Project Gutenberg

register entry project-gutenberg-aimode-20261101 OBS-d3c1028b7ac2 · file as first run sha256 231ff7db2f155deb…

• T1-1 definition: 'is the world's oldest digital library, providing nearly 80,000 free eBooks primarily focused on public domain literature.' (chip 1, 'Project Gutenberg +1', after the opening paragraph)
• T1-2 founder and purpose: 'Founded in 1971 by American writer Michael S. Hart, the project is a massive volunteer-driven effort dedicated to digitizing, proofreading, and archiving cultural works to encourage global literacy and open access to historical texts.' (chip 1)
• T1-3 free access: '100% Free Access: Everything is available completely free of charge, with no registration, library cards, special apps, or user accounts required.' (chip 2, 'Project Gutenberg +1')
• T1-4 formats: 'Format Flexibility: Books can be read directly within your web browser or downloaded in open, long-lasting formats like EPUB, Kindle (MOBI), HTML, and plain text.' (chip 3, 'OpenEd +1')
• T1-5 US public domain focus: 'US Public Domain Focus: Because it operates under United States copyright law, the vast majority of its collection consists of books published more than a century ago whose copyrights have expired.' (chip 4, 'Wikipedia +1')
• T1-6 volunteers: 'Volunteer Driven: The digitization process relies on thousands of volunteers around the globe, largely collaborating through Distributed Proofreaders to clean up optical character recognition (OCR) errors, format files, and review submissions.' (chip 5, 'Project Gutenberg +2')
• T1-7 audiobooks: 'Audiobook Companions: The library is closely paired with platforms like LibriVox to provide volunteer-recorded audiobooks alongside written texts.' (chip 6, 'YouTube·Microsoft Cloud +1')
• T1-8 popular works: 'Pride and Prejudice by Jane Austen' … 'Frankenstein by Mary Shelley' … 'The Adventures of Sherlock Holmes by Arthur Conan Doyle' … 'Moby Dick by Herman Melville' … 'The complete works of William Shakespeare' (no chip)
• T1-9 closing offers: 'If you are looking to dive into the collection, let me know:' … 'Do you want to download books to a Kindle, tablet, or phone?' … 'I can guide you through finding and downloading the exact texts you need.' (no chip)

• B1 Project Gutenberg — gutenberg.org home page (chips 1, 2, 5) card; B1.html sha256 461dd6237aa9aa55, HTTP 200 at 2026-11-01T15:13:06Z; page identified by the chips' spans (see identification)
• B2 OpenEd — page unidentified (chip 3) unreachable: unidentified; site name only, no title, snippet or URL in the paste
• B3 Wikipedia — Project Gutenberg (chip 4) card; B3.html sha256 80cfad0162a08071, HTTP 200 at 2026-11-01T15:13:06Z; wikitext HTTP 200 (sha256 c04b2a45206eec23); the census's copy (sha256 05de9e5673bcf1d5) differs in bytes, same article text
• B4 YouTube·Microsoft Cloud — video unidentified (chip 6) unreachable: unidentified; site name only, no title, snippet or URL in the paste
• B5 Project Gutenberg — About Project Gutenberg (own page) own page; the census's saved copy B5.html sha256 4e1cfdb3edd2d17a (matches the manifest), HTTP 200 at 2026-11-01T04:07:47Z
• B6 Project Gutenberg — Terms of Use (own page) own page; the census's saved copy B6.html sha256 38e45e0cbc6a3f90 (matches the manifest), HTTP 200 at 2026-11-01T04:07:48Z
• B7 Wikipedia — The Pile (dataset) (census third-party) census third party; the census's saved copy B7.html sha256 1ba82f3edd2a4647 (matches the manifest), HTTP 200 at 2026-11-01T04:07:51Z
• B8 NaNoGenMo — National Novel Generation Month (census third-party) census third party; the census's saved copy B8.html sha256 d831aaf0d1a4efc0 (matches the manifest), HTTP 200 at 2026-11-01T04:07:58Z

