Erebus

The Origins of SARS-CoV-2 · claim COV-C001 · Observation rung · Early cases and geography

The Huanan Seafood Wholesale Market was the geographic center of the earliest documented cluster of SARS-CoV-2 infections in Wuhan in December 2019.

The earliest known cases cluster tightly around one market. This is the sturdiest finding in the case, and both sides now largely accept it — what they dispute is what it means.

AI-extractedorigin: casework/covid-origins digest — C001

credibility — is the claim itself true?

Well supported

Multiple channels converge: the spatial analysis of December 2019 cases, the concentration of positive environmental samples inside the market, and the Chinese CDC's own surveillance data placing 87.5 percent of interior positives in the market's west zone. The WHO scientific advisory group calls a zoonotic origin the best supported hypothesis on accessible evidence while declining to conclude the market was the first spillover site rather than a site of amplification.

diagnosticity — how much does it decide the thesis?

moderate

An epicenter is predicted by market spillover, by upstream spillover with subsequent market arrival, and by amplification of a virus already circulating in humans. It discriminates against non-market origins collectively, not among the hypotheses that matter most here.

Evidence

Supports (2)

Supports· strongCOV-E001

Early cases clustered around the market, including cases with no market link

what the source states
"We considered three categories of cases, which were all significantly closer to the Huanan market than expected: (i) all cases (median distance 4.28 km; P < 0.001), (ii) cases linked directly to the Huanan market (median distance 5.74 km; P < 0.001), and (iii) cases with no evidence of a direct link to the Huanan market (median distance 4.00 km; P < 0.001)." And, from the Discussion: "this group of cases resided significantly closer to the market than those who worked there, indicating that they had been exposed to the virus at or near the Huanan market. For market workers, the exposure risk was their place of work, not their residential locations, which were significantly farther afield than those cases not formally linked to the market."
what we infer
This is the observation that ascertainment bias has the hardest time explaining. If early cases were found because investigators looked around the market, the unlinked cases should not be the ones living nearest it. The comparison is significant at p = 0.029, which is a modest margin for a load-bearing result.
limitations
  • Rests on Chinese case ascertainment that has never been independently audited; the complete December 2019 line list has not been released.
  • Not independently reproduced; the code and raw geospatial data have not been re-run by a third party.
  • A published eLetter disputes the spatial result on distributional grounds and the authors replied identifying flaws in that critique; the exchange is unresolved.
Michael Worobey (2022)AI-verified citation
Supports· strongCOV-E021

WHO's conclusion, and its statement that the laboratory hypothesis could not be assessed

what the source states
"The weight of available evidence reviewed by SAGO suggests zoonotic spillover of SARS-CoV-2 into the human population, either directly from bats or through an intermediate host. However, SAGO cannot conclude with certainty where and when this occurred, nor if the HSM was indeed the first instance of spillover into the human population, or the site of further spillover and amplification." And: "Information and evidence is also lacking to assess the possibility of a laboratory origin - either the evidence is not available or has not been provided to the scientific community. As a result, SAGO has been unable to adequately assess this route for human infection and therefore is not in a position to rule this out as a possibility." On its own access: "It is important to note that SAGO did not have access to original raw data from any source in preparing this report."
what we infer
The two halves must always travel together. Note also the qualification "reviewed by SAGO," which ellipsis usually removes and which is the whole point: the weight of the evidence that could be seen. The report's own statement that it had no access to raw data from any source is the most important sentence in it for calibrating everything else.
limitations
  • The file WHO now serves is a post-publication corrected version, so page locators can differ from the June 2025 original; the version read is recorded in the source.
  • Not independent of the papers and assessments it reviews.

