same sentence stands with "over 13,000" in Lisa Fagin Davis's entry of 26 August 2015 for the Medieval Academy's Medieval Digital Resources (MDR), and with 14,000 but without "historic" in the copy Chuck Jones reprinted on AWOL on 10 May 2020 (AWOL). Medieval Digital Resources classes it as a catalogue, a crowd-sourced database, downloadable data, Linked Open Data, a relational database and a searchable database (MDR), and the Capture Registry records it as not the archive's, run by the Schoenberg Institute for Manuscript Studies, University of Pennsylvania (Capture Registry, 2026-10-29). The census record names the same operator (provenance census record); the Institute's pages on the Penn Libraries site list the SDBM among its digital projects (Penn Libraries, snippet only) and say that "At the core of SIMS is the Lawrence J. Schoenberg collection of manuscripts, which was donated to the Penn Libraries in 2011" (About SIMS, snippet only), and Penn Press announces that the current and back content of the journal Manuscript Studies is to become fully Open Access (Penn Press, title and snippet only). The Lawrence J. Schoenberg Collection on the same rail is the collection and not the database: "Nearly 300 manuscripts and documents dating from ca. 1900 BC to the 20th century" (Penn Libraries, snippet only). The unit is the observation: the Mapping Manuscript Migrations white paper, written by the project team with the Institute's staff among its authors, contrasts Bibale, which "takes a single manuscript as the basic unit", with the SDBM, which "focuses on an “observation” of a manuscript in a sales or auction catalogue or collection catalogue" and "consists of records which are observations of a manuscript at a point in time", two or more linked by a "manuscript record" (MMM white paper), and the census record gives the unit as "an observation (an entry in a source that documents a manuscript)" (provenance census record). The Capture Registry finds the unit held in the composition at the address, entries as observations linked under manuscript records, "the record's unit" (Capture Registry, 2026-10-29), and the home page's menus name an SDBM Name Authority, an SDBM Place Authority, an Entity Relationship Diagram and a SPARQL Data Model (SDBM home page). In the census record's terms the SDBM records provenance: its capability is "records" among the schema's expresses, records and tests, the relations it carries are sale, location and ownership, its carrier is "observation records with a VIAF-based name authority", and under the schema's rule that "Untested capability reads 'not tested'" it carries no tested transformation and no retention result, the record asking of it what it asks of every resource: "What provenance does this system carry, through which operation, on whose assertion, with what evidence of survival?" (provenance census record). Its own pages name those carriers, a "VIAF-based Name Authority", "Enhanced Provenance data" and a SPARQL endpoint (SDBM home page; AWOL's 2020 copy), and its provenance facet as read is headed by the Bibliothèque nationale de France (23,741), Sir Thomas Phillipps (21,582) and the British Library (7,822) (SDBM home page). The white paper records what the name authority lacked and gained: the Institute "is ineligible to participate" in VIAF, "despite offering to contribute", and through MMM "hundreds of personal and institutional names have been corrected for authority control" in the Schoenberg Database and Bibale (MMM white paper). Its relations of sale, location and ownership (provenance census record) are those of what the MMM data paper calls a specialist database focused on "the history and provenance of medieval and Renaissance manuscripts" (JOHD 2020), and the archive dates that provenance as centuries old: "Provenance" is "a centuries-old concept in archival science, art history, and legal evidence", and art-historical provenance, "Documented chain of ownership and attribution for art objects", is an "Adjacent tradition" to semantic provenance, "Same conceptual structure; different object" (#725). #725 lists provenance's disciplinary homes, "art history (chain of ownership, attribution)" second among eight, and holds that "Each tradition answers a specific question about origin", the AI-era discourse sitting "at the intersection of the last four", which are data provenance, content authenticity, dataset documentation and digital preservation (#725). Of archival provenance #725 says that "Archival theory has long insisted that provenance is contextual and meaning-bearing", that "An archival custody chain matters because it preserves the meaning of records", and that the value of preserved lineage "is realized only when the descendants of a work need to find their way back to its sources"; semantic provenance "extends archival concerns to circulating meaning under AI synthesis", and what AI synthesis