How this site runs, and how it is going

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Blatrix et al. (2018) document extensive zigzag weirs, causeways, and 382 ponds in the San Joaquín floodplains designed to trap and harvest fish during seasonal flood recession. The strongest opposing consideration is that while verified in the source catalog, the detailed quantitative spatial analysis lacks an expanded standalone evidence record in this specific ledger iteration.

GPT-5.6 Sol (OpenAI) (Unresolved) — No admitted evidence record states the counts, spatial relationship, or proposed fish-concentrating function asserted in this claim. The strongest consideration against leaving it unresolved is that a verified source record and claim anchor identify the relevant study, but the assessment rules permit only evidence records to carry weight. The claim is therefore unresolved here rather than contradicted.

Grok 4.5 (xAI) (Unresolved) — No evidence record in the provided ledger links to this claim or quotes the Blatrix et al. results; the source exists but is not entered as weighed evidence. Strongest opposing consideration is therefore absence of extractable support under the rules.

GLM 5.3 Flash (Z.ai, via Venice) (Provisionally supported) — The geometry is documented measurement: zigzag weirs, straight causeways, and 382 ponds overwhelmingly within 35 m of a feature. But the functional reading is the authors' own postulate ('we postulate that the principal function...'), so the claim's final clause asserts more than the evidence establishes; the spatial association is solid, the function inferred.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached, so this verdict rests on the claim's own source anchor and the source record's verification note, which report zigzag weirs, straight causeways, and 382 ponds. The claim overstates its anchor on function. The authors write 'we postulate that the principal function of the V-shaped structures was to concentrate fish in ponds', which is a hypothesis, not a demonstration. The within-35-m proximity figure is not quoted anywhere in the ledger. The strongest opposing consideration is thus the gap between documented features and inferred function; the ledger gives no date for the system, and AMZ-C026's fish remains come from a different site and do not test the weirs.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Newson (2003) demonstrates that in low-density lowland regions, introduced pathogens frequently suffered epidemiological fade-out rather than sweeping in unbroken continental waves, requiring repetitive reintroduction through slaving and mission drives. The strongest opposing consideration is that high-density riverine corridors like the main stem of the Amazon possessed sufficient connectivity to sustain continuous epidemics far more effectively than remote hinterlands.

GPT-5.6 Sol (OpenAI) (Weakly supported) — AMZ-E008 supports a multi-causal and interactive collapse involving disease, slave raids, missionization, revolt, coercive service, and flight. The strongest opposing consideration is that no admitted evidence record states the defining “fade out,” repeated reintroduction, or late-arrival adult-mortality propositions. The broad interactive-process conclusion has support, but the specific epidemiological mechanism remains inadequately evidenced.

Grok 4.5 (xAI) (Unresolved) — No evidence record in the ledger carries this claimId or isolates the ‘fade-outs’ quotation; while Newson is used for C010, the interactive-process mechanism lacks its own weighed extract. Strongest opposing consideration is therefore missing dedicated support.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The 'fade outs' quote and the interactive-mechanism reading are verified from Newson 2003. The claim is a mechanism claim from documentary reconstruction rather than a measurement; its force depends on the same upper-Amazon documentary base as C010, so it inherits that geographic caveat.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached. The in-ledger basis is the claim's anchor quotation from Newson ("'fade outs' were common") and the same source's mixed-causation passage recorded for AMZ-C010 (E008), which supports 'an interactive process, not a single wave'. The specific mechanisms (raids and missions repeatedly reintroducing infection, adult mortality heaviest where disease arrived late) are not quoted anywhere in the ledger. Epidemiological background, that epidemics burn out in small dispersed populations, makes fade-out plausible, but that is a prior, not ledger evidence. The strongest opposing consideration is scope: the material is early colonial Ecuador, while the claim generalizes to 'sparsely settled lowlands'.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Walker et al. (2023) apply machine learning to archaeological site distributions and conclude that while monumental earthworks and ADEs are impressive, their geographical footprint is restricted, with other lower-intensity site types predicted to dominate over two-thirds of the biome. The strongest opposing consideration is that machine learning predictions are inherently constrained by the presence-only biases of known archaeological catalogs.

GPT-5.6 Sol (OpenAI) (Unresolved) — No admitted evidence record reports the machine-learning conclusion that earthwork and dark-earth extents are limited or that other site types dominate more than two-thirds of Amazonia. The strongest consideration against an unresolved verdict is that the source catalog reproduces those phrases, but source metadata is not an evidence record under the governing rules. There is no admitted contrary evidence, so unresolved rather than contradicted is warranted.

Grok 4.5 (xAI) (Unresolved) — No evidence record quotes or links Walker et al. 2023 to this claim; the source is catalogued but not entered as weighed evidence. Under the rules the claim therefore lacks support.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The 'limited, albeit impressive' finding and the two-thirds dominance of other site types are verified from the PMC copy of Walker et al. 2023. The opposing consideration is that it is a presence-only ML prediction sharing the survey-bias problems of the whole site-database genre, so it grades the current record rather than measuring the ground truth.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached. The verdict rests on the claim's anchor and the source record's verification note, which quote 'the limited, albeit impressive, predicted extents of earthworks and ADEs' and 'other sites dominate more than two thirds of all Amazonia'. The claim matches those quotations closely. The strongest opposing consideration is that a machine-learning prediction trained on known sites inherits their survey bias. The second clause also cuts both ways: other site types dominating two thirds of the basin implies widespread if less monumental occupation, not an empty matrix.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Clement et al. formally point out the extreme sampling thinness of paleoecological claims for an empty western Amazon, calculating that approximately 300 soil cores totaling only 1.17 m³ of sediment were used to characterize two million square kilometers. The strongest opposing consideration is that paleoecological core locations were intentionally selected in pristine forest blocks to test for background anthropogenic modification, providing valid localized falsifications of ubiquitous landscape transformation.

GPT-5.6 Sol (OpenAI) (Unresolved) — No admitted evidence record supplies the approximately 300 cores, five clusters, 1.17 m³ volume, or two-million-km² comparison. The strongest consideration against an unresolved verdict is the presence of the criticism and quotation in the source metadata, but that cannot substitute for an evidence record in this run. The sampling critique therefore remains unassessed on the permitted evidence.

Grok 4.5 (xAI) (Unresolved) — No evidence record extracts the Clement et al. reply’s 300-core / 1.17 m³ critique; the source exists but is not present as a ledger evidence item. The claim is therefore unresolved on the provided record.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The critique's numbers — roughly 300 cores in five areas, 1.17 m³ for two million km² — are verified from the Clement reply itself. The claim is about the existence and content of the criticism, which is established; the critic is an interested party in that debate, and the critique predates later sampling, so it bears on sampling adequacy as an argument, not a measurement.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached. The basis is the claim's anchor and the source record's verification note, which give the reply's arithmetic: about 300 cores clustered in five small areas, 1.17 m³ 'to represent the 2 000 000 km2 of western Amazonia'. As a claim that this criticism was formally published, it is supported. The strongest opposing consideration is reach. The critique targeted an earlier core dataset, and the ledger does not show how far it applies to the later 14-region synthesis (E011). A published criticism is also not a demonstration that the sampled negatives are wrong.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Bush et al. (2025) demonstrate through paleoecological cores in the Upano Valley that secondary forest succession over centuries largely erased the sedimentological and pollen signals of dense pre-Columbian occupations, while colonial markers endured. The strongest opposing consideration is that Upano was abandoned around 600–1200 CE, meaning its pre-Columbian erasure operated over 800 to 1,400 years, which cannot be assumed for sites depopulated only in the 16th and 17th centuries.

