Generated by Fable · below the evidence/publication boundary
308 of 1001 posed · 158 shepherd-triaged · 150 provisional · 0 frontier · 20 predictions · 9 resolved (6 supported / 3 killed) — the 1001st will be posed at Ars Inquirendi, Oxford, 20 November 2026.
Cross-domain conjectures generated noetically by Fable — a frontier AI proposing, from its own knowledge, surprising connections between two well-known domains that it judged likely to be both novel and important. Each pairs a specific claim with a quantitative prediction and a dataset that could prove it wrong; each was then checked against the literature to flag the ones with known priors.
This is one form of lead generation for Inferpedia, the encyclopedia of the missing — and this page is an early preview.
Why these conjectures matter: charting the noetome — its structure, limits & potential →
Nothing here is claimed as verified-novel. Each sits below the evidence/publication boundary: a connection already known in the literature is shown honestly and tagged Prior, and every prediction is registered before it is scored. Spotted a prior yourself? Open any conjecture and weigh in.
Filter
Clear all filtersShowing 1–32 of 32 matching conjectures.
Joins measurement theory (the error bar as a signal in its own right) to third-century political history: the width of the date range a modern editor can assign to a document is a thermometer of the dating formulas, regnal clarity, and bureaucratic…
Greenland lead deposition from Roman smelting is a proxy for silver output; predict lead flux co-moves with denarius fineness and leads Egyptian papyri price inflation by roughly a decade. Falsify: ice-core lead vs papyrological price series.
Hoard deposition is antiquity's VIX: deposition rates should lead invasions in narrative sources by months (people bury on rumor), and non-recovery rates should scale with conflict mortality. Falsify: dated hoard databases vs chronicle chronology.
Coin wear per year (calibrated on dated issues) measures circulation velocity; predict velocity spikes in crises as hoards dishoard. Falsify: 3D wear metrology on stratified site finds.
Provenance distance-decay exponents should scale with value density: Mayen lava millstones should decay steeper than fine pottery by precisely the ratio of transport cost to unit value. Falsify: sourced distribution maps.
Computational errors in dated horoscopes are inherited from specific copies of astronomical tables — so recomputing surviving horoscopes reconstructs the copy-tree of the tables themselves. Falsify: recompute and cluster.
The Aṣṭādhyāyī's rule ordering sits near the minimum-description-length optimum over permutations of its own rules. Falsify: computational reordering search.
Generalizing the famous 190 Themistocles sherds cut by a few hands: handwriting clusters on ostraka should predict deposit location — ward-level bloc voting, measurable. Falsify: paleographic clustering plus findspots.
Pompeian wall dialogues should follow the same reply-depth distribution as modern forum threads — conversational universals across two millennia. Falsify: CIL IV thread reconstruction.
His distance errors should grow log-normally with informant hops from the Aegean — rumor as multiplicative noise. Falsify: reported vs true distances.
The map's segment errors should cluster by province, revealing the distinct source itineraries it was compiled from — stemmatics of a map. Falsify: error-variance clustering.
The Shannon entropy of shipwreck cargo composition falls a generation before political fragmentation — markets disintegrate first. Falsify: Parker/Strauss wreck databases.
Iron Age fort intervisibility graphs are small-world, and abandonment sequences preserve connectivity — graceful degradation. Falsify: DEM viewsheds plus dating.
Route alignments fit Wardrop user-equilibrium (individually shortest time), not system optimum; imperial upgrades target exactly the congestion externalities. Falsify: least-cost GIS vs surveyed alignments.
Gradient error accumulates as √distance between inspection shafts — surveying as a bounded random walk with resets. Falsify: laser-level surveys of Nîmes and Segovia channels.
Dimensional standardization (CV of amphorae and lamps) falls as a power law of cumulative regional output — the learning curve, two millennia early. Falsify: typology metrics vs production estimates.
The recycling fraction in Roman glass (antimony–manganese mixing) rises with political crisis and distance from Levantine furnaces — a business-cycle indicator in silica. Falsify: compositional time series through the 3rd century.
Roman colonial grid azimuths match sunrise on foundation-festival dates, so grids encode founding dates we've lost. Falsify: azimuths vs the colonies whose dates we know.
Portable Roman dials cut for the wrong latitude reveal central workshops exporting blind; misfit errors should cluster by workshop signature. Falsify: latitudinal misfit clustering in the dial corpus.
Foggara abandonment should propagate radially outward from the depleting fossil aquifer's center in the exact sequence a drawdown model predicts. Falsify: settlement dates vs hydrogeological modeling.
Vomitoria design achieves evacuation times meeting modern stadium standards — safety engineering by iteration. Falsify: agent-based egress simulation on the actual plans.
Greek and Roman astragalus throw scores are inversely proportional to empirically measured face probabilities — implicit odds-pricing two millennia before probability theory. Falsify: toss experiments vs attested scoring.
Roman die asymmetry clusters by workshop tradition, not period — craft transmission of unfairness. Falsify: 3D-scan corpora.
Early received hexagram records should match the asymmetric yarrow-stalk probabilities and later ones the symmetric coin method — a statistical date for the method change. Falsify: frequency analysis of Zuozhuan vs later corpora.
Recorded oracle ambiguity (countable branch interpretations) scales with the stakes and uncertainty of the question — hedged forecasting as institutional survival. Falsify: coding the Fontenrose catalog.
Joins actuarial insurance pricing to maritime archaeology: fourth-century BCE Athenian bottomry rates were risk-priced to break even, so the loss probability implied by loan spreads must match the wreck rate implied by the archaeological record, two completely independent instruments reading the same…
Joins eigenvector centrality to Roman jurisprudence: the Law of Citations of 426 CE, which privileged five jurists, mostly legislated what network centrality would compute from the jurists' own cross-citations, with Gaius as the deliberate, measurable exception smuggled in from the classroom.
Joins modern campaign geography (yard-sign and canvassing decay around a candidate's home) to Pompeian electoral epigraphy: painted programmata are spatial campaign artifacts, so endorsements should decay exponentially with walking distance from the candidate's own house, just as lawn signs do around a…
Friedman's plucking model of business cycles meets documentary papyrology: aggregate documentary output runs along an administrative-capacity ceiling from which crises pluck it downward, so the depth of a collapse predicts the speed of the subsequent recovery, while the size of a boom…
Lanchester's attrition laws meet ancient naval history: ramming warfare concentrates force like aimed fire, so galley battles obey the square law, while phalanx battles, fought rank against rank on a fixed frontage, obey the linear law — two regimes of the same…
Price-ceiling economics meets Roman epigraphy: Diocletian's Price Edict of 301 CE worked exactly as shortage theory predicts — goods capped below prevailing market prices withdrew from recorded exchange, while goods capped at or above market stayed visible, so the Edict's failure is…
Service-capacity planning meets Greek civic demography: theaters were sized to a fixed service fraction of the citizen body — a percentile capacity rule — making seat counts a fossilized census tighter than any other urban size proxy.