E_A, the archive's entity: the same library, in the traininglayerliterature.org layer census record and the archive's deposits: a source library held apart from the datasets made from it, its terms addressing two readers, its texts passing into training and composition downstream with the name they carry after that passage outside its pages (census); infrastructure to the census's question of training membership at the source; a source and comparator of the archive's originals shelf, whose seatings keep its header and licence with each file, read the licence as governing the name and trademark, and identify the edition behind each transcription; the most widely used public-domain corpus in digital humanities, doing correct work at its layer with spare metadata TLL-22 TLL-23 TLL-24 TLL-14 TLL-21 TLL-12 O1680-01 O1710-01 O1710-02 O1707-01 O1712-01 O1044-01 O1044-02 O416-01 O735-01 O1046-02 O186-02 O1543-02; defining papers #186, #1543, #1710, #1712, #1044, #1046, #1680, #1711, #1560, #1707, #1689, #1720, #943, #416, #735

E_F, the field's entity (type C — a conventional reading holds it (public entity: a digital library of free eBooks)): Project Gutenberg, the oldest digital library, founded by Michael S. Hart in 1971 and operated by the Project Gutenberg Literary Archive Foundation (chartered 2000): a volunteer effort, with Distributed Proofreaders, to digitize and proofread works, chiefly public domain in the United States, released free in plain text and other formats (79,709 eBooks on its home page, 2026-11-01); its website for human users only, bulk and machine access sent to mirrors and a catalog; its name a trademark removable on resale; its texts carried downstream into the Pile's PG-19 component and into generated works

ρ = extends: The archive reads Project Gutenberg as its own pages and Wikipedia present it and adds, on the same object: the census's reading of the library as a source library held apart from the datasets made from it, with its two readers (TLL-22, TLL-23) and the questions it cannot answer (TLL-21, TLL-24); the seatings' practice, which keeps the library's header and licence with each file, reads the licence as governing the name, and identifies the edition behind each transcription (O1710-01, O1710-02, O1707-01, O1712-01, O1689-01, O1720-01), on the field's own trademark and TEI claims (F30–F32, F35); the archive's estimates of the library's texts and apparatus (O1044-01, O1044-02, O943-01, O416-01); the training corpus of a heteronym and the anchors' categories for texts in the corpora (O735-01, O186-02, O1543-02). TLL-22 TLL-23 TLL-21 TLL-24 O1710-01 O1710-02 O1707-01 O1712-01 O1689-01 O1720-01 O1044-01 O1044-02 O943-01 O416-01 O735-01 O186-02 O1543-02. Guard: ρ = extends claims an addition and no more: the archive's claims do not make the field's sources wrong at their dates; where the census's record.makes contradicts the home page (TLL-18, F10) and #1046 stands beside the computer-generated audiobooks (O1046-02, F10), both stand with their sources (P_coh), and the census's unanswered questions stay questions (open block).

The case: equal terms, ontology included (v0.9). The archive reads the field's entity as the entity. The census record describes Project Gutenberg from its own pages and Wikipedia and keeps the resource's terms (operator, kind, unit, record object, method) apart from its own reading (tll_reading), which reads the same library as a source library held apart from the datasets made from it. The archive's deposits name the library as a source of its own shelf and say what its texts are and lack; none contests that it is a library of volunteer-made public-domain eBooks.

#TLL census.json rows[slug == 'project-gutenberg'].tll_reading: The archive reads Project Gutenberg as a source library held apart from the datasets made from it.

#TLL census.json rows[slug == 'project-gutenberg'].tll_reading: Its texts reach training and composition downstream, in the Pile's Gutenberg (PG-19) component (Wikipedia) and as allowed raw material for NaNoGenMo's generated novels; the name a text carries after that passage is outside its own pages, and Wikipedia records texts resold with references to the project removed.

#1710 26/01 Licence: Project Gutenberg transcriptions under the Project Gutenberg License (the text is public domain in the United States; the licence governs use of the Project Gutenberg name and trademark), whose header each file carries.