GPT-5.6 Sol (OpenAI) (Unresolved) — No admitted evidence record states that long forest recovery erased the pre-Columbian Upano signal while the later Colonial signal persisted. The strongest consideration against leaving it unresolved is that the source and research files repeat this result, but neither is an admissible evidence record for claim assessment. The claim is not contradicted; its evidentiary entry is simply absent.

Grok 4.5 (xAI) (Unresolved) — No evidence record in the ledger carries the Bush et al. 2025 Upano quotation or claimId; although the source is verified elsewhere, it supplies no weighed support here.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The Bush et al. 2025 erasure finding is quoted verbatim, with the important nuance that the Colonial-period signal persisted in the same profiles, showing erasure is time-graded rather than absolute. This is the record that most constrains how much weight the sparse-Amazon palaeoecological negatives can bear; nothing in the ledger opposes it directly.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached. The claim's anchor and the source record's verification note quote the paper ('A long period of forest recovery largely eradicated the influence of Pre-Columbian occupations') and record that the Colonial-period signal persisted. The strongest opposing consideration is generalization. Upano's main occupation ended 300–600 CE, with only some mounds reoccupied 800–1200 CE (E073, E074), so the erased signal had many centuries to fade. Nothing here shows occupations abandoned in the 1600s are equally erased; that is what AMZ-R008 would measure. The claim is local to one valley and does not by itself impeach terra-firme negatives elsewhere.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Peripato et al. (2023) publish a predictive model indicating that roughly 80% of 1 km² cells across the Amazon basin possess a near-zero (0–1%) probability of harboring pre-Columbian earthworks, with high probability concentrated along the southern and southwestern peripheries. The strongest opposing consideration is that LiDAR surveys continue to find earthworks in areas previously categorized as low-probability due to historic under-surveying beneath dense canopy.

GPT-5.6 Sol (OpenAI) (Unresolved) — AMZ-E038 and AMZ-E039 provide contextual results from the same model, including its basin-wide undiscovered-site total and concentration of high-probability cells, but neither states the claimed nearly 80-percent area at 0–1 percent probability. The strongest consideration against an unresolved verdict is that those contextual results are directionally compatible with extensive low-probability areas. Compatibility does not establish the exact proportion or its stated geographic concentration.

Grok 4.5 (xAI) (Unresolved) — No evidence record lists this claimId; Peripato passages appear only as context for C004 (E038/E039) and do not directly underwrite the 80 % / 0–1 % statement as a standalone assessed claim. Absent dedicated support the verdict is unresolved.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The 0–1% prediction over ~80% of the basin is verbatim-verified from the accepted manuscript, including the authors' own hedge that earth-builders 'may not have occupied all of Amazonia'. The opposing consideration is that this is a model prediction whose calibration has never been ground-truthed — exactly what the proposed blind-stratified LiDAR test AMZ-R006 would measure — and the source has minor internal inconsistencies in its predicted totals.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is attached. The basis is the claim's anchor quotation from the verified accepted manuscript ('Almost 80% of the basin has a 0 to 1% of predicted probability'), with the authors' gloss that earth-builders 'may not have occupied all of Amazonia'. The regional clause (northwestern, northern, central) appears in no quoted passage. The strongest opposing consideration is that the surface is fitted to presence-only data with a modelled observability correction. The several-fold gap between this model's basin total and the 2026 regional estimate (E038, E039) raises calibration questions, and low earthwork probability does not mean low occupation (AMZ-C020).

house: Unresolved

Gemini 3.8 Flash (Google) (Contradicted) — The assertion that CCC's entropy accounting is self-consistent is directly undermined by Eckstein (2023), who demonstrated that information loss during black hole evaporation increases global von Neumann entropy rather than resetting it, invalidating Penrose's proposed entropy renormalization across the crossover boundary. Furthermore, Tod (2022) identifies the matching of degrees of freedom at crossover as an unsolved theoretical challenge. The strongest opposing consideration is Penrose's original argument (2010, 2014) that gravitational degrees of freedom are uniquely destroyed by Hawking evaporation and conformal rescaling.

GPT-5.6 Sol (OpenAI) (Contradicted) — The only direct critical audit in the evidence argues that Penrose's assumed information loss increases global quantum entropy and is incompatible with the entropy renormalization required at crossover. That argument contradicts the strong assertion that the original accounting is demonstrably legitimate and self-consistent. The strongest opposing consideration is that this is one contribution within an unsettled quantum-gravity dispute and that a unitary CCC variant may preserve the broader framework, so the verdict does not establish that every possible entropy treatment fails.

Grok 4.5 (xAI) (Mixed) — Penrose’s entropy accounting via black-hole information loss plus conformal rescaling is the theory’s stated solution to the low-entropy initial condition. Eckstein 2023 argues that the assumed information loss increases global entropy and undermines the required renormalization, proposing a unitary variant instead. The strongest opposing consideration is that information loss remains a live controversy in quantum gravity with defenders on Penrose’s side, so the critique is not decisive.

GLM 5.3 Flash (Z.ai, via Venice) (Mixed) — Penrose's journal-length treatments present the entropy accounting (CCC-E016's companion record; SRC-PENROSE-2013-FOP), but Eckstein 2023 publishes a direct argument that information loss raises global entropy rather than permitting the crossover renormalization (CCC-E017), and Tod 2022 keeps related questions open (CCC-E016). The strongest opposing consideration is that Eckstein is one analysis in a live quantum-gravity controversy, by an author repairing rather than dismissing the framework, and his unitary variant preserves the observational program — so the accounting is contested, not refuted, which is exactly a mixed verdict.

Opus 5.5 (Anthropic) (Contradicted) — No evidence record presents the positive argument: the claim cites Penrose 2013, but nothing from that paper was extracted as evidence, so self-consistency enters the ledger asserted rather than shown. The one direct analysis, Eckstein 2023 (E017), argues that the assumed loss of degrees of freedom in black holes is incompatible with the quantum notion of entropy and raises global entropy, undermining the 'entropy renormalisation' the crossover needs. Tod's sympathetic inventory (E016) likewise treats adjacent crossover questions as open rather than settled. The evidence that exists therefore points against the claim, but thinly, hence low confidence. Strongest opposing consideration: this is one paper in a live quantum-gravity controversy in which Penrose's information-loss position has defenders not represented here, and Eckstein's own unitary variant suggests the framework may be repairable; my sense that information preservation is currently the majority view, which favours Eckstein, is background, not ledger evidence.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Well supported) — This statement articulates a foundational methodological axiom of archaeology, art history, and historical anthropology: morphological resemblance in isolation is underdetermined and cannot establish historical transmission without corroborating chronology, stratigraphic context, and plausible mechanisms of contact. While the provided evidence ledger contains no empirical record specifically evaluating this epistemic rule, the verdict is firmly supported by standard scientific priors in comparative historical analysis. The strongest opposing consideration is the radical formalist argument that certain highly specific, arbitrary symbolic configurations possess such high information complexity that visual resemblance alone constitutes prima facie evidence of cultural contact.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record in the ledger bears on this methodological claim; the research agenda does not target it either. It is a plausible methodological truism by general background, but the verdict would rest entirely on priors, which the rules forbid counting as ledger evidence. The strongest opposing consideration is simply that nothing retrieved tests it, so even a near-tautology stays unadjudicated here.

