← All conjectures · Philology, editions & stemmatics

The Catalogue Confesses Alike

Medieval Dated c. 500 CE – 1000 CE East Asia Islamic world

Status: Falsified

The verdict’s fine print — quoted from the resolution record: “KILLED robustly: every defensible reading surfaced by two independent implementations puts BOTH qualifying catalogues below the 15% floor, and the two definitional divergences the spot-check found (Sengyou's que-conflation, corrected DOWNWARD; the Sui shu deno…” 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)

A catalogue compiled downstream of an inheritance must reconcile titles known by report with books in hand, and the shortfall it confesses is set by the generational loss rate of hand-copying — which ran at comparable speed everywhere — not by cultural candor. So three unrelated reconciliations, Sengyou's Chu sanzang ji ji at Jiankang (515 CE) with its registers of scriptures missing from his collection, the Sui shu bibliographic monograph at Chang'an (656 CE) noting titles the Liang court had possessed that were since lost, and Ibn al-Nadim's Fihrist at Baghdad (987 CE) recording books known by title but not seen, should confess loss inside one band: between roughly a sixth and a bit under half of inherited entries, never a few percent, never a majority.

Prediction clause (verbatim)

Prediction: in each of the three catalogues, flagged entries (missing, lost-since-Liang, or not-seen) constitute between 15% and 45% of entries in the surveyed division — every share at least three times a nominal 5% clerical-noise floor, none beyond 45%, all three inside the same 30-point band; coverage guard: only divisions where the cataloguer flags systematically, with at least 200 entries counted per catalogue.

Kill-dataset (verbatim)

Kill: count flagged versus total entries in CBETA T2145 (Sengyou's registers, including the missing-scriptures lists), in the Sui shu jingji zhi on ctext.org (entries annotated as held under the Liang and now lost), and in the Flügel edition of the Fihrist (1871–72, archive.org; not-seen notices); any qualifying catalogue outside 15–45% kills.

On Inferpedia

This conjecture is linked to the following pages on Inferpedia, an encyclopedia of the missing — working atlas pages, some still early scaffolding.

Gestalt

Provenance

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

Fresh blind generation by claude-fable-5, 2026-07-20, for the Old-World-invariance (same-laws-everywhere) wave, posed entirely from weights with no repository or web consultation. I favored paired law-and-named-break shapes — the copying funnel against calendar-driven liturgy, oral king-list rounding against Egyptian annalistic precision, open-library age-curves against the sealed Dunhuang deposit — plus catalogue grammars of loss and doubt, cross-canon vocabulary growth, commentary expansion ratios, translator concentration, pharmacopoeia accretion tempo, primate archives, hapax poets, and the day-wage price of a book; ordinals 2, 4, 8, 13, and 16 are the divergence wagers.

Novelty / leakage triage

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

Catalogue-confessed loss is established scholarship on each side separately: Dudbridge 2000 builds precisely on the Sui shu monograph's 'the Liang had it, now lost' annotations to quantify medieval Chinese textual loss; Sengyou's registers of scriptures missing from his collection are treated in the history of Chinese Buddhist bibliography (Storch 2014); and the Fihrist's known-by-title-but-not-seen notices are a commonplace of Arabic book history (Dodge 1970). What is not located is any cross-tradition census placing the three confession rates inside one 15-45% band - Dudbridge quantifies the Chinese side only, and I know of no comparable tabulation of the Fihrist's not-seen notices. The factual frame checks out: Chu sanzang ji ji at Jiankang c. 515 and T2145 correct, the Sui shu monographs completed 656, the Fihrist 987-88, the Fluegel edition 1871-72. Verdict: adjacent - the pinned three-way band census is new.

Sources cited by the triage

Glen Dudbridge, Lost Books of Medieval China (The British Library, 2000) — Quantifies loss via the Sui shu bibliographic monograph's annotations - the closest located prior on one leg.
Tanya Storch, The History of Chinese Buddhist Bibliography: Censorship and Transformation of the Tripitaka (Cambria Press, 2014) — Sengyou's catalogue and its registers of missing scriptures.
Bayard Dodge (trans.), The Fihrist of al-Nadim: A Tenth-Century Survey of Muslim Culture (Columbia University Press, 1970) — Standard English apparatus for the not-seen notices.

Predictions

Killed registered 2026-07-20 calibration prediction (parent triage: leaked/adjacent)

Resolution: Killed

Caveats: KILLED robustly: every defensible reading surfaced by two independent implementations puts BOTH qualifying catalogues below the 15% floor, and the two definitional divergences the spot-check found (Sengyou's que-conflation, corrected DOWNWARD; the Sui shu denominator granularity) both move shares away from the band, never toward it - the kill never depends on a contested definition. SPOT-CHECK LAYER FINDING, recorded: the blind compute's QC claimed to exclude catalogue-omission que marks but excluded only one of ~8-16 such cases; the independent re-implementation caught it; verdict direction unaffected (the corrected numbers kill harder), but the incident is the spot-check layer visibly earning its place. WHAT THE KILL MEANS: the confession MECHANISM is real and the direction right - entry-level loss-flagging runs at ~10-14% in both catalogues, an order of magnitude above the 5% clerical-noise floor the claim itself posited - but the wagered 15% band floor over-pinned it; the familiar calibration shape, now on an invariance-band floor. Disclosed-no-weight, threshold-shopping made visible and inert by pre-registration: looser post-hoc definitions WOULD reach the band (Sengyou's juan-4 embedded missing-scriptures list reads 35.5%; the Sui shu you-you...wang variant reads 16.20%) - the ex-ante pinned division and formula govern, and no v2 on these corpora (outcomes known). FIHRIST, two lessons: (i) LESSON-4 REFINEMENT MINTED - the registration verified the Fluegel edition EXISTS on archive.org but not that its OCR layer is READABLE (zero Arabic in the text layer); OCR-dependent instrument pins must include a text-layer usability probe at registration time - reachable is not readable. (ii) The substitute-edition color is independently valuable and non-verdict-bearing: al-Nadim's catalogue grammar contains essentially NO systematic sight-confession (4/7001; the lam-aqif formant absent entirely from both editions checked) - the claim's premise fails at the Baghdad end by ABSENCE of the convention, a deeper break than a share miss, and direct grist for sibling registration-candidate B-006 (catalogue grammar as institutional, not universal). Blind compute claude-sonnet-5; spot-check claude-opus-4-8; verdict claude-fable-5.