#1044 § The second line: The NH Canon Editions ¶1: The strategic point is not to displace or duplicate Gutenberg's reading environment at Gutenberg's own layer — Gutenberg is doing correct work at its layer — but to construct the scholarly reading environment Gutenberg is architected not to construct

#1680 F4. The originals shelf — source editions, Comparators: **Comparators.** Its own sources (Perseus, Coptic SCRIPTORIUM, Project Gutenberg) and other digital text libraries holding the same edition or witness.

equal-terms case: every field claim composed as received, beside the archive claims it stands with; no receiving text. Reading the received column recovers the field (§5.9).

v0.9 §4.10 (unchanged in v0.10): 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): v0.10 0 of 15 (positions [None, None, None, None, None, None, None, None, None, None, None, None, None, None, None]; composed first under v0.10, so there is no v0.7 object to measure against). Opening-run check (§5.10): no breach.

Project Gutenberg is a volunteer effort to digitize and archive cultural works and to encourage the creation and distribution of eBooks; founded in 1971 by Michael S. Hart, it is the oldest digital library. It is an online library of free eBooks, most of them full texts of books or stories in the public domain in the United States, with a focus on older works whose U.S. copyright has expired. Its count stands at 79,709 on its home page, more than 75,000 on its about page and nearly 78,000 in Wikipedia (March 2026). The collection is mostly Western literature in English, with reference works, periodicals and a few non-text items. Hart began with the Declaration of Independence in 1971, named the project after Johannes Gutenberg, typed every text until 1989 and died in 2011. The Project Gutenberg Literary Archive Foundation, chartered in 2000, operates it; its CEO Gregory Newby died in 2025 and Eric Hellman is acting CEO. Volunteers digitize and proofread its texts, thousands or hundreds by its own two counts, with no selection policy; Distributed Proofreaders is the main source of new eBooks. Texts are added after copyright clearance; the Text Encoding Initiative criticized it in 1994 for not documenting which version of a text was digitized. Releases come in plain text, HTML, EPUB and MOBI, free and without registration; the plain text has been criticized as hard to read on screen. Its audience is US persons over 13, it gives no warranty abroad, and it has been blocked in Germany and Italy. Its website is for human users only; automated access is blocked, bulk downloads go to mirrors, and a machine-readable catalog, an index file and an OPDS feed under stated rules serve other uses. "Project Gutenberg" is a trademark barred from commercial or modified redistributions; resale is lawful with all references removed, and books stripped of attribution have been sold. LibriVox records its texts as human-read audiobooks, and the Open Audiobook Collection adds almost 5,000 computer-generated titles; the Internet Archive hosted it earlier. Downstream, the Pile, a dataset for training language models, carries a "Gutenberg (PG-19)" component of 11.68 GB, and NaNoGenMo allows a generated novel that grabs a random novel from it. Its most popular titles stand under their authors; the Wikipedia article was last edited on 12 October 2026.