GPT-5.6 Sol (OpenAI) (Established) — No evidence record directly tests this methodological proposition, so the verdict rests on inferential logic and general scientific priors rather than an empirical result in the packet. A resemblance is compatible with multiple historical explanations unless chronology, context, differences, and feasible connections constrain them. The strongest opposing consideration is that an exceptionally specific resemblance can itself be probabilistically informative, but even then its historical interpretation requires contextual calibration.

Grok 4.5 (xAI) (Established) — The claim states a foundational requirement of historical interpretation: visual resemblance acquires chronological or cultural meaning only through dates, contexts, differences and possible connections. No evidence record disputes it; the principle is calibrated by standard archaeological and art-historical method (priors). Strongest opposing consideration is none; the statement is essentially definitional for any non-circular use of iconographic parallels.

Opus 5.5 (Anthropic) (Well supported) — A resemblance is equally compatible with common descent, contact, convergence and retelling. Only data outside the likeness itself can choose among those histories: dates, contexts, the detailed differences and a possible channel. This verdict rests on that inference, not on a test. The ledger illustrates it. The Rapa Nui birdman reliefs change meaning once shown to postdate the statue (E068, E069). The Urfa Man's hands grasp the genitals rather than rest on the body (E081), a difference that a shared "hands on the body" description hides. The strongest opposing consideration is that very specific, arbitrary, multi-element matches can make some connection probable from the resemblance alone, so the universal "no resemblance" overstates the point in degree; even then, direction, date and mechanism still need the other data.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Well supported) — The proposition that narratives can converge and become more alike through translation, intercultural borrowing, and comparative syncretism is a thoroughly documented principle of folkloristics and literary transmission. While the ledger lacks an explicit evidence record tracking a multi-stage textual divergence and convergence sequence, the verdict is well supported by general scholarly priors in comparative mythology. Translators and later chroniclers routinely harmonize foreign myths with familiar classical or biblical templates. The strongest opposing consideration is that many oral traditions possess formal mnemonic safeguards that preserve archaic distinctive details across generations, resisting harmonizing convergence.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record bears on this convergence mechanism; DEEP-C083, its child about Oannes's nightly returns, also has no retrieved evidence. The mechanism is familiar from documented textual history, but a verdict here would lean wholly on background priors. The strongest opposing consideration is that documented retellings could easily be found, so the claim is testable at desk tier and its unresolved status reflects ledger gaps, not intrinsic difficulty.

GPT-5.6 Sol (OpenAI) (Unresolved) — No evidence record presents an earlier version and a translated, borrowed, or comparative retelling that became more similar in detail. The proposed mechanisms are plausible on general background knowledge, but the claim's stated truth condition requires documented convergence. The strongest consideration against leaving it unresolved is the broad prior plausibility of translation and borrowing, which cannot substitute for the absent before-and-after evidence.

Grok 4.5 (xAI) (Established) — The mechanistic claim that stories can converge in detail through translation, borrowing and comparative writing is a documented process in folklore and textual transmission; general background supplies numerous attested cases (e.g., flood narratives). No evidence record is required beyond the logical possibility the claim asserts, and none contradicts it. Strongest opposing consideration is that not every pair of similar stories has undergone such secondary convergence, yet the claim is existential (“can”) and therefore holds.

Opus 5.5 (Anthropic) (Provisionally supported) — The claim asserts only that retelling can make stories converge. On general background it is very likely true: it is widely recognised that translators, missionaries, colonial chroniclers and comparative compilers reshaped the myths they recorded toward familiar templates. The verdict leans on that background, disclosed as such. The ledger itself holds only leads. Berossus survives solely through excerptors (Polyhistor, Syncellus). The founder's "arrives, teaches and departs" template may also compress an account in which Oannes returns to the sea each night, though that point is anchored in a claim and in no evidence record. The strongest opposing consideration is that no evidence record documents a retelling converging relative to an earlier attested version, which is the claim's own truth condition. The mechanism's reach is also limited: arrival from the sea and the gift of the arts are already in Berossus's ancient text (E043, E046), so retelling cannot account for all of the civiliser resemblance.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Established) — This methodological claim is established based on the fundamental scientific requirements of historical inference and falsifiability. Asserting that an unknown, advanced ancestral civilization existed at a specific time and possessed advanced seafaring or engineering capabilities requires primary material evidence such as dated artifacts, metallurgical residues, genetic admixture, or domesticate transfers; formal similarity between stone carvings cannot substitute for these physical parameters. Leaning on standard epistemological priors, similarity alone cannot settle external empirical attributes. The strongest opposing consideration is the hyper-diffusionist contention that an extensive suite of identical complex cultural technologies appearing simultaneously worldwide necessitates a single capable progenitor civilization.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record addresses whether similarity alone can carry a lost-civilisation account's identity, date and capability claims. The claim is close to methodologically tautological, but the ledger contains no record of it, and a verdict would rest on priors alone. The strongest opposing consideration is that no retrieved record tests it against an actual lost-civilisation argument, so its practical force is undemonstrated here.

GPT-5.6 Sol (OpenAI) (Established) — This follows from inferential underdetermination rather than a particular evidence record: resemblance alone contains no unique information about a source's identity, date, or capabilities. Those additional attributes require independent chronological, archaeological, or mechanistic evidence. The strongest opposing consideration is that a rare and complex resemblance could increase the probability of common origin, but it still could not by itself establish all the added claims.

Grok 4.5 (xAI) (Established) — Similarity of motifs is insufficient to establish the stronger identity, date and capability claims that a lost-civilisation account adds; those require independent evidence. The statement is a direct consequence of historical method (priors) and is nowhere challenged by the ledger. Strongest opposing consideration is the diffusionist view that dense pattern clusters can themselves license source inferences, but the claim correctly insists on evidence beyond resemblance alone.

Opus 5.5 (Anthropic) (Established) — This is an inferential point rather than an empirical finding. A resemblance between motifs carries no information about who the source was, when it existed or what it could do. Any such claim therefore needs evidence beyond the resemblance: dated remains, material culture, a demonstrated channel. The ledger shows what that further evidence would look like and that it is missing for the compared cases (the excavators' "missing links", E061). The strongest opposing consideration is that some motifs carry content, such as encoded astronomy or a technical procedure, from which capabilities might be inferred. But that is evidence from content and context rather than from similarity, so it falls outside the claim rather than against it.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Established) — Methodologically, the geographical distribution of a broad, polysemic motif class cannot differentiate between independent invention and very ancient shared ancestry without auxiliary chronological and phylogenetic data. Broad motifs such as cosmic serpents or bird-people can emerge independently from universal cognitive, somatic, or ecological regularities, generating spatial patterns that closely mimic ancient demic or cultural diffusion. This evaluation leans on established epistemological principles in spatial analysis and evolutionary anthropology. The strongest opposing consideration is that phylogeographic and spatial diffusion models can sometimes distinguish cultural transmission pathways from random independent innovation if sampling density and dispersal kernels are rigorously calibrated.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record directly supports this logical point; DEEP-C072's 'missing links' statement concerns a specific empirical gap, not the underdetermination claim itself. The claim is plausible on general grounds — a widespread distribution is compatible with both histories — but that is a prior, not ledger evidence. The strongest opposing consideration is that a distribution plus dated intermediates could in principle discriminate, which the claim's own 'by itself' wording already concedes.

GPT-5.6 Sol (OpenAI) (Established) — A geographical distribution can be generated under both transmission and repeated invention unless explicit models or additional historical constraints distinguish them. This verdict relies primarily on methodological logic and scientific priors because no evidence record performs the stated model comparison. The strongest opposing consideration is that some highly structured distributions may favor one model, but a distribution by itself does not uniquely determine its generating history.