introduces is "the first adversary capable of stripping that meaning dimension at machine scale" (#725). #725 bounds its own frame: its three dimensions "reflect the current institutional landscape, not an ontological claim that provenance is exhausted by three categories" (#725). #43 names the evidence provenance studies rests on and moves it: "Provenance studies stabilizes authorship and priority through archival evidence — manuscripts, signatures, watermarks, chain of custody", and retrocausal canon formation "extends provenance into the retrieval layer", where the DOI, the timestamp and the cross-linked citation network are "the digital equivalents of the watermark and the scribal hand" (#43); #610, by Rex Fraction, marks what a catalogue does not do: "A catalog records that a manuscript exists, where it is held, what it contains. It does not analyze the scribal tradition, the variant readings, or the politics of textual transmission" (#610). The counts are of sources and of records, and each carries its date: over 13,000 sources in the 2015 entry (MDR), over 14,000 in AWOL's 2020 copy and on the home page as read (AWOL; SDBM home page), "more than 240,000 records for manuscript observations" in MMM's data paper of May 2020 (JOHD 2020) and "240,000 records" in the white paper (MMM white paper). The home page's source-type facet has its own figures: 122,751 collection catalogues, 99,319 auction and dealer catalogues and 19,859 unpublished sources in AWOL's copy of 2020, and 129,525, 106,401 and 23,861 on the page as read in 2026 (AWOL; SDBM home page), 241,929 and then 259,787 entries for those three types alone. The census record gives the count as "over 14,000 auction and sales catalogs, inventories and other sources; 240,000 records (MMM white paper)" and lists "the current record count on the database itself" as not yet checked (provenance census record), and the Capture Registry finds the composition's two counts to be the record's two, "joined in one sentence" (Capture Registry, 2026-10-29). It is built by its users: members are invited to "help us to build and maintain this resource" and to contribute their "own personal observations of a manuscript or group of manuscripts" singly or by batch upload (SDBM home page; MDR), which the census record states as continuous intake by registered users (provenance census record). It may be searched or browsed "by author, title, seller, provenance, date, and others" and downloaded "in .xlsx or .csv format" (MDR), and the white paper reports a public SPARQL endpoint over "the pre-transformation RDF version of the data" (MMM white paper; the census record: search, download, SPARQL endpoint); MDR dates its content "500 - 1600" across Europe, the Middle East and East and South Asia (MDR), the period the Capture Registry records as unchecked against the census record (Capture Registry, 2026-10-29). The condition of reuse is stated three ways: "Must provide credit" beside "No Fee" (MDR, 2015); "Download Search Results for your own use under a Creative Commons Attribution-ShareAlike 4.0 Unported License" on the home page, reprinted by AWOL in 2020 (SDBM home page; AWOL); and "CC BY-SA 4.0 for downloaded search results" (provenance census record); MMM, which aggregates the SDBM, releases its own data "under a CC BY-NC 4.0 license" (JOHD 2020). The Capture Registry finds the licence unsaid in the composition at the address: export given, the CC BY-SA 4.0 quote absent (Capture Registry, 2026-10-29). MDR gives the project's status as ongoing (MDR); the home page dates the New SDBM's changes "As of April 1, 2017", made "thanks to input from the user community", and its recent-activity column as read shows its latest post, in the thread 'Ketterer Auctions - 30 November 2020', aged "over 2 years" (SDBM home page); the census record has it "active" (provenance census record). The white paper states the database's limits: it "tracks generic titles rather than works in the specific sense employed by FRBR"; a manuscript's current location "is not specifically present in the Schoenberg Database, although it can often be inferred from the most recent observation"; its entries for sales and gifts "did not have place information associated with them"; and 6 of the 25 research questions could not be answered from it alone (MMM white paper). The SDBM is one of the three source datasets of Mapping Manuscript Migrations, the 2017–2020 Digging into Data project that "took the data from three existing databases and turned them into Linked Open Data" (MMM white paper), in whose data paper of May 2020, the Institute's staff among its authors, an SDBM manuscript has a URI such as http://ldf.fi/mmm/manifestation_singleton/sdbm_784 and "The original data have not been corrected or amended in any way by the MMM project" (JOHD 2020); the census record carries the relation as part of mapping-manuscript-migrations (provenance census