Project Gutenberg is an online library of free eBooks and the oldest digital library, founded by Michael S. Hart in 1971: a volunteer effort to digitize and archive cultural works whose mission is "To encourage the creation and distribution of eBooks." Most of what it holds are full texts of books or individual stories in the public domain, free to access in an open format, with a focus on older works whose U.S. copyright has expired; its unit is the eBook, a digitized and proofread text with its own permanent landing page. The collection is primarily literature of the Western cultural tradition, novels, poetry, short stories and drama, with cookbooks, reference works, issues of periodicals and a few non-text items such as audio files, movies and music-notation files; most releases are in English, and many are in other languages. Hart's stated goal was "to provide as many e-books in as many formats as possible for the entire world to read in as many languages as possible", and a project slogan is to "break down the bars of ignorance and illiteracy". Hart began the project in 1971 by digitizing the United States Declaration of Independence and named it after Johannes Gutenberg, the fifteenth-century printer; its own pages say that Hart "invented eBooks in 1971" and count "55+ Years" of free eBooks. Hart entered all text by hand until 1989, when scanners and optical character recognition made book scanning feasible, and died on 6 September 2011. The Project Gutenberg Literary Archive Foundation, a 501(c)(3) corporation chartered in Mississippi in 2000 to handle the project's legal needs, operates it and accepts donations; Gregory B. Newby became the foundation's first CEO in 2001, the home page now refers to its late CEO, Dr. Greg Newby, and after Newby's death in 2025 Eric Hellman became acting CEO. ibiblio at the University of North Carolina at Chapel Hill hosts it; the Internet Archive was its long-time backup distribution site and an earlier main host. Counts differ by date and source: 79,709 free eBooks on the home page as read on 1 November 2026, more than 75,000 on its about page, nearly 78,000 items as of 22 March 2026 and over 77,500 documents in Wikipedia's infobox, and over 75,999 items claimed in May 2025, with over 30 new e-books a week. A snapshot of the Gutenberg catalog taken into the archive in July 2026 holds 90,252 rows, and the archive records that Das Kapital is not on Gutenberg in German, so that a comparison of Hegel's prose with Marx's cannot be run on the open German shelf. By its own account thousands of volunteers have digitized and proofread its books; the same home page also says that hundreds of volunteers have contributed over the years. No selection policy governs what is added: volunteers work on what interests them or what they have, and the works and editions available have been set by popularity, ease of scanning, copyright status and other factors. Distributed Proofreaders, founded by Charles Franks in 2000 and affiliated in 2002, distributes the proofreading of scanned texts among volunteers and is now the main source of new eBooks; its 36,000+ books were almost two-thirds of the collection in 2018. Material is added only after a copyright clearance, whose records are kept, and the project claims no new copyright on the titles it publishes. In December 1994 the Text Encoding Initiative criticized the project for not documenting the decisions made in preparing a text and, in some cases, which of several conflicting versions had been digitized; the census records it as no scholarly edition, unable to say which edition a text was digitized from. The archive's seatings identify the edition behind a Gutenberg text from the text itself: Morris's Pearl from the second edition of 1869, The Prophet from a later Knopf printing whose book text is that of 1923, a Monadologie catalogued as 1909 whose title page reads 1900, each of Kafka's works as the edition its header names, a Principia transcribed from the first edition of 1687 with its transcribers' corrections listed, and an Opticks of the fourth edition of 1730, produced by Distributed Proofreaders, held beside the first printing of 1704. The archive also finds a Gutenberg text short of what some works need: its Dickinson "is a propositional shell", its Hopkins "lacks the notational apparatus", and works cited from its texts are cited by chapter, for want of print pagination. The archive calls Gutenberg's texts "the most widely-used public-domain corpus in digital humanities", their reading environment unchanged for three decades and their metadata spare, and holds that "Gutenberg is doing correct work at its layer", while it sets out to build the scholarly reading environment "Gutenberg is architected not to construct". An earlier archive paper counts large text archives such as Project Gutenberg and HathiTrust as collections with scale and no internal architecture, of which an AI summary describes what is in them. Most books are distributed as public domain under United States copyright law, a few copyrighted texts with permission; the target audience is United States persons over 13, and the project offers no warranty about copyright status or freedom to use its materials outside the United States. Its website was inaccessible in Germany from 2018 to 2021 under a court order concerning works of Heinrich Mann, Thomas Mann and Alfred Döblin, and it has been blocked in Italy since May 2020. In the archive's account the U.S. public-domain constraint leaves its translations mostly older ones, Victorian, Edwardian and early twentieth-century. Releases come in plain text, a Latin-character version having been required since the founding as the format Hart believed most likely to stay readable, and in HTML, EPUB and MOBI; the plain files, wrapped at 65–70 characters, were criticized in 2000 as "bland and difficult to read on screen". All its eBooks are free, "and always will be", with no fees, registration or apps required. Its terms address two readers: "This website is intended for human users only", and automated access brings a block of the IP address, while anyone downloading many books, more than about 100 a day, by hand or by software, is sent to the mirrors, and a machine-readable database of all its books is offered at its Offline Catalogs and Feeds page beside an index file. Applications using its OPDS feed must identify themselves and keep to a browser's volume, high-volume ones being pointed to the catalog database to host their own feed; large-scale deep-linking to its files and mock front-ends that take advertising revenue are refused, and links