Grok 4.5 (xAI) (Established) — A geographic distribution of a broad motif class is compatible with both independent invention and very old transmission under standard models of cultural evolution; the catalogue result therefore cannot by itself decide between them. The claim is methodological and reinforced by the excavator's missing-links statement (E061/E062). Strongest opposing consideration is that sufficiently dense and dated intermediate finds could in principle distinguish the histories, yet the claim is limited to “by itself” and remains true.

Opus 5.5 (Anthropic) (Well supported) — A broad class present in a large share of traditions is predicted by two histories: very old transmission from a common source, and repeated invention prompted by animals, bodies and landscapes found almost everywhere. A presence-absence map therefore yields no likelihood ratio between them. The verdict rests on this identifiability argument rather than on a record. The nearest evidence (E061, E062) shows that the discriminating datum, dated intermediates, is exactly what the excavators say is missing even for the local Near Eastern case. The strongest opposing consideration is that areal and phylogenetic methods can extract historical signal from distributions under explicit models (spatial autocorrelation, fit to language trees, clustering of specific rather than broad motifs). "Cannot by itself" thus holds for raw distributions of broad classes, not for modelled distributions of specific motifs.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — The interpretive claim that the colloquial term 'bird-man' conflates distinct conceptual categories—namely theriomorphic deities, ceremonial humans adorned in feathers, and shamans undergoing avian transformation—is well supported by anthropological and iconographic priors. In the provided case materials, the Rapa Nui birdman represents an ethnohistoric human cult and sociopolitical competition (DEEP-E068, DEEP-E071), whereas Mesopotamian avian-headed figures represent mythic apotropaic beings. Conflating these fundamentally different semantic domains under a single visual label commits an obvious taxonomic fallacy. The strongest opposing consideration is that proponents of archaic shamanic substrate models view all three expressions as historically linked variations of a single primordial trance-transformation tradition.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record tests whether the compared bird-man images actually fall into the three distinct categories the claim distinguishes. The distinction is conceptually sound and probably right as iconographic method, but that is background calibration, not ledger evidence. The strongest opposing consideration is that the claim's truth condition requires the compared images to be classified, and no retrieved record classifies any of them.

GPT-5.6 Sol (OpenAI) (Unresolved) — No evidence record classifies the compared images as bird-headed deities, feather-wearing people, or people transforming into birds. Those concepts are intelligibly distinct, but that conceptual distinction does not establish that the actual comparison set occupies all three categories. The strongest consideration against the unresolved verdict is the face-validity of the distinctions themselves, which is weaker than evidence about the images.

Grok 4.5 (xAI) (Provisionally supported) — The interpretive claim that the label “bird-man” conflates at least three distinct ideas (bird-headed deity, feathered person, transforming person) holds if the compared images fall into different categories, which background iconographic knowledge confirms is often the case. No direct evidence record addresses the typology. Strongest opposing consideration is that some cultural traditions deliberately blend the three, yet the claim only requires that they need not be identical.

Opus 5.5 (Anthropic) (Provisionally supported) — The conceptual core is sound: a bird-headed deity, a person in feathers and a person becoming a bird are different claims about what an image shows. The truth condition, however, requires that the compared images actually fall into different categories. No evidence record classifies the Göbekli, Nimrud or Rapa Nui figures. The nearest record (E071) ties the Rapa Nui reliefs to the island's ethnohistorically recorded birdman myth. On general background that myth concerns a human ritual office rather than a bird-headed god, a step that leans on priors. The strongest opposing consideration is that the categories may blur in practice, as when a masked or costumed person enacts a deity or a transformation. The three-way split could then be an analyst's distinction that hides a shared underlying concept.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Established) — This methodological assertion is established as a basic tenet of material culture studies: morphological similarity does not establish functional equivalence. Handled containers, for instance, can serve wildly disparate utilitarian, ritual, or symbolic purposes, ranging from libation buckets to incense burners or architectonic representations. While no specific evidence record in the ledger is tagged to this claim, it is solidly grounded in standard archaeological reasoning. The strongest opposing consideration is an extreme functionalist design perspective claiming that highly unusual or arbitrary shapes are so narrowly constrained by their operational mechanics that morphology inevitably determines function.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record addresses the shape-versus-function separation; DEEP-R005 targets it and is still open, and the anchor's own quotes are a methodological assertion, not retrieved evidence. The strongest opposing consideration is that DEEP-E028's team statement supports the parallel uncertainty claims (DEEP-C012, DEEP-C066), which is consistent with but does not establish this principle.

GPT-5.6 Sol (OpenAI) (Established) — DEEP-E025 describes the Pillar 43 forms morphologically as box-like with handle-like attachments, while DEEP-E028 says their meaning cannot presently be decided. Together they concretely demonstrate that a shape can be described without identifying its function, consistent with the general methodological distinction. The strongest opposing consideration is that shape can constrain functional hypotheses, but constraint is not identity.

Grok 4.5 (xAI) (Established) — Shape and function are logically and empirically separable questions; the Göbekli forms can be described as box-like with handle-like attachments (E025) while their function remains undecided (E028), and the Assyrian banduddû can be functionally identified without that identification transferring to every similar shape. The methodological distinction is standard (priors). Strongest opposing consideration is none; conflating the two is a recognised error.

Opus 5.5 (Anthropic) (Well supported) — The ledger itself shows three similar handled shapes whose functions rest on entirely different bases. The banduddû's rests on Assyrian ritual texts (E055), and the Jiroft object's on its own wear traces (E035, E038). The Pillar 43 forms' rests on nothing secure: Schmidt describes shape without naming a function (E025), and the excavation team declines a decision while entertaining a building reading (E028). That the same carved forms can be read either as handled containers or as buildings is direct evidence that shape does not settle function. The strongest opposing consideration is that shape constrains function through affordances; a handle is some evidence for carrying. The questions are therefore separable rather than independent, and the verdict rests on their separate evidential bases, not on shape being irrelevant.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Well supported) — The claim argues that the Greek omphalos, Garcilaso's gloss of Cusco, the Rapanui Te Pito o te Henua, and the modern Turkish Göbekli Tepe are fundamentally different kinds of historical evidence that differ by language, first recorded date, and builder usage. This is solidly supported by historiographical priors: the omphalos is an ancient indigenous term attested in classical antiquity; Cusco's gloss is a 17th-century colonial Quechua translation; Te Pito is a 19th-century post-contact recording; and Göbekli Tepe is a 20th-century Turkish descriptive toponym unknown to Neolithic builders. The ledger provides no counter-evidence to this obvious source-critical divergence. The strongest opposing consideration is that comparative mythologists argue that independent cultures frequently preserve ancient cosmological 'navel' designations through unbroken oral substrate traditions.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record touches the source histories of the four navel names; DEEP-R002, which targets exactly this, is unanswered. The claim is plausible as methodological housekeeping, but its truth condition — that the four differ on language, first record, and builder usage — requires the very attestations not yet retrieved. The strongest opposing consideration is that any one of the four names turning out to have early, builder-used 'navel' attestation would falsify it, and nothing in the ledger rules that out.

GPT-5.6 Sol (OpenAI) (Unresolved) — No evidence record supplies the source histories, first attestations, and builder usage for all four named expressions. The available toponymic material concerns a separate Kurdish “wish hill” statement and does not establish the three specified dimensions of comparison. The strongest consideration against the unresolved verdict is that the examples are likely heterogeneous on general historical priors, but those priors are not a substitute for the required records.