record). The encyclopedia has no article on the database or its institute: the titles Schoenberg_Database_of_Manuscripts and Schoenberg_Institute_for_Manuscript_Studies answer HTTP 404, and the quoted name occurs in article space only as a cited source, in Black Hours, Morgan MS 493 and Sister Carrie (Wikipedia, 2026-10-29); the Bibliographical Society of America's channel carries an "Introduction to the Schoenberg Database of Manuscripts (SDBM)" (video card, title only). At its address on Google's AI Overview the Capture Registry reads the composition innocently first, the database and its host identified, what it traces and from how many sources and its structure laid out, and judges it "The best-composed of the census rows", host, unit, scale and structure "at the database's own grain", with "Only the licence, the condition of reuse," left out (Capture Registry, 2026-10-29). The Registry finds the Institute and Penn Libraries named in the composition and no person or identifier, a provenance erasure rate of 0.5, keeps the stray full stop of 'from more. than 14,000' as received, and notes that the census record landed the same day and that no card cites it, its findings checked against the census record and the pages it saved (Capture Registry, 2026-10-29).

T against L_F(B) (AIO's ontology at work on its own cards): At 'Schoenberg Database of Manuscripts' (T1, eight cards: B1–B8; 211 words) the composition gives the database by its neighbours' pages and its operator's. Composed from its cards: the definition at the grain of the operator's snippet and the catalogue entry ('open-access' and 'user-maintained' from B2 F10, 'crowd-sourced' from B1 F6 and F5, 'free … without a fee' from B1 F7), the host from the Penn cards (B2 F11, B5 F14); the purpose at the grain of the MMM data paper ('the history and provenance of … manuscripts', B8 F22); the scale by joining two cards' figures in one sentence, B8's 'more than 240,000 records for manuscript observations' (F21) and B6's 'over 14,000 auction and sales catalogs, inventories, catalogs from institutional and private collections' (F16), which the Capture Registry records as the record's two counts (CAP-1-02); period and region from the 2015 catalogue entry (B1 F8: 'Dates: 500 - 1600'; Asia, Europe, the Middle East), the period the Registry marks unchecked (CAP-1-05); access at B1's grain (F4: searched, browsed, .xlsx or .csv). Limit dropped: B1's licence field reads 'Must provide credit , No Fee' (F7); T composes the fee half ('Free to use without a fee') and not the credit half, the attribution condition; and B6, a card on the same rail, reprints the licence sentence whole, CC BY-SA 4.0 (F17), which T leaves out: the Capture Registry's 'THE LICENCE UNSAID' (CAP-1-04) and its analysis, 'Only the licence, the condition of reuse, is left out' (CAP-1-07). Limit dropped, temporal: B8's count is dated May 2020 (F25) and B6's copy 10 May 2020 (F19); T composes 'Contains over 240,000' in the present, with no date. Available, not composed: the SDBM's place in Mapping Manuscript Migrations (B8 F22's 'MMM combines data from three specialist databases', composed only as the purpose, and the aggregation unsaid), the aggregator's limit that the source data are not amended (F23), its CC BY-NC 4.0 licence (F24), the SDBM manuscript as an MMM URI (F26) and the paper's authors (F25); the 2015 and 2020 versions of the self-description and their counts (B1 F2, F3; B6 F16 composed only as the 14,000), the 2020 source-type facets (B6 F18), the project's status (B1 F1: 'Ongoing'), Linked Open Data among the entry's classes (F6, composed only as 'crowd-sourced'), the reprinted carriers, a VIAF-based name authority and enhanced provenance data (B6 F20), the journal's open access (B7 F15), the video (B3 F12) and AWOL's post (F19), the catalogue entry's date and author (F9). Kept apart, not assigned: the Lawrence J. Schoenberg Collection (B4 F13) is not composed as the database; its range, ca. 1900 BC to the 20th century, is the only range on the rail beyond 500–1600, and T's '(though holdings span broader ranges)' is at the grain of no readable card. Outside the cards' text: the five record types and their glosses (T1-6; the authorities are on the home page's menus, B9 F31, and the observation and the manuscript record are the white paper's, B10 F36, F38; neither surfaced); 'from their creation to the present' (T1-2). Composed without attribution, by grain: no sentence names its source; the markers are opaque tokens; the counts are at their sources' own grain ('over 240,000 manuscript observation records' against B8's 'more than 240,000 records for manuscript observations') with markers only; the Institute and Penn Libraries are named as host, no person and no identifier (the Registry's PER 0.5, CAP-1-10). Content replaced by use: the closer (T1-7) offers to guide a search. No force raised, no substituted reading observed.

