← All conjectures · South Asian text cultures

The catalogue keeps a graveyard wing

Medieval Dated c. 100 CE – 800 CE South Asia

Status: Supported

The verdict’s fine print — quoted from the resolution record: “SUPPORTED, robust across the entire defensible reading band (the sefaria-004 standard): both finite readings give R = 8.35-10.99 against the registered >= 3, and the conservative never-split readings yield a Sui-Tang missing share of ZERO -- undefined R only b…” Read the full caveats ↓

Status is derived only from the shepherd-authored triage/prediction data above -- community submissions and claims are a separate overlay and can never change it (see the participation panel below).

This is a conjecture imagined by a language model — drawn from its trained weights and held to falsifiability, novelty, and plausibility, not to any one method: it may join two or more fields, or none. It is not an article and not evidence: it sits below the evidence/publication boundary. A quantitative prediction and a named kill-dataset are attached (when registered) so the claim stays falsifiable rather than merely evocative.

Claim (verbatim)

India kept no register of its own textual losses; China kept one for it. Zhisheng's Kaiyuan Shijiao lu (730 CE; Taishō 2154), the culmination of the catalogue sequence begun by Sengyou's Chu sanzang ji ji (515 CE), includes a queben ("missing books") register: Chinese translations of Indian works that earlier cataloguers had recorded but that could no longer be produced by 730 — a dated loss-ledger for the Indian corpus's carrier fleet. Custody explains the decay schedule: translations made before state canon-formation (Han through Jin) circulated privately, without institutional custody, and should be missing at multiples of the later rate, with translator-attribution itself decaying (the "translator lost" class concentrating in the earliest strata). The same ledger records the pipeline's narrowness at the source: of the 657 texts Xuanzang famously carried home in 645, his imperial bureau translated 75 before the master died — the rest are catalogue smoke. Prediction: tabulating the Kaiyuan Shijiao lu's own entries by dynastic period of translation, the missing (queben) share for pre-420 translations will be at least three times the share for Sui-Tang translations, and at least 80% of anonymous-translator entries will fall before the Sui (primary clause: the threefold queben ratio; the verdict follows it). Exact computation: parse Taishō 2154 from the open CBETA XML corpus, classify entries by period, translator status, and extant/missing status, and count. Kill: the CBETA XML-P5 open corpus (github.com/cbeta-org/xml-p5) texts of the Kaiyuan Shijiao lu (T. 2154) and Sengyou's Chu sanzang ji ji (T. 2145).

Prediction clause (verbatim)

Prediction: tabulating the Kaiyuan Shijiao lu's own entries by dynastic period of translation, the missing (queben) share for pre-420 translations will be at least three times the share for Sui-Tang translations, and at least 80% of anonymous-translator entries will fall before the Sui (primary clause: the threefold queben ratio; the verdict follows it). Exact computation: parse Taishō 2154 from the open CBETA XML corpus, classify entries by period, translator status, and extant/missing status, and count.

Kill-dataset (verbatim)

Kill: the CBETA XML-P5 open corpus (github.com/cbeta-org/xml-p5) texts of the Kaiyuan Shijiao lu (T. 2154) and Sengyou's Chu sanzang ji ji (T. 2145).

Provenance

Run: Fresh agent generation · model: claude-fable-5

Fresh blind generation by claude-fable-5, 2026-07-16, breadth wave weighting India/South Asia by inferred textual production rather than survival; every item grounded in real works, authors, catalogues, and testimonia, including the real evidence of loss (citing authors, catalogue entries, translation corpora, rediscovery cases); no fabricated citations.

Novelty / leakage triage

anticipated in the literature — this exact test has never been run

The queben (missing books) register and Zhisheng's extant/lost notations are well studied in the bibliography literature (Storch), and the concentration of anonymous 'translator lost' entries in the earliest strata is a recognized feature of the catalogues (Nattier on the earliest translations). A period-wise tabulation of missing shares — the threefold pre-420 ratio computed over T. 2154 — was not located.