Undermines (1)

Undermines· moderateCOV-E003

The earliest reported case had no market link, per the first clinical series

what the source states
"By Jan 2, 2020, 41 admitted hospital patients had been identified as having laboratory-confirmed 2019-nCoV infection. … 27 (66%) of 41 patients had been exposed to Huanan seafood market." And: "The symptom onset date of the first patient identified was Dec 1, 2019. None of his family members developed fever or any respiratory symptoms. No epidemiological link was found between the first patient and later cases."
what we infer
The same paper carries evidence in both directions, which is why it is recorded once rather than split into a favorable half. Two-thirds of the earliest confirmed patients had direct market exposure — strong for the market's centrality — while the earliest case by onset had none, which is what an upstream or non-market index event predicts.
limitations
  • A single case's reported exposure history in an early clinical series is weak evidence about the outbreak's origin.
  • The identification of this patient as the earliest case, and the December 1 onset date, have been disputed in later reporting this case has not adjudicated.
Chaolin Huang (2020)AI-verified citation

strongest unresolved objection

The finding rests entirely on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released. Cases were identified by clinical review, but the review itself is not inspectable.

what would change our mind

  • Release of the complete December 2019 case line list showing a materially different geographic distribution.
  • Evidence that early case-finding was concentrated around the market for administrative reasons.

research that would move this claim

Assessment history

Append-only AI overlay records; the canon claim file is never mutated.

Well supportedconfidence: high · anthropic/claude-opus-5 (scheduled maintenance run) · run 2026-08-26-auto-a2ud · unreviewed draft

Convergent channels including the custodian's own surveillance data and SAGO's reading (COV-E004, COV-E021). The single unlinked earliest case in COV-E003 qualifies but does not overturn the geographic centring. Unchanged; no new material bears on it.

Well supportedconfidence: high · anthropic/claude-opus-5 (scheduled maintenance run) · run 2026-08-26-auto-kg4n · unreviewed draft

Multiple channels converge, including the custodian's surveillance data (COV-E004) and SAGO's reading (COV-E021). The single unlinked earliest case in COV-E003 qualifies but does not overturn the geographic centring.

Well supportedconfidence: high · anthropic/claude-opus-5 (scheduled maintenance run) · run 2026-08-26-auto-l4iy · unreviewed draft

Geographic centring of the earliest recognised cluster is corroborated by the custodian's own surveillance data and by SAGO's reading (COV-E004, COV-E021). Nothing new contests it; COV-E003's single unlinked earliest case qualifies rather than overturns it.

Well supportedconfidence: high · Gemini 3.1 Pro (Google) — independent check run via gemini-3.1-pro-preview · run 2026-08-26-check-gemini-r2 · unreviewed draft

The Huanan market's status as the early geographic epicenter is corroborated by independent spatial analysis, environmental sampling, and the Chinese CDC's surveillance data. Both major origin hypotheses now largely accept this premise. The strongest opposing consideration is that the underlying December 2019 clinical case list relies entirely on Chinese ascertainment data that has never been independently audited or fully released.

Well supportedconfidence: high · Gemini 3.1 Pro (Google) — independent check run via gemini-3.1-pro-preview · run 2026-08-26-check-gemini-r3 · unreviewed draft

Early case spatial analysis and the Chinese CDC's own environmental sampling strongly converge on the Huanan market as the geographic center of the outbreak. The clustering is corroborated by multiple analytical layers, including the heavy concentration of positive environmental swabs in the west zone. The strongest opposing consideration is that the underlying line list of December 2019 cases has never been released or independently audited for ascertainment bias.

Well supportedconfidence: high · Gemini 3.1 Pro (Google) — independent check run via gemini-3.1-pro-preview · run 2026-08-26-check-gemini · unreviewed draft

Convergence of spatial case data, environmental positives, and internal CDC data firmly establishes the market as the geographic epicenter of early cases. The WHO SAGO report concurs on this geographic centrality based on accessible data. The strongest opposing consideration is that this spatial conclusion rests entirely on early Chinese case ascertainment methodology and line lists that have never been independently audited or fully released.

Well supportedconfidence: high · GPT-5.1 (OpenAI) — independent check run via gpt-5.1 · run 2026-08-26-check-gpt-r2 · unreviewed draft

Multiple independent evidence records (COV-E001, COV-E003, COV-E004, COV-E021) consistently show that the earliest recognized December 2019 cases and the bulk of positive environmental samples center on the Huanan market, particularly its west section. The strongest opposing consideration is that all of this rests on Chinese case ascertainment and line lists that have not been independently audited, so systematic biases in early case-finding cannot be entirely excluded.

