Verify a decision

{"actor":"judge","decided_at":"2026-08-27T06:09:23.158Z","decision":"review_revise","investigation_id":"15c90e9f-a964-4aa4-8358-196a48fbbd4d","new_score":48,"page_slug":"ambient-finance","prev_score":58,"reason":"Of 24 fact-checked claims, 5 were disputed and 4 partially supported, giving a disputed rate of 20.8% (5/24), which falls in the moderate-minor band. The most notable defect (claim_findings[12]) is that the page's sentence deriving a 'net gain of roughly $110,600' from 50 ETH seeded and 83.72 ETH withdrawn does not follow arithmetically (83.72 - 50 = 33.72 ETH, ~$56,700) -- but the $110,600 total-loss figure itself is independently corroborated elsewhere on the page (claim_findings[11]), so this is a flawed derivation sentence rather than a wrong headline number. Separately, the page understates Ambient's cumulative trading volume by roughly 10x against its own cited DefiLlama source (claim_findings[15]), and three timeline dates around the 2023 mainnet launch and the October 2024 DNS-hijack recovery are off by days to weeks against contemporaneous or primary sources (claim_findings[20], [22], [23]) without changing the substance of what happened. Weighing in the page's favor, it avoided two of the most common defects in this review process: it did not assert the single-contract design as a confirmed cause of the exploits (claim_findings[13]) and it correctly avoided treating an investor relationship (Quantstamp) or an automated scoring tool (CertiK Skynet) as a real audit (claim_findings[16]). No citations were found to be dead or stale. A high-priority coverage gap was flagged: the page never examines who controls admin/upgrade keys on the single contract, which is directly relevant to its own stated risk thesis and should be addressed even though it does not itself constitute a factual error.","score_delta":-10,"sequence_num":3,"submission_content_hash":null,"submission_id":null,"submission_kind":null,"submission_valence":null,"v":1}