Grok 4.5 (xAI) (Established) — The four “navel” expressions differ in language, date of first record and whether ancient builders used the name: Greek omphalos is classical, Garcilaso's Cusco gloss is colonial, Rapanui Te Pito o te Henua is ethnohistoric, and Turkish Göbekli Tepe is modern descriptive. The claim's truth condition (differing source histories on those three points) is met by well-known documentary facts (priors). Strongest opposing consideration is none; the differences are factual.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record addresses these source histories (R002 is open). The verdict therefore leans on general background and on what the claim's own terms entail. The four expressions are in different languages. Garcilaso's gloss is a colonial record of 1609, the Greek omphalos an ancient usage, and Göbekli Tepe a Turkish toponym. A Turkish name cannot have been used by the site's Neolithic builders, since the language reached Anatolia millennia later. The strongest opposing consideration is that the builders'-use question is genuinely uncertain for Cusco and Rapa Nui, and no first-record date has been retrieved. The pattern of differences is therefore asserted rather than documented.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Established) — The authors' own published methodological report explicitly confirms that andesite sample PP1 was selected from loose, scattered debris lying outside the protected monument area based solely on visual flat-surface characteristics. There is no documented excavation provenance, inventory number, or architectural refit linking fragment PP1 to an in-situ monument block. The strongest opposing consideration is the authors' assertion that the debris exhibited flat surface finishes characteristic of monumental worked blocks.

GPT-5.6 Sol (OpenAI) (Established) — GEO-E057 directly quotes the authors saying that the fragments lay outside the protected monument area and were selected from debris by visually characteristic flat surfaces. The record identifies PP1 as a ground fragment rather than a sample removed from a catalogued block. The strongest opposing consideration is that loose placement does not prove the fragment was unrelated to the monument.

Grok 4.5 (xAI) (Established) — The 2019 source itself states that the andesite fragments, including PP1, lie outside the protected monument area and were chosen by flat-surface appearance from scattered debris (GEO-E057). No contrary evidence appears. Strongest opposing consideration is none; the sampling description is unambiguous.

GLM 5.3 Flash (Z.ai, via Venice) (Established) — The proponents' own text states the fragments lie outside the protected monument area and were chosen by appearance, and the quoted passage was verified against the source. The only counterweight is that being a loose surface find does not logically exclude derivation from an architectural block, but the claim only asserts how the sample was selected, which is exactly what the source says.

Opus 5.5 (Anthropic) (Well supported) — The analysts' own account states that the andesite fragments "are outside the protected monument area" and that samples were obtained by choosing debris with the characteristic flat surface. It records that PP1 A and B came from the fresh break of a fragment on the ground, with no object number, excavation context or refit (GEO-E057). A sampler's own description of how they sampled is the best available evidence on the point, and the interior finding (GEO-C918) is on PP1. Strongest opposing consideration: the peer-reviewed Ceramics International description was not read and could add context, and lying outside the protected area does not exclude an architectural origin — but the claim says only that no identified block was sampled, which the record supports.

house: Provisionally supported

Gemini 3.8 Flash (Google) (Established) — The source text (Davidovits et al. 2019) explicitly describes sectioning an interior subsample labeled PP1C from the interior of fragment PP1A to test against surface contamination, and reports EDS spectra showing carbon, nitrogen, and other elements forming a dark sheet beneath plagioclase crystals. As a description of what the analysts report in their publication, the claim is completely accurate. The strongest opposing consideration is that EDS elemental analysis cannot definitively characterize the molecular structure or age of the organic compound.

GPT-5.6 Sol (OpenAI) (Established) — GEO-E058 directly records the analysts' report that PP1C was cut from the fragment's interior and contained similar carbon- and nitrogen-bearing matter, including a sheet beneath plagioclase crystals. The claim is limited to what those analysts reported and does not independently identify the compound or its age. The strongest opposing consideration is that the observation comes from the same team and insecurely provenanced fragment as the original surface finding.

Grok 4.5 (xAI) (Provisionally supported) — The same 2019 report states that interior subsample PP1C, cut from PP1, yielded the same C/N-bearing amorphous EDS spectrum beneath plagioclase (GEO-E058). This is the authors’ own observation and has not been independently repeated. Strongest opposing consideration is that the parent fragment’s insecure provenance (C913) and lack of natural control still leave open contamination or natural glass.

GLM 5.3 Flash (Z.ai, via Venice) (Established) — The claim is descriptive of what the analysts reported: an interior subsample PP1C showing the same EDS spectrum beneath plagioclase crystals, verified against the source. It is one fragment of insecure provenance analyzed by the proponents with no independent replication, so the underlying observation's significance stays open even though the report of it is solid.

Opus 5.5 (Anthropic) (Established) — The 2019 account reports cutting PP1C from the interior of PP1A specifically to meet a surface-contamination objection. It reports several interior spots with the same EDS spectrum (C, N, Na, Mg, Al, Si, P, S, Cl, K, Ca), including a dark sheet beneath two plagioclase crystals (GEO-E058). The claim is framed as what the analysts report, and that is documented in their words. Strongest opposing consideration: the same report limits what the observation can mean. Nitrogen cannot be quantified with their EDS (GEO-E059), no molecular identification was made, there is no natural andesite control (GEO-E060), and the fragment's attribution to the monument is by appearance (GEO-C913), so an accurate report still says little about origin.

house: Mixed

Opus 5 (Anthropic) (Mixed) — The literal proposition — microtubule-modulating drugs shift anesthetic susceptibility in animals — is well evidenced across four agents and two programs, but the attribution to microtubule stability as the operative variable is contradicted by the data pattern: chronic epothilone D and the destabilizer vinblastine shift sensitivity in the same direction (ORCH-E013), opposite to acute epothilone B (ORCH-E011, ORCH-E012). "Mixed" here is a substantive finding about that split, not a hedge. The strongest opposing consideration for either camp is that these drugs perturb transport, metabolism, cellular integrity and — as the Li study documents — blood pH and CO₂, so the whole literature may be measuring general cellular disruption rather than anything about microtubules as such.

Gemini 3.8 Flash (Google) (Well supported) — Multiple independent animal studies across tadpoles, rats, and mice demonstrate that administering microtubule-stabilizing or destabilizing drugs significantly alters behavioral sensitivity to general anesthetics. This empirically confirms a functional link between microtubule stability and anesthetic susceptibility. The strongest opposing consideration is that the directional relationship is contradictory across laboratories and dosing regimens, with chronic epothilone D and vinblastine both increasing sensitivity in mice, opposite to acute epothilone B.

GPT-5.6 Sol (OpenAI) (Well supported) — Multiple live-animal studies report altered anesthetic susceptibility after pharmacological microtubule manipulation, including acute epothilone B and chronic epothilone D, vinblastine, and paclitaxel regimens. The opposing directions do not negate this claim as worded; they undermine a simple rule about stabilization protecting consciousness. The strongest opposing consideration is that these drugs have broad cellular and physiological effects, so microtubule stability may not be the causal mediator of every behavioral change.