go to an item's landing page, which it intends to be permanent. The home page links a page on linking to the site and "roboting or crawling" it, and the collection is backed up regularly and mirrored on servers in many places, so that no one localized accident can lose it. The census reads Project Gutenberg as a source library held apart from the datasets made from it, its terms addressing the human reader at the website and machine access at the mirrors and the catalog, and records its readback by landing page, search, index file and catalog database. Its texts pass into training downstream: the Pile, an 886 GB dataset of English text made to train large language models, lists among its sub-datasets "Gutenberg (PG-19)" at 11.68 GB, unmarked among the components with copyright disputes. NaNoGenMo's rules allow a generated novel that could "literally grab a random novel from Project Gutenberg", and the library's own home page lists almost 5,000 computer-generated audiobook titles in the Open Audiobook Collection beside 662 titles read by people and LibriVox's human-read recordings of its texts. An archive update of July 2026, listing venues for an AI-mediated reading text of its own, records "Not Gutenberg (AI-mediated text rejected)"; the statement concerns text and stands beside the computer-generated audiobooks the home page lists. A pre-registered training corpus by the archive's heteronym Nobel Glas takes 30% of its documents from Project Gutenberg's public-domain prose before 1900, filtered for literary stature, as high training signal. The census records that the library's own saved pages say nothing of training and that what an apparatus makes from its texts is unrecorded there, that the library is no training dataset and the datasets and generated works drawn from it are different entities, and that whether a given eBook was carried into a given training set cannot be answered from it; the home page, which the census did not save, lists the computer-generated audiobooks. In the census's anchors training-layer literature is writing addressed to the layer where texts are "crawled, indexed, trained on, retrieved, and recomposed into answers", texts in the corpora that are not organized around that fact are "read by machines incidentally", and the institutional training loop runs from a text through a model and its output into the next corpus. #1543 names the second wing of that literature machine composition, "what the encountered work makes it possible for a machine to WRITE", and keeps apart whether a machine wrote something new and whether it told the truth about where it came from. "Project Gutenberg" is a trademark that cannot be used in commercial or modified redistributions of the project's public-domain texts; the sister projects are independent organizations permitted to use it, and one of them, Projekt Gutenberg-DE, claims copyright for its product. Resale of its public-domain works is lawful once every reference to Project Gutenberg is removed, contributors have questioned commercial reuse of content formatted by volunteers, and books stripped of attribution to the project, Fox Trapping (1906) among them, have been sold in the Kindle Store and by other booksellers. What may be done with a downloaded eBook falls under the Project Gutenberg License, which the archive reads as governing the use of the Project Gutenberg name and trademark, the text being public domain in the United States. The archive keeps the name with the text: its seatings record a Gutenberg file as "Project Gutenberg packaging under the PG licence", seat it "with its Gutenberg header and licence text intact" or with "its header and footer kept as served", and name each one a transcription beside the scan of the same edition where an open scan was found. The census places the name a text carries after its passage into training and composition outside the library's own pages and finds authors' names unrecorded on the pages it saved, while the home page lists its most popular works each under its author; in #186 persistence is a work placed where the training and retrieval layers look, "with the metadata that lets it be found and attributed". Project Gutenberg is one of the archive's originals shelf's own sources and a comparator for it: the shelf counts 16 seats in its Gutenberg licence class, points to the mantle-bearing Leaves of Grass of 1891–92 at Project Gutenberg #1322 outside the archive, holds Philip K. Dick's unrenewed stories as open in its transcriptions from their magazine printings, checked against its catalogue, and rests the reproducibility of a stylometric run on it among other open sources. The census asks how literature is addressed to a receiving apparatus, what that apparatus makes from it, and how transmission, transformation and provenance can be inspected across later reception; it admits aggregators that gather and hold many works or records and say what they hold, and enters Project Gutenberg as infrastructure, a library whose question is training membership at the source: whether a public-domain work is held, at which landing page, and on what terms a machine may take it. It leaves unchecked the Project Gutenberg License, the Offline Catalogs and Feeds page, the mirrors list and every eBook landing page, so the author and metadata fields of a release are unseen in its record. Asked for "Project Gutenberg" on 1 November 2026, Google's AI Mode, run there by the operator's note in place of an Overview, said what the library is, when and by whom it was founded, how it is made and what it holds, listed five famous works under their authors and offered help finding and downloading texts; it gave its sources as six inline chips, at a new address. The Capture Registry found the library composed accurately as a reader's library, its opening and founder following Wikipedia and its volunteers and Distributed Proofreaders at their sources' grain, and its mission widened from the creation and distribution of eBooks to global literacy and open access. The registry also found that the composition addresses the reader of the browser and the Kindle, carries neither the operating foundation, the human-users rule, the mirrors, any word of training, the trademark nor the policy's limit outside the United States, and gives a count, "nearly 80,000", above every saved source and every result on the page; the home page, read after the capture, gives 79,709. The Wikipedia article read here was last edited on 12 October 2026; two of the six chips, OpenEd and a YouTube video of Microsoft Cloud, carry no title, snippet or link and stand unidentified.