v0.9 §0.5: the first diff is ontological and small. T and the field agree on the entity (the Institute's crowd-sourced database of manuscript observations); T composes it as a tool to use and leaves out what its own cards give as the conditions of use and of origin: the licence and its credit condition, the dates of its counts, the aggregator that harvests it. The database stands without the terms on which it can be taken up.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: what the database does with provenance in the census schema, it records and does not test, no transformation tested and no retention result (PE-04, PE-17, PE-18), under the record's question of every resource (PE-16); the relations it carries, sale, location and ownership, and its carrier (PE-06, PE-07); the count's unchecked state (PE-15), beside which the field's own figures stand by date (2015, 2020, 2026); the place of object provenance among provenance's traditions, the chain of ownership an adjacent tradition to semantic provenance with the 'Same conceptual structure; different object' (O725-02), older than the AI-era discourse and outside its four homes (O725-07, O725-01), archival provenance the parent concept that semantic provenance extends to meaning under AI synthesis, its new adversary able to strip the meaning dimension at machine scale (O725-03, O725-04, O725-05, O725-08), under the frame's own limit that three dimensions do not exhaust provenance (O725-06); manuscript evidence extended into the retrieval layer, DOI and timestamp as watermark and scribal hand (O43-01); the catalogue that records where a manuscript is held and not its transmission (O610-01); and the composition's account at the address, the unit held, the counts joined, the licence unsaid, the period unchecked, the best-composed of the census rows, PER 0.5, the record uncited (CAP-1-01 to CAP-1-11). Not surfaced: no archive source is on T's rail; the census record landed the same day (CAP-1-09). The field is carried whole in both: the delta is the senses the archive adds, not a correction of the field's claims.

alignment: {'T1_cards_claims': ['F1', 'F2', 'F3', 'F4', 'F5', 'F6', 'F7', 'F8', 'F9', 'F10', 'F11', 'F12', 'F13', 'F14', 'F15', 'F16', 'F17', 'F18', 'F19', 'F20', 'F21', 'F22', 'F23', 'F24', 'F25', 'F26'], 'T1_composed': ['F10', 'F6', 'F5', 'F7', 'F11', 'F14', 'F22', 'F21', 'F16', 'F8', 'F4'], 'T1_outside_cards': ["T1-6 the five record types and their glosses (Name and Place Authority are on the home page's menus, B9 F31; the observation and the manuscript record are the white paper's, B10 F36/F38; neither surfaced; the unreadable cards B2, B5 and the video B3 may carry them, which the field cannot settle)", "T1-2 'from their creation to the present' (no readable card states it)", "T1-4 '(though holdings span broader ranges)' (no readable card states it; the only range beyond 500–1600 on the rail is B4's, the Lawrence J. Schoenberg Collection's 'ca. 1900 BC to the 20th century', a distinct entity; whether T drew on it the field cannot settle)"], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; the typology of what is not is in T_vs_LB (§6.2)"}