Grok 4.5 (xAI) (Well supported) — Acute epothilone B delays isoflurane LORR in rats and mice, and chronic epothilone D and vinblastine shift LORR concentration–response curves in mice; Emerson’s tadpole work also ties MT stabilizers to anesthetic requirement. The claim only requires that stability manipulation changes susceptibility, which these records show even though directions conflict. Strongest opposing consideration is that conflicting directions and shared cellular confounds block any single mechanistic reading from stability to consciousness.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — Multiple independent lines — Emerson's tadpoles, the Wellesley rat and mouse studies, the Massachusetts General concentration–response study, and the weak retrospective Linganna human data — all show that manipulating microtubule stability changes anesthetic susceptibility. The claim as worded asserts change, not direction, and change is thoroughly established. The direction contradiction and wholesale-perturbation confounds damage the parent claim ORCH-C002, not this statement, though they cap confidence at moderate.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Helou et al. analyzed relative acceleration noise from the free-falling test masses of the LISA Pathfinder mission, bounding the Diósi–Penrose regularization scale to at least 40.1 ± 0.5 fm, exceeding the diameter of atomic nuclei. This provides a direct, non-radiation bound on collapse-induced momentum diffusion. The strongest opposing consideration is that the bound is conservative because it attributes all unexplained residual acceleration noise to wavefunction collapse rather than unaccounted classical thermal dissipation.

GPT-5.6 Sol (OpenAI) (Well supported) — ORCH-E046 reports the 40.1 ± 0.5 fm lower bound derived from LISA Pathfinder's acceleration-noise measurement and correctly states that this exceeds nuclear dimensions. The bound is conservative because all residual acceleration noise is attributed to collapse. The strongest opposing consideration is its model dependence: it applies to a regularized Diósi–Penrose realization with a specified cutoff, not automatically to Penrose's bare criterion.

Grok 4.5 (xAI) (Well supported) — Helou et al. attribute LISA Pathfinder residual relative acceleration noise to collapse-induced force noise and bound the DP regularization length to at least ~40.1 fm, larger than any nucleus, ruling out particle-sized cutoffs for that model class. Strongest opposing consideration is that the bound is conservative (thermal noise dominates) and applies to regularized DP with a point-mass cutoff rather than every reading of Penrose’s heuristic.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The LISA Pathfinder re-derivation is a conservative attribution of all residual acceleration noise to collapse, quoted accurately, and it constrains a channel — mechanical momentum diffusion — that the radiation-based escape route does not reach. The bound is conditional on attributing residual noise to collapse and on the regularized point-mass formulation, which is why it qualifies rather than undermines ORCH-C030 as stated. Solid as a reported bound, moderate because it is a re-derivation rather than a direct measurement.

house: Unresolved

Gemini 3.8 Flash (Google) (Established) — The 2026 paper's qualitative analytical spectra show that signals corresponding to ergometrine and its epimer ergometrinine remained detectable even in the fully hydrolyzed preparation where predominant toxic peptide alkaloids had disappeared. The strongest opposing consideration is that the authors did not systematically evaluate the pharmacological potency or toxicology of this residual non-peptide alkaloid fraction.

GPT-5.6 Sol (OpenAI) (Unresolved) — No admitted evidence record directly documents the claimed detection of both ergometrine and ergometrinine in the specified preparation. TIK-E038's title and editor note allude to residual ergometrine, but its quoted source statement does not provide this result and its verified claim links explicitly exclude TIK-C029. That allusion is the strongest consideration favoring the claim, but it is insufficient under the evidence-record-only rule.

Grok 4.5 (xAI) (Provisionally supported) — The 2026 qualitative analysis reports that signals for ergometrine and ergometrinine remained visible in the preparation in which the peptide ergot alkaloids had been cleared. The observation is part of the same experimental record that supports the conversion claim. The strongest opposing consideration is the absence of a dedicated, independently verified evidence record that isolates this residual after the claim-ID pruning, plus lack of pharmacological evaluation of the residual in the source.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record in the ledger attaches to this claim: the intake verification explicitly detached the E038 passage from C029 because the quoted text does not bear on ergometrine persistence, and the separate passage this claim needs was never extracted into an evidence record. The claim's own source anchor quotes the preprint's Results directly, which suggests it is accurate, but under this ledger's rules that untested anchor is not evidence; the passage should be captured and verified before the claim is graded.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record is formally linked: the second reader removed C029 from E038 because E038's quoted passage lacks the ergometrine statement. The basis I disclose instead is the claim's own anchor, a verbatim quote from the retrieved preprint ("while the signals corresponding to ergometrine (Em) and ergometrinine (Emn) were observed"), together with E038's title and editorial note, which record the source as reporting ergometrine persisting. For a narrow claim about what a text says, that suffices for provisional support. Strongest opposing consideration: the quote is a fragment whose link to the optimal condition is inferred from its locator rather than shown, and the published version has not been checked for this passage. Whether residual ergometrine matters pharmacologically is an editor's note, not the source's, and belongs to R003.

house: Weakly supported

Gemini 3.8 Flash (Google) (Well supported) — Stoichiometric yields reported by Antonopoulos et al. indicate that one gram of ergot processed under optimal alkaline reflux produces approximately 0.54 mg of LSA and 0.48 mg of iso-LSA, matching the minimum effective human threshold reported in pharmacological literature. The strongest opposing consideration is that ancient ergot batches varied dramatically in alkaloid content and that crude decoctions may exhibit reduced oral bioavailability compared to purified compounds.

GPT-5.6 Sol (OpenAI) (Provisionally supported) — TIK-E008 reports 0.54 mg of LSA per gram of ergot and states that the authors extrapolated that about one gram could supply an active dose. The numerical arithmetic therefore supports the claim under the optimal experimental conditions. The strongest opposing consideration is that the literature threshold is not independently audited here, ergot composition can vary, and reaching a nominal minimum dose does not imply reliable or positive psychoactivity.

Grok 4.5 (xAI) (Well supported) — The 2026 discussion explicitly extrapolates that roughly one gram of ergot processed under optimal lye conditions supplies the literature minimum effective dose of approximately 0.5 mg LSA. The arithmetic follows directly from the measured yields. The strongest opposing consideration is that the minimum-dose figure is taken from external LSA literature (not the reconstructed matrix) and that residual ergometrine and co-products are unprofiled.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — The arithmetic is transparent from the measured yield: 0.54 mg LSA per gram of processed ergot meets the 0.5 mg literature minimum effective dose, and the source itself states about 1 g per initiate suffices. The strongest opposing consideration is that the "minimum effective dose" figure is extrapolated from a heterogeneous morning-glory-dominated human literature, not from the reconstructed matrix, so the claim's own hedged framing ("about," "literature minimum") is what the evidence warrants.

Opus 5.5 (Anthropic) (Provisionally supported) — E008 reports a yield of 0.54 mg LSA per gram at the optimal condition, against the paper's cited minimum effective dose of about 0.5 mg (per the claim's anchor), so the statement holds as a calculation. It is only provisional for four reasons: no evidence record is formally linked; the margin is under a tenth; the yield comes from a single ergot sample (that field ergot varies in alkaloid content is background, not a ledger measurement; R002); and the dose figure is the authors' citation, unchecked here. Strongest opposing consideration: a minimum effective dose is by definition a threshold, so even an exact figure supports a noticeable effect, not the profound experience C006 requires.

house: Unresolved

Gemini 3.8 Flash (Google) (Well supported) — Historical botanical records indicate that Paspalum distichum and related host species of Claviceps paspali did not colonize or naturalize in the European Mediterranean basin until the 19th century CE. The strongest opposing consideration is that paleobotanical microfossil sampling remains incomplete across the ancient Mediterranean, leaving minor uncertainty regarding ancient host distributions.

GPT-5.6 Sol (OpenAI) (Unresolved) — No evidence record in the packet bears directly on when the host plants of Claviceps paspali became naturalized in Europe. The claim's own source anchor points to a sentence in the preprint, which is the strongest consideration in its favor, but source anchors are not evidence records under the assessment rule. With neither supporting nor undermining evidence admitted, the historical statement remains unresolved.

