AI-generated conjecture · below the evidence/publication boundary

The script's fingerprint

Status: Anticipated · untested

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 proposed connection between two domains, generated by a language model. 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)

Zipf's law — a few signs used constantly, most rarely — is usually taken as a fact about languages. Cuneiform ran a natural experiment no other script can offer: the same sign inventory served Sumerian, then Akkadian, then a heavily logographic late scholarly idiom, across three millennia. The conjecture: the rank-frequency profile of signs is conserved across the language switch — same exponent, same share of text carried by the top signs — because the distribution was fixed by the script's training regime and scribal habit, not by the host language, making the writing system a statistical organism independent of what it encodes. If it holds, the standard assumption that a script's token statistics reflect its underlying language breaks, with direct consequences for how undeciphered scripts' languages are inferred from sign counts.

Prediction clause (verbatim)

Computed over transliterated corpora for Uruk III, Ur III, Old Babylonian, Neo-Assyrian, and Late Babylonian subcorpora, the sign-level Zipf exponent and the text share of the 50 most frequent signs will each vary by no more than 10 percent across the Sumerian-to-Akkadian transition, while the same statistics computed over words shift by more than 25 percent across the same boundary. Primary clause, which decides the verdict: the contrast — sign-level invariance within 10 percent against word-level shift above 25 percent — holds for the Ur III versus Old Babylonian comparison. Secondary clause: the invariance persists into the Neo-Assyrian subcorpus.

Kill-dataset (verbatim)

CDLI (transliterated corpus across all periods) with ORACC period subcorpora for lemmatized word-level comparison.

Nobody has run this test. The kill-data is named above. If you can run it — or you know the paper that already settles it — claim the kill or submit the prior. Kills and priors are credited here, by name, as they come in.

In the atlas

This conjecture is bridged, as an L1 lead, onto these Inferpedia subject pages.

Provenance

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

Composed blind by claude-fable-5 from internal knowledge only, with zero tool calls, and emitted directly as a single JSON text message.

Novelty / leakage triage

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

Searched sign-level rank-frequency statistics of cuneiform. Sign-frequency distributions have been computed for cuneiform-family scripts (Elamite signs fit a broken power law; Indus sign statistics fit Zipf-Mandelbrot), but the invariance-across-host-language test on the Sumerian-to-Akkadian transition was not located.

Predictions

No prediction registered yet.

Weigh in

No community feedback yet.

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.