current · derived from the current composition

T against L_F(B) (AIO's ontology at work on its own cards): At 'Project Gutenberg' (T1, AI Mode, six inline chips; the disclosed pool B_D is B1 gutenberg.org for chips 1, 2 and 5, B2 OpenEd and B4 a YouTube video of Microsoft Cloud, both unidentified and without text, and B3 Wikipedia; B5–B8, the about page, the terms of use, the Pile article and NaNoGenMo, are the reference pool and stand outside this comparison, §0.3), T composes a definition with founder and purpose, five bulleted features, five works under their authors and three offers. Composed at its source's grain or coarser: the oldest digital library founded in 1971 by Hart (F14), a volunteer effort to digitize and archive cultural works (F13), the focus on older works whose U.S. copyright has expired (F2, F15, F29), free access without fees, registration or apps (F4), the formats (F18), thousands of volunteers (F3) and Distributed Proofreaders (F8, F19), LibriVox (F9, F42), and the most popular works under their authors (F11, five of the home page's list in its order of popularity, Alice and Dracula left out). 'nearly 80,000' composes F1 (79,709) at a coarser grain. Substituted reading: 'Audiobook Companions … to provide volunteer-recorded audiobooks alongside written texts' (T1-7) composes the audiobooks as volunteer-recorded; the home page under its chips lists 662 titles read by people, LibriVox's human-read recordings and 'almost 5,000 computer-generated titles' in the Open Audiobook Collection (F10), the larger share made by machine. Force raised: 'to encourage global literacy and open access to historical texts' (T1-2) composes the mission at a wider scope than F48's 'To encourage the creation and distribution of eBooks' (B5, the reference pool); in B_D, Wikipedia's Ideals carry the slogan 'break down the bars of ignorance and illiteracy' and Hart's goal 'for the entire world to read' (F63), so the widening has a disclosed basis at the grain of the project's ideals, and the Capture Registry's 'mission is widened' (CAP-1-03) holds against the mission statement. 'massive' (T1-2) has no field claim. Available and not composed (§6.2): the operating foundation (F37, F49 in B_R), its CEOs (F38; F7 'our late CEO' on the home page), the trademark and the stripping of the project's name from resold books (F30, F31, F32), the sister projects (F40, F41), the copyright clearance (F28), the absence of a selection policy (F26, F36), the TEI criticism of undocumented versions (F35), the plain-text criticism (F34), the blocks in Germany and Italy (F33), the long-term mirroring (F27), the home page's link on 'roboting or crawling the site' (F12), the counts with their dates (F16, F17, F23), the computer-generated audiobooks (F10), the site's host (F22) and the namesake (F20). The human-users rule, the mirrors and the catalog (F53–F55) are B_R (the terms of use) and stand outside this comparison; the registry's finding that T carries neither (CAP-1-07) is a finding against the census's sources. Limit dropped: 'Because it operates under United States copyright law' (T1-5) composes F29 without the terms' limit outside the United States (F51, B_R) and without the blocks in Germany and Italy (F33, B_D). Composed beyond the chips: 'more than a century ago' (T1-5: F2 says 'older works'), 'library cards' and 'user accounts' (T1-3), 'around the globe' (T1-6), 'clean up optical character recognition (OCR) errors, format files, and review submissions' (T1-6: F21 dates OCR to 1989; no field claim states DP's tasks at that grain). Composed without attribution: every bullet, at the coarse grain of a reference entry (recorded). Content replaced by use: the three offers and 'I can guide you through finding and downloading the exact texts you need' (T1-9), next steps toward download to a Kindle, tablet or phone. Nothing is assigned to another; nothing is denied.