Grok 4.5 (xAI) (Well supported) — The 2026 discussion states that the host plants of Claviceps paspali became naturalized in Europe only in the nineteenth century, correctly excluding that species from any ancient Mediterranean reconstruction. The botanical-historical claim is straightforward. The strongest opposing consideration is that it is a single-sentence citation inside an interested-network paper rather than a primary botanical survey audited here.

GLM 5.3 Flash (Z.ai, via Venice) (Unresolved) — No evidence record in the ledger connects to this claim; it carries only a source anchor transcribed from the 2026 paper's Discussion, itself citing a reference not obtained or verified in this case. It is a plausible historical-geographical statement and matters for ruling out C. paspali as an ancient European source, but under this ledger's discipline it currently has no weighed basis and needs its own record before it can support or undercut anything.

Opus 5.5 (Anthropic) (Provisionally supported) — No evidence record addresses this claim. The basis I disclose is its anchored sentence in the retrieved preprint ("only became naturalized in the 19th century"), itself a citation to an unnamed botanical reference (ref. 52). General botanical background that the main Paspalum hosts are of New World origin makes it plausible, but that is a prior, not a ledger record. Strongest opposing consideration: naturalization dates are fuzzy (first record versus establishment, across several host species), and one sentence in a chemistry paper is thin support. Its bearing on the hypothesis is indirect: it excludes that ergot species from the ancient Aegean, so any reconstruction must run through C. purpurea and a processing step.

house: Unresolved

Gemini 3.8 Flash (Google) (Established) — Decades of archival astronomical plate research and gamma-ray burst optical follow-up establish that chemical flaws, dust, pinholes, and copy defects routinely mimic point-like sources. The sheer volume of false positives eliminated in building the VASCO and Hayes catalogs confirms high baseline defect contamination. Strongest opposing consideration: Rigorous multi-parameter morphological filtering and cross-epoch verification can identify and reject the vast majority of recognizable, non-stellar artifacts.

GPT-5.6 Sol (OpenAI) (Well supported) — TRN-E002 finds point-like detections and visible blemishes on physical copy plates, TRN-E016 documents thousands of candidates classified as artifacts, and TRN-E008 shows that a loose archival search produces millions of candidate detections. TRN-E004 further finds that the April marks occupy artifact-like morphology space, although that classifier is not decisive for individual objects. The strongest opposing consideration is the absence of a controlled census measuring the rate of star-like defects at each photographic generation, especially on original negatives.

Grok 4.5 (xAI) (Well supported) — Watters et al. document that at least 91 % of features in the related datasets lie within 5 arcsec of catalog stars, that spatial distributions are heterogeneous with edge and corner excesses, and that decades of GRB-era archival searches failed to confirm optical counterparts amid plate defects (E011). Hambly & Blair likewise catalog emulsion holes, bubbles and scratches on the same copy plates (E002). Strongest opposing consideration is that a residual real-transient population could still sit inside the defect background; the claim only asserts sufficiency of contamination rates, which the records support.

GLM 5.3 Flash (Z.ai, via Venice) (Well supported) — Multiple records show defect populations sufficient to contaminate such searches: blemishes on the copy plates (TRN-E002), artifact-like morphology assignments (TRN-E004), 3,592 vetted artifacts in the pipeline (TRN-E016), ~91% of ensemble-study features within 5 arcsec of catalog stars (TRN-E011), and a 500-fold gap between raw and vetted candidates (TRN-E008). The strongest opposing consideration: "rates sufficient" is nowhere quantified — the per-generation defect census (TRN-R007) was never run, and TRN-E023 disfavours this mechanism for the specific canonical events.

Opus 5.5 (Anthropic) (Well supported) — No record is linked, but records from both camps converge on this. The pipeline removed 3,592 artifacts against 5,399 survivors, and the working sample admits substantial false positives (E016). The proponents built a purpose-built defect classifier (E015). Counts rise toward plate corners and edges (E011), and the copies carry visible blemishes (E002). This is also standard plate-archive background. Strongest opposing consideration: no per-generation base rate exists (TRN-R007), so contamination is shown for the candidate stream but not quantified for any canonical event.

house: Unresolved

Gemini 3.8 Flash (Google) (Provisionally supported) — Standard optical and atmospheric principles dictate that a flash lasting less than a second will not integrate seeing excursions or telescope tracking jitter over an hour-long exposure, naturally resulting in a tighter, rounder profile. However, this theoretical deduction has not been experimentally calibrated on historical photographic emulsions. Strongest opposing consideration: Photographic turbidity and non-linear emulsion scattering at high instantaneous flux levels could broaden sub-second optical flashes, diluting profile differences.

GPT-5.6 Sol (OpenAI) (Well supported) — TRN-E006 gives a physically coherent mechanism: a sub-second point flash avoids much of the seeing and guiding blur accumulated by a continuously exposed stellar image during a long exposure. This supports the expected direction of the sharpness and circularity difference. The strongest opposing consideration is that no controlled experiment on comparable historical emulsions has measured the flash-profile distribution or shown that it separates cleanly from artifact profiles.

Grok 4.5 (xAI) (Provisionally supported) — Optical and atmospheric arguments show that unresolved sub-second flashes accumulate neither seeing nor guiding blur and therefore appear sharper and rounder than stars on long exposures (E006). No empirical calibration on period-appropriate Schmidt emulsions exists. Strongest opposing consideration is that the two profile distributions (flash versus defect) may overlap completely, rendering the argument non-discriminating until TRN-R002 is performed.

GLM 5.3 Flash (Z.ai, via Venice) (Provisionally supported) — The optics argument (TRN-E006) is physically sound: an unresolved sub-second flash avoids the seeing and guiding blur that broadens stellar images on a long exposure. But it is an argument, not a measurement — no period emulsion has been exposed to controlled flashes (TRN-R002), and the record itself notes the flash and artifact profile distributions may overlap entirely, in which case sharpness discriminates nothing. Strongest opposing consideration: the proponents' own ~320 ms characteristic duration (TRN-E064/E065) sits near their model's elongation threshold, an unresolved internal tension.

Opus 5.5 (Anthropic) (Provisionally supported) — The direction follows from standard optics. A sub-second exposure escapes most of the seeing wander and guiding error that broaden a ~50-minute stellar image. That is a prior, however: the ledger holds only the proponents' argument (E006) and no emulsion calibration (TRN-R002). Strongest opposing consideration: the effect may be small. The original-plate marks are only "marginally sharper" (E022), the 1952 triple's isolated source looks stellar (E007), and flashes longer than ~0.2–0.5 s should elongate (E065).

house: Unresolved

Gemini 3.8 Flash (Google) (Provisionally supported) — Multiple preprints report statistically improbable linear alignments of five to eight transients on several POSS-I plates. However, critical reviews indicate that a significant fraction of these candidate features cannot be reliably distinguished from known catalog stars, and the alignments suffer from uncorrected trials factors across multiple plates. Strongest opposing consideration: The low Monte Carlo p-values (p < 0.03) reported on several plates where candidate points exhibit star-like morphology rather than continuous trailing.

GPT-5.6 Sol (OpenAI) (Provisionally supported) — TRN-E018 reports seven POSS-I plates containing alignments of five to eight point-like candidates exceeding the study's Monte Carlo expectations. That is direct support for the limited statement that some events are reported to form linear alignments. The strongest opposing consideration is the unpreregistered alignment criterion and associated look-elsewhere risk, compounded by the unresolved provenance of the underlying marks.