Sources cited by the triage

T. Storch, The History of Chinese Buddhist Bibliography: Censorship and Transformation of the Tripitaka (Cambria Press, 2014)
J. Nattier, A Guide to the Earliest Chinese Buddhist Translations (IRIAB, Soka University, 2008)

Predictions

Supported registered 2026-07-17 calibration prediction (parent triage: leaked/adjacent)

Resolution: Supported

Caveats: SUPPORTED, robust across the entire defensible reading band (the sefaria-004 standard): both finite readings give R = 8.35-10.99 against the registered >= 3, and the conservative never-split readings yield a Sui-Tang missing share of ZERO -- undefined R only because the denominator vanishes, i.e. maximal support, not ambiguity. The graveyard wing is real: pre-420 translations carry a 10-36% missing rate in the catalogue's own bookkeeping while Sui-Tang carries 0-3.3%. CALIBRATION NOTE: the first MULTIPLICATIVE pin in the track record to survive, and with room to spare (wagered 3x, world gave >=8.4x) -- the dead class died by overshooting reality; here the wager sat well inside it. Registration-vintage note: pre-dates lessons 3+ (the pre-420 gloss leaves the Sixteen-Kingdoms rump ambiguous), disclosed and immaterial -- both resolutions clear the bar. DISCLOSED-NO-WEIGHT: 60-77% of entries' queben status is structurally unresolvable (batch-level summary prose); extraction caution cannot rescue the conjecture's negation, since the strictest reading is the MOST supportive. Secondary (anonymous-translator share 1.000 pre-Sui) reported non-binding at low reliability per the compute's own coverage-imbalance flag.

Registered against the open CBETA XML-P5 text of Zhisheng's Kaiyuan Shijiao lu (Taisho 2154). Claim under test (primary clause): among the catalogue's own entries, the MISSING (queben) share of pre-420-CE translations is AT LEAST THREE TIMES the missing share of Sui-Tang translations.

Resolution criteria — the registered fine print

Resolution criteria: DATA: parse T.2154 from CBETA XML-P5 (github.com/cbeta-org/xml-p5); freeze a sha256. POPULATION: catalogue entries, each classified by (a) DYNASTIC PERIOD of translation, (b) EXTANT vs MISSING (queben 闕本 / 闕 register), (c) named vs anonymous translator. PRE-420 = periods before the Liu-Song (Han through Western/Eastern Jin up to 420 CE); SUI-TANG = Sui + Tang. QUEBEN_SHARE(period) = entries-marked-missing / total-entries in that period. PRIMARY R = queben_share(pre-420) / queben_share(Sui-Tang). CLAUSE PRECEDENCE: (1) INCONCLUSIVE_BY_DESIGN if T.2154 cannot be parsed, OR the missing/period/translator classification cannot be reliably extracted from the catalogue's own structure, OR fewer than 30 classifiable entries fall in either period. (2) SUPPORTED if R >= 3. (3) KILLED if R < 3. Report both shares, entry counts, R, the extraction method, and the SECONDARY (non-binding): the share of anonymous-translator entries falling before the Sui (conjecture predicts >=80%). If the queben-status extraction admits more than one reasonable reading, report the R range across them (Tier-1 operationalization-stability lesson). computed_at postdates registered_at (rule 6).

Known-priors disclosure — what the registrant already knew

Known priors disclosure: Held: the conjecture's reasoning (China kept a register of Indian Buddhist textual losses that India did not) and general knowledge that T.2154 is in CBETA with a queben register and per-entry translator/period data. NOT computed: the actual period-by-period missing shares or R. Threefold threshold is the conjecture's own.

Method and dataset — how it was measured

Firewalled blind compute (claude-sonnet-5, launched by the resumed Opus loop, blind to threshold and direction) parsed the catalogue's own structure only: dynasty headings for period, closing summary lines for queben status under TWO pre-declared operationalizations (self-consistent-split vs never-split-mixed), reported as a range per instruction. Verdict authored by claude-fable-5 per the registered clause precedence: guard passed; SUPPORTED iff R >= 3.

Dataset: CBETA XML-P5 Taisho 2154 (Zhisheng's Kaiyuan Shijiao lu), raw fetch sha256 1d910dd1ac1938711677ca815bd824f923ad496159fb60e7b2c5e8329919894c, juan 1-9 General Catalogue: 2,836 entries, period classification 100% structural. Registered guard (>=30 classifiable entries per period) cleared under EVERY reading: pre-420 462 (split) / 270 (conservative), Sui-Tang 245 / 184.

computed 2026-07-20

Weigh in

No community feedback yet.

New here? Create an account first

Create an account or sign in and your feedback is tied to you — you can track it, get replies, and claim this conjecture so others know you’re working on it. Prefer not to? Just leave your take below as a guest — only the name you type is shown.

Add your take

Posted immediately (spam is removed). Community feedback is never an adjudicated verdict and never changes this conjecture's triage label or status above.