OPEN prediction (invariance band): three unrelated catalogue reconciliations - Sengyou's Chu sanzang ji ji (515 CE), the Sui shu bibliographic monograph (656 CE), Ibn al-Nadim's Fihrist (987 CE) - each confess loss on 15-45% of inherited entries in their systematically-flagged divisions: one confession band from Jiankang to Baghdad.

Resolution criteria — the registered fine print

Resolution criteria: VERBATIM PREDICTION: 'in each of the three catalogues, flagged entries (missing, lost-since-Liang, or not-seen) constitute between 15% and 45% of entries in the surveyed division - every share at least three times a nominal 5% clerical-noise floor, none beyond 45%, all three inside the same 30-point band; coverage guard: only divisions where the cataloguer flags systematically, with at least 200 entries counted per catalogue.' OPERATIONALIZATION (pinned ex ante): division selection is logged BEFORE any counting, justified from each catalogue's own flagging practice; counts second. (1) Sengyou, Chu sanzang ji ji = CBETA T2145 (record CBETA release version): flagged = entries carrying que/queben (missing-text) markers or standing in Sengyou's explicitly missing-scriptures registers; denominator = total entries of the selected division(s). (2) Sui shu jingji zhi via ctext.org (record fetch date): flagged = entries carrying the 'Liang you ... wang' formula (held under the Liang, since lost); denominator = total entries of the division(s) where the formula is applied systematically. (3) Fihrist, Fluegel edition 1871-72 (archive.org; record the page range used): flagged = entries carrying al-Nadim's not-seen notices (lam arahu / lam aqif alayhi and equivalent sight-formulas as printed); denominator = entries of the division(s) where sight-notices are systematic. The 'three times a 5% floor' and '30-point band' phrases restate the [15%,45%] bounds; there are no separate clauses. VERDICTS: any qualifying catalogue outside [15%,45%] -> KILLED; all three qualifying and inside -> SUPPORTED; fewer than three catalogues qualifying under the guards -> INCONCLUSIVE-BY-COVERAGE.

Known-priors disclosure — what the registrant already knew

Known priors disclosure: Triaged adjacent 2026-07-20: per-catalogue loss-confession conventions (que markers, the Liang-you...wang formula, al-Nadim's sight notices) are established philology and are the claim's premise; the cross-tradition band computed on a common definition is unlocated. The registrant knows the conventions exist and has computed no shares. Neighbors disclosed: breadth-india-003 (Kaiyuan graveyard wing, SUPPORTED 2026-07-20) concerns a different catalogue (T2154) and a different quantity (era-composition ratio of the missing-books register); its numbers are not carried over. Sibling: gestalt-invariance-006 pins T2154's doubt divisions - kept disjoint from this registration. Risk disclosed: division selection is the operationalization's soft joint; hence the log-selection-first rule.

Method and dataset — how it was measured

Register-before-compute held (registration 15:05:52Z, packet 9c516ba). Division selection was logged before counting per the registered rule. Compute results: Sengyou 67/489 = 13.70%; Sui shu 339/3086 = 10.99% under the pinned main-entry unit and liang-you...wang formula; Fihrist limb coverage-failure. Independent spot-check (full re-implementation, different source route, prior output unread): Sengyou entries 489 EXACT; Sengyou flags corrected to 52-57 (= 10.63-11.66%) - the compute had counted ~10-15 catalogue-omission que marks ('an earlier catalogue lacked this title', text demonstrably extant, e.g. the Sishierzhang jing whose own summary says it currently circulates) as now-missing flags; Sui shu flags 335-341 bracketing 339 (strict-title variant 307); Sui shu entries 3206 vs 3086 (juan-33 mega-cell expansion; the treatise's own subtotals are internally inconsistent, so both are defensible). DECISION SWEEP over every defensible reading: Sengyou 10.63-13.70%, Sui shu 9.58-11.05% - all outside [15%,45%], all below the floor. Both catalogues qualify (>= 200 entries, systematic flagging). REGISTERED KILL CLAUSE: 'any qualifying catalogue outside [15%,45%] -> KILLED.' KILLED, on both qualifying catalogues independently.

Dataset: Three catalogues per the registration. (1) Sengyou, Chu sanzang ji ji: CBETA T2145 release 2026.R1 (compute) and Chinese Wikisource's T2145 text (spot-check, textual equivalence corroborated by the exact 489-entry match); division selected ex ante = juan 2's two primary registers. (2) Sui shu jingji zhi via ctext.org, all four bibliographic juan (32-35), 2s-throttled. (3) Fihrist: the PINNED Fluegel 1871-72 archive.org edition proved unusable - its OCR text layer contains ZERO genuine Arabic characters across both volumes (sampled at 6 positions spanning volume 1 plus volume 2's opening, evidence recorded); substitute-edition evidence (OpenITI transcriptions of Tajaddud, cross-validated against a Fluegel-paginated transcription) is reported as non-verdict-bearing color only.

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.