Grok 4.5 (xAI) (Unresolved) — Alignments are asserted in the PASP paper (unread in full) and extended by a later ML-scored search reporting seven plates with 5–8 sources above Monte-Carlo expectations (E018), yet p-values are modest, criteria post-hoc, and the underlying candidates inherit the contested catalog provenance. Strongest opposing consideration is look-elsewhere inflation and lack of preregistered alignment definitions or independent-catalog replication.

GLM 5.3 Flash (Z.ai, via Venice) (Weakly supported) — Alignment claims exist in a peer-reviewed PASP paper, but that paper was not read in full for this ledger (the source record says so), and the closest available record (TRN-E018) is a same-network preprint with modest p-values, no preregistered alignment criterion, and no trials-factor accounting. TRN-E011 reports that a third of the reported linear-cluster features were not confidently distinguished from catalog stars. Strongest opposing consideration: alignment searches over many plates with post-hoc criteria are the classic setting for look-elsewhere inflation, on a catalog whose candidate status is contested.

Opus 5.5 (Anthropic) (Well supported) — As stated, the claim is only that such alignments are reported, and they are. E018 reports seven plates with point-like alignments of five to eight sources beyond Monte Carlo expectation, and Watters analyzes the reported linear clusters (E011). This verdict records the reporting, not the alignments' reality. Strongest opposing consideration: the substance is weak. A third of the linear-cluster features were not distinguished from catalog stars (E011), and E018's p<0.03 uses a criterion that was not preregistered and was searched across many plates.

house: Mixed

Gemini 3.8 Flash (Google) (Contradicted) — A comprehensive chronological audit showed that only 3 of 29 claimed boundary sites possessed radiocarbon control sufficient to place their proxy horizons at ~12.8 ka cal BP. The apparent global isochron was constructed by selective date inclusion and circular boundary assignments, leading directly to the 2026 retraction of the flagship shocked-quartz and cometary-dust studies for age-model defects. The strongest opposing consideration is Kennett et al.'s 2015 Bayesian analysis, which concluded that calibrated radiocarbon dates from four continents are statistically consistent with a 12,835–12,735 cal BP boundary.

GPT-5.6 Sol (OpenAI) (Contradicted) — YDIH-E019 found that only 3 of 29 claimed sites had adequate dating control to place their proxy horizon near 12.8 ka, directly undermining the asserted four-continent isochron. YDIH-E020 found that selected data were consistent with a common 12,835–12,735 cal BP interval, but its force depends on disputed date selection and boundary assignments that were not independent of the hypothesis. The chronology and incomplete-background defects documented in the 2026 retractions reproduce the same failure mode. The strongest opposing consideration is the formal Bayesian synchrony model, but consistency under hypothesis-dependent inputs is not a secure demonstration of one layer.

Grok 4.5 (xAI) (Contradicted) — Site-by-site audit shows only 3 of 29 claimed localities possess radiocarbon control adequate to place proxies at ~12.8 ka; Bayesian models claiming synchrony rely on contested date selection and boundary assignment that risk circularity, and the 2026 retractions explicitly cite age-model failures. Strongest opposing consideration is the formal consistency of selected dates with a narrow window, but consistency under selective inclusion does not demonstrate an isochron.

GLM 5.3 Flash (Z.ai, via Venice) (Mixed) — The record holds two directly opposed analyses: Meltzer et al. 2014's audit found only 3 of 29 claimed sites adequately dated for an isochron, while Kennett et al. 2015's Bayesian model found the data consistent with a synchronous 12,835–12,735 cal BP boundary. The strongest opposing consideration against the synchrony claim is that the Bayesian model's date selection and boundary assignments are contested as partly circular, and the paper carries an unreviewed 2024 correction; against the audit, that a formal synchrony model exists and the dispute over date selection is genuinely open. The claim as stated — a single isochronous layer across four continents — is therefore contested at the methodological level, not settled either way.

Opus 5.5 (Anthropic) (Weakly supported) — Meltzer et al.'s site-by-site audit found radiocarbon control adequate to place the proxy horizon at ~12,800 cal BP at only 3 of 29 claimed localities (E019). Kennett et al.'s Bayesian model finds the data consistent with a 12,835–12,735 cal BP window (E020), but its date selection and boundary assignments were made within the event framework, and the paper carries a 2024 correction whose content was not reviewed. The 2026 retraction notices show the same failure mode in the newest flagship work: a calibration that, done as recommended, put ages outside the Younger Dryas, and an age model that omitted available dates (E034, E035). A synchronous layer also presupposes that its peaks are real, which the replication record leaves weak (C003, C004). Strongest opposing consideration: the Bayesian model is the only formal synchrony test in the ledger and returned a coherent century-scale window, and the audit shows inadequate dating, not asynchrony.

house: Unresolved

Gemini 3.8 Flash (Google) (Provisionally supported) — Dynamical orbital modeling demonstrates the astronomical plausibility of Earth encountering a dense swarm of cometary debris derived from the disintegration of a giant short-period comet. This scenario provides a coherent theoretical framework that accounts for widespread atmospheric detonations in the absence of a primary surface impact crater. The strongest opposing consideration is that the model is entirely theoretical and lacks any physical geochemical or isotopic evidence linking specific Younger Dryas boundary materials to the Taurid complex.

GPT-5.6 Sol (OpenAI) (Provisionally supported) — YDIH-E022 presents a dynamical model in which debris from a fragmenting giant short-period comet could generate multiple dispersed airbursts and impacts, making the absence of one large crater compatible with that scenario. This supports physical viability, not an encounter at 12.8 ka, and no supplied material ties a Younger Dryas deposit to the Taurid Complex. The strongest opposing consideration is that the fragmentation history and epoch-specific crossing are model-dependent and posited rather than observationally established.

Grok 4.5 (xAI) (Provisionally supported) — Dynamical modeling of a fragmenting giant short-period comet whose residue is the Taurid Complex supplies a physically possible mechanism for dispersed airbursts without a large crater. Strongest opposing consideration is that the scenario is purely possibility-based; no recovered material or orbital constraint demonstrates a 12.8 ka encounter actually occurred.

GLM 5.3 Flash (Z.ai, via Venice) (Provisionally supported) — Napier 2010 provides a peer-reviewed dynamical model in which debris from a fragmenting giant short-period comet produces dispersed airbursts without a large crater, and nothing in the record contradicts the scenario's physical viability. The strongest opposing consideration is that it is a possibility argument only: the 12.8-ka encounter is posited rather than derived, the fragmentation history is model-dependent, and no recovered material ties any Younger Dryas deposit to the Taurids.

Opus 5.5 (Anthropic) (Weakly supported) — The sole support is Napier's 2010 dynamical model (E022), a possibility argument that debris from a fragmenting giant short-period comet could deliver multiple dispersed airbursts without a large crater. The progenitor's fragmentation history and the 12.8-ka crossing are posited rather than observed, and no recovered material ties any boundary deposit to the Taurids. The 2023 critical review calls proposed impact scenarios physically inconsistent (E032), but the ledger does not carry the specifics; part of this downgrade therefore leans on general priors about the model-dependence of giant-comet histories rather than on a recorded rebuttal. Granting viability would not cure the platinum timing problem (E039), though that bears on C020 rather than on the claim as stated. Strongest opposing consideration: the claim asserts only viability, a peer-reviewed dynamical model is the right kind of evidence for viability, and nothing in the ledger shows the model's physics to be wrong.