AI-generated conjecture · below the evidence/publication boundary
← All conjectures · Method, bridges & the corpus itself
The Instrument Reinvents Where It Reads Most
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).
A conjecture about this corpus itself
This is one of five meta-conjectures the generating model posed, blind, in the corpus’s final wave — predictions not about the pre-print world but about this collection of 1001 conjectures: how its verdicts would distribute, where its novelty would concentrate, whether its own governing architecture would survive its own data. Its kill-dataset is the corpus’s adjudication records, which is why it sits outside the ordinary triage ladder (“out of scope” below): it could only be judged once everything else had been.
Those records now exist, and on 11 July 2026 the model formally resolved this conjecture against them. Verdict: killed
Mapping the architecture's quadrant assignments to an ordinal operationalization score (Q1=4, Q3=3, Q2=2, Q4=1, straddles averaged) across the 17 quadrant-assigned waves W07-W22+W25, Spearman rho between operationalization rank and leaked rate = +0.38 (t ~ 1.59, df 15, one-tailed p ~ 0.066) — below the pre-registered rho>=0.6 and not significant at 0.05. Secondary clause fails decisively: the two named densest strata rank 12th (W13 scholasticism, 5.7%) and 14th (W11 Byzantine, 2.9%) of 17 in reinvention rate; the actual top two are W20 exact sciences (16.7%) and W07 medieval literature (14%).
The test is genuinely runnable — leaked is the rubric-stable 'already established' measurement and the pre-registration supplies an operationalization ordering — and it fails on both clauses. Operationalization choices flagged: wave=stratum; quadrant-to-ordinal mapping is mine (no numeric index was pre-registered, only the quadrant placements); cross-cut waves W23/W24 excluded for lacking quadrant assignments — including them as low-operationalization points raises rho to only ~0.45, so the kill is robust to that choice. The substantive finding is sharper than a mere miss: reinvention tracks the DIGITIZATION axis, not the question-operationalization axis — the corpus's two pre-registered axes dissociate in its own failure data, with the most question-saturated strata (scholasticism, Byzantine transmission) among the LEAST reinvented and the digitized-quantitative ones (exact sciences, cuneiform at 13.3%) the most. Caveat: wave-level leaked counts are small (0-7 per wave), making individual ranks noisy — but noise cannot rescue a point estimate of 0.38 against a 0.6 bar plus a secondary clause whose two named strata land in the bottom third. Note the direction is at least positive; the calibration-curve idea is weakened, not reversed.
Claim (verbatim)
A generative model posing research conjectures is an instrument, and an instrument's systematic error should be characterizable. I conjecture that this corpus's dominant systematic error — reinvention, the confident restatement of results already standard in a field — rises monotonically with the field's degree of question-operationalization, so that ranking the strata by how thoroughly scholars have already quantified them predicts, with high rank correlation, the rate at which adjudicators mark this corpus's conjectures "already known." The mechanism is that a model cannot easily distinguish what it knows because the answer is established from what it infers freshly; the denser the secondary literature in training, the more established answers masquerade as conjectures. If this holds, reinvention rate becomes a usable calibration curve for the instrument: given any new target field, its operationalization rank predicts in advance how much of a model-generated conjecture set will be recycled knowledge — which is precisely the correction factor future waves of machine-generated science will need.
Prediction clause (verbatim)
Ranking the corpus's strata by a pre-specified operationalization index and computing each stratum's "already established" verdict rate from adjudication records, the Spearman rank correlation between operationalization rank and reinvention rate will be at least 0.6. Primary clause (the verdict follows it): Spearman rho >= 0.6, significant at p < 0.05. Secondary clause: the two densest strata (Latin scholasticism and Byzantine transmission) occupy the top two positions in reinvention rate.
Kill-dataset (verbatim)
The One Thousand and One Conjectures corpus and its triage/adjudication records; kill is a statistical test (Spearman rank correlation of reinvention rate against operationalization rank).
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 with zero tool use, emitted as a single JSON text message per the fresh-lane blindness protocol.
Novelty / leakage triage
out of scope (a meta-conjecture about the corpus itself)
Meta-conjecture characterizing the generating model's systematic error using this corpus's adjudication records as the kill-dataset; out of shepherd scope per the W26 rule.
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.
Working on this?
Sign in to claim this conjecture and let others know you're working on it.