The double monolith

Claim (verbatim)

The cuneiform we hold is not the cuneiform of the textbooks. In the in-house CDLI copy (126,000 rows), a single genre and a single fifty-year period each swallow the corpus: Administrative documents are 89,304 rows (70.9 percent), and the Third Dynasty of Ur (Ur III, about 2112-2004 BCE) is 59,692 rows (47.4 percent). The genres that fill our histories of Mesopotamian thought - Literary (1,907 rows, 1.5 percent), Divination (446), Medical (444), Prayer (290), the whole scholarly-scientific stream - are together a rounding error against the receipts and ledgers. Survival and excavation, not ancient importance, shaped this: bureaucracies produced clay in industrial volume and buried it in findable institutional dumps, while literature lived in fewer copies in fewer places. The catalogue's shape is a preservation artifact wearing the mask of a civilization's output. The mechanism is a double filter - production volume times institutional burial - that privileges administrative Ur III clay over everything the tradition itself prized.

Prediction clause (verbatim)

Prediction (in-house CDLI copy): administrative documents are >= 60 percent of catalogued tablets (in-house: 70.9 percent) AND the combined literary, divination, medical, prayer, and astral-science genres are < 5 percent (in-house: about 3 percent) (primary clause: administrative share >= 60 percent with the famous-genres share < 5 percent); independently, Ur III alone is >= 35 percent of the dated corpus (in-house: 47.4 percent). Disambiguation: shares are computed over rows carrying an assigned genre or period respectively; genre and period are read from CDLI's own fields. Coverage guard: this is explicitly a claim about the CATALOGUED world, not the ancient one - the item's value is to quantify the distortion, and the in-house rate is confirmed against the full cdli.earth API.

Kill-dataset (verbatim)

Kill: the CDLI genre and period fields - genre-composition and period-composition shares. Runs THIS WEEK in-house.

Novelty / leakage triage

already answered in the literature

Pure self-measurement: every clause is stated in the item from the in-house run - administrative share 70.9 percent against a threshold pinned at >= 60, famous-genres share about 3 percent against < 5, Ur III 47.4 percent against >= 35 - so nothing in the primary clause is left that the generator has not already computed, the register admitting as much. The field has also published the double dominance: Streck's corpus quantification and Molina's Ur III census state the administrative and Ur III preponderance the item re-derives.

• In-house CDLI catalogue copy (126,000 rows, about 30 percent of CDLI's 421,501 artifacts, harvested 2026-07-04), the wave's own registered aggregates: Administrative 89,304 rows (70.9 percent), Ur III 59,692 rows (47.4 percent)
• M.P. Streck, 'Grosses Fach Altorientalistik: Der Umfang des keilschriftlichen Textkorpus', MDOG 142 (2010)
• M. Molina, 'The Corpus of Neo-Sumerian Tablets: An Overview', in S.J. Garfinkle and J.C. Johnson (eds.), The Growth of an Early State in Mesopotamia (Madrid, 2008)

Predictions

No prediction registered yet.