Establishedconfidence: high · GPT-5.1 (OpenAI) — independent check run via gpt-5.1 · run 2026-08-26-check-gpt-r3 · unreviewed draft

Multiple independent lines of evidence from early case mapping and environmental sampling (COV-E001, COV-E003, COV-E004, COV-E021) all show December 2019 infections clustering around the Huanan market, and no evidence record suggests an alternative epicenter. The strongest objection is that these analyses rest entirely on Chinese case ascertainment and an unreleased December 2019 line list, so hidden ascertainment bias cannot be excluded, but the convergence of geography, market-linked exposures, and internal positivity makes a different early center highly implausible on the available record.

Well supportedconfidence: high · GPT-5.1 (OpenAI) — independent check run via gpt-5.1 · run 2026-08-26-check-gpt · unreviewed draft

Multiple independent strands—Worobey’s early‑case spatial analysis, the Chinese CDC environmental sampling, and WHO SAGO’s synthesis—all align in placing the earliest recognized cluster around the Huanan market. The strongest opposing consideration is that all of this rests on Chinese case ascertainment and line‑list data that have never been independently audited, leaving open the possibility of systematic bias or omission.

Well supportedconfidence: high · Grok 4.6 (xAI) — independent check run via grok-4.6 · run 2026-08-26-check-grok-r2 · unreviewed draft

Multiple channels place the earliest documented December 2019 cluster on the Huanan market: Worobey 2022 finds all three case categories significantly closer to it than expected; Huang 2020 reports that 27 of 41 early confirmed patients had market exposure; Liu 2024 records dense environmental positivity after closure. SAGO 2025 treats the market as central while declining to say it was the first spillover rather than an amplification site, which is a limit on interpretation rather than on the geographic observation as stated. The strongest opposing consideration is that the finding rests entirely on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released.

Well supportedconfidence: high · Grok 4.6 (xAI) — independent check run via grok-4.6 · run 2026-08-26-check-grok-r3 · unreviewed draft

Worobey et al. report that all three early-case categories, including cases with no market link, resided significantly closer to the Huanan market than expected, and Huang et al. report that 27 of 41 of the first confirmed patients had been exposed there. Liu et al. separately place 87.5 percent of interior environmental positives in the west zone, so geography, clinical exposure fractions, and environmental RNA all point at the same site. The claim as stated is about the center of the earliest documented cluster, not about the first spillover, and on that narrower wording the record is consistent. Strongest opposing consideration: the finding rests entirely on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released.

Well supportedconfidence: high · Grok 4.6 (xAI) — independent check run via grok-4.6 · run 2026-08-26-check-grok · unreviewed draft

Multiple independent analytic layers on the same underlying Chinese data converge: December 2019 cases were significantly closer to the market than expected, two-thirds of the first published clinical series had direct market exposure, and environmental positives concentrate inside the market. WHO's 2025 advisory group treats zoonotic spillover as the best-supported hypothesis on reviewed evidence while declining to conclude that the market was the first spillover rather than an amplification site. The strongest opposing consideration is that the finding rests entirely on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released.

Well supportedconfidence: high · Kimi K3 (Moonshot, via Venice) — independent check run via kimi-k3 · run 2026-08-26-check-kimi-r2 · unreviewed draft

Three channels converge on the market as the geographic center of the earliest documented cluster: the spatial analysis of December 2019 cases, the interior concentration of positive environmental samples, and the Chinese CDC's own surveillance data placing 87.5 percent of interior positives in the west zone. As stated — geographic center of the earliest documented cluster, not site of spillover — the claim is modest and multiply supported. The strongest opposing consideration is that every channel rests on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released, so the observation is one custodial act away from revision. That qualifies confidence in permanence, not the current verdict.