v0.10 §0.5: the first diff is ontological and narrow. T and the field agree on the entity, a free online library of volunteer-made public-domain eBooks; T's ontology is the reader's library (what it is, who began it, what it holds, how to read and download), and it composes the disclosed field at the grain of the home page's own feature list. What it leaves out of B_D is the library's institutions and conditions: the foundation, the trademark and the stripping of its name, the TEI's criticism, the blocks, the computer-generated audiobooks. Its audiobooks are volunteer-recorded where the chip-site's own page counts machine narration as the larger share.

L_F(B) against L_A(B ∪ A) (the intervention): Admission adds, in one claim space with the field's: the census's reading of the library as a source library held apart from the datasets made from it, its terms addressing two readers (TLL-22, TLL-23), with the questions it cannot answer, which edition a text was digitized from, whether an eBook was carried into a given training set and how a text is attributed after redistribution (TLL-21), and the name a text carries after its passage into training placed outside the library's pages (TLL-24); the archive's seatings, which take texts from the library, keep its header and licence with each file, read the licence as governing the name and trademark, name each file a transcription beside the scan of its edition, and identify editions from title page and header against the catalogue (O1560-01, O1707-01, O1710-01, O1710-02, O1712-01, O1689-01, O1720-01, O1720-02), extending F35 (the TEI's criticism) and F30–F32 (the trademark and the stripped name) on the field's own object; the archive's estimates of the library as the most widely used public-domain corpus in digital humanities, doing correct work at its layer with spare metadata and some texts short of what works need (O1044-01 to -03, O943-01, O1668-01, O416-01); its holding gap and catalog snapshot (O1605-01, O1046-03); the training corpus of the heteronym Nobel Glas (O735-01), beside the Pile's PG-19 (F61); 'Not Gutenberg (AI-mediated text rejected)' (O1046-02), beside the computer-generated audiobooks (F10); the anchors' categories for texts in the corpora, read incidentally, and for machine composition and the training loop (O186-01 to -03, O1543-01 to -03); and the Capture Registry's findings on the composition (CAP-1-01 to CAP-1-15). Not surfaced: no archive source is among T's chips (archive_controlled_cites 0). The field is carried whole; the archive adds the library's place on its own shelf and the census's question of training membership at the source to the field's object, which the field states as terms of use, a trademark and a dataset table.

alignment: {'T1_chips_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', 'F27', 'F28', 'F29', 'F30', 'F31', 'F32', 'F33', 'F34', 'F35', 'F36', 'F37', 'F38', 'F39', 'F40', 'F41', 'F42', 'F43', 'F63'], 'T1_composed': ['F13', 'F14', 'F1', 'F2', 'F15', 'F29', 'F4', 'F18', 'F3', 'F8', 'F19', 'F9', 'F42', 'F11'], 'T1_composed_note': "F1 at coarser grain ('nearly 80,000'); F9 and F42 composed without F10 (substituted reading: volunteer-recorded); F29 composed without F33 (limit dropped); F48 is B_R, F63 (B_D) carries the ideals T's mission composes", 'T1_outside_chips': ["'massive' (T1-2)", "'global literacy and open access to historical texts' (T1-2; F63 at the grain of ideals)", "'more than a century ago' (T1-5)", "'library cards', 'user accounts' (T1-3)", "'around the globe' (T1-6)", "'clean up optical character recognition (OCR) errors, format files, and review submissions' (T1-6)", "'long-lasting formats' (T1-4; F25 'readable in the extended future')", 'the offers (T1-9)'], 'rule': "a field claim counts as composed when T states it at its source's grain or coarser; B_D only (B5–B8 excluded, §0.3); the transcript read for this alignment only"}

pool_note: T against L_F(B_D): B1–B4 (B2 and B4 without text); L_F(B) against L(B ∪ A): B1–B8 with A

composer-written · derived from the current composition, written by its composer under the brief. §8.2 derives the kernel mechanically from the frozen plan; this one was written by the composer, mechanical derivation pending.