Well supportedconfidence: high · Kimi K3 (Moonshot, via Venice) — independent check run via kimi-k3 · run 2026-08-26-check-kimi-r3 · unreviewed draft

Three channels converge: the published spatial analysis of December 2019 cases, the Chinese CDC's own environmental data placing 87.5 percent of interior positives in the west zone, and WHO SAGO's synthesis — and the claim is carefully scoped to the earliest *documented* cluster, which both camps largely accept. No competing geographic account exists anywhere in the record. The strongest opposing consideration is that the entire case set rests on Chinese ascertainment that has never been independently audited, with the December line list unreleased, so "documented" could in principle encode a search pattern; that possibility caps the grade at well_supported rather than established.

Well supportedconfidence: high · Kimi K3 (Moonshot, via Venice) — independent check run via kimi-k3 · run 2026-08-26-check-kimi · unreviewed draft

The geographic concentration of the earliest documented cases around the market is supported by the published spatial analysis, the Chinese CDC's own environmental data, and is accepted in substance by both camps — the dispute is over interpretation, not the fact. The strongest opposing consideration is that the entire finding rests on Chinese case ascertainment that has never been independently audited, and the complete December 2019 line list has never been released. That is a provenance caveat, not a contradiction in the record; nothing provided shows a different distribution. I agree with the dossier's grade.

Well supportedconfidence: moderate · Opus 5 (Anthropic) — independent check run via claude-opus-5 · run 2026-08-26-check-opus-r2 · unreviewed draft

Three channels agree that the earliest documented cluster centres on the market: the spatial analysis of December 2019 cases, the Chinese CDC's environmental positivity distribution, and the first clinical series in which 27 of 41 patients had direct market exposure. The strongest opposing consideration is that all three inherit the same case-ascertainment process, which has never been independently audited and whose complete December 2019 line list has never been released — so a review that looked hardest around the market would produce a similar picture. Confidence is moderate rather than high for that reason, not because any contrary data has been produced.

Well supportedconfidence: high · Opus 5 (Anthropic) — independent check run via claude-opus-5 · run 2026-08-26-check-opus-r3 · unreviewed draft

Three partly separable channels converge: the spatial analysis of December cases, the Chinese CDC's own environmental positivity inside the market, and SAGO's synthesis. The claim is narrowly worded — the centre of the *documented* cluster — which makes it close to self-verifying on the available case set. The strongest opposing consideration is that the case set itself is a product of Chinese ascertainment applied with a case definition that at points included market exposure, and the complete December 2019 line list has never been released, so the documented cluster could be partly an artefact of where investigators looked. That objection is blunted, though not removed, because the environmental positivity is not case-derived and points to the same site.

Well supportedconfidence: moderate · Opus 5 (Anthropic) — independent check run via claude-opus-5 · run 2026-08-26-check-opus · unreviewed draft

Three channels converge — the spatial analysis of December 2019 cases, 74 of 923 environmental positives inside the market, and two-thirds of the first 41 confirmed patients reporting market exposure — and both camps largely concede the point. The strongest opposing consideration is that all three channels depend on Chinese case ascertainment that has never been audited and a December 2019 line list that has never been released; if early case-finding was itself organised around the market, the cluster is partly self-generated. That concern is real but does not fully dissolve the finding, because the environmental positivity and the unlinked-case geography are not products of the same selection mechanism. Confidence is held at moderate rather than high solely because of the single-custodian dependency.

Well supportedconfidence: high · Claude Fable 5 (Cursor cloud agent) · run 2026-08-26-fable-covid-1 · unreviewed draft

Multiple channels converge, including the custodian's own surveillance data. Rests on case ascertainment that has never been independently audited.

Well supportedconfidence: high · anthropic/claude-opus-5 (reconsideration — deliberately non-blind: engages the panel's dissents) · run 2026-08-26-reconsider-m39k · unreviewed draft

Multiple channels converge, including the custodian's own surveillance data and SAGO's reading (COV-E021). Held against no dissent; the unaudited ascertainment caveat is retained but does not reach the geographic centring itself, which COV-E003's single unlinked earliest case does not overturn.