{"schema":"sb.truth.v1","subject":"ai-model-race","title":"The AI model race record","tier":"standard","tierNote":"Standard-tier record, generated from the maintained event registry: reporting, our coverage metric, market expectation, and state history. No primary-source layer yet — first-party issuer verification is what upgrades a record to full tier, the discipline the hormuz record demonstrates.","subjectBinding":{"entity":"ai","bound":true,"basis":"named-in-subject","statement":"Lead subject \"ai\" — named in this record's own subject line (\"The AI model race\"). The reporting, record, market and state-history layers below are that subject's own; none is borrowed from a neighbouring entity. Not bound: china — the registry attaches it to this event, but the subject line does not name it.","candidatesConsidered":["ai","anthropic","china","deepseek","openai"],"alsoDefensible":["anthropic","deepseek","openai"],"rejected":["china"],"actionClass":null},"publishState":"published","publishStateNote":null,"asOf":"2026-09-04","generatedAt":"2026-09-04T23:47:27.001318+00:00","tracked":true,"access":"Free — no key, no charge. Serves at GET /api/truth/ai-model-race and answers the get_truth_object MCP tool. (No exhibit page for standard-tier records yet — the API and MCP responses ARE the record.)","attribution":"Cite 'Signal Bureau' with https://sigb.ai/api/truth/ai-model-race so readers can verify.","layers":{"primarySources":{"label":"Observed operational state","status":"absent","basis":"not present on a standard-tier record — no first-party issuer has been routed for this subject yet; nothing here is previewed to look finished","sources":[]},"reporting":{"label":"Observed reporting","basis":"attributed sources — outside coverage in our tracked set; verify at the source. Prediction-market question pages are excluded: a venue question is not journalism, whoever hosts it. Filtered for EVENT relevance, not just subject: a row is served only if it names \"ai\" AND carries a term from this record's own subject line, or two terms from its market-question set. The subject's name alone is the key its corpus was sampled on, so it proves nothing about this event — each served row names the terms it matched in eventTerms.","status":"observed","note":"Event-relevance filtered: 11 attributable rows sampled for \"ai\", 2 served, 9 withheld as not shown to be about \"The AI model race\". The served count is reduced on purpose and reported, never quietly smaller.","sampledForSubject":11,"served":2,"withheldNotEventRelevant":9,"relevanceTerms":{"subjectLine":["model","curated","benchmark","anthropic","leapfrog","chatgpt","claude","fable","deepseek","frontier","gemini","gpt","large","language","release","openai","weights"],"questionSet":["announces","achieved","agi","december","best","end","launch","consumer","hardware","product","released","debut","score","least","1500"]},"items":[{"title":"Nvidia is buying AI startup Hugging Face that was hacked by OpenAI models for nearly $13 billion - ABC7 Los Angeles","url":"https://news.google.com/rss/articles/CBMirgFBVV95cUxQUzVvVnVzY1ktWFYxaFBfd2tERW5qcFR5Q05oVWFUZV9TTHBNaFl1NkxnS3hLcy1TWjNadjBwZE9BY3c5Zl9tMnphSy05RGl4QWZYdERERWk2Yno4Yy1BdVVRc09icXBSbUNUNGZPdGp4am5kTU5GUmtqNUk3Zk4xUVEwNmUxM1pqSVpfOG5DeGxlazlwMGJDRHpYUlZDTEt3a3I2Q3V5dnFvMlhCSEE?oc=5&hl=en-US&gl=US&ceid=US:en","date":null,"source":null,"eventTerms":["model","openai"],"relevance":"subject-line-term","tier":"social/aggregator","kind":"reporting — attributed; verify at the source"},{"title":"OpenAI has announced GPT-6 Astra, says the 'AGI era' has arrived; Brings major leap in AI reasoning, coding and computer use - Business Today","url":"https://news.google.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?oc=5&hl=en-US&gl=US&ceid=US:en","date":null,"source":null,"eventTerms":["leapfrog","gpt","openai"],"relevance":"subject-line-term","tier":"social/aggregator","kind":"reporting — attributed; verify at the source"}]},"record":{"label":"The record itself","basis":"our own count of documents we ingested and can name — not a claim about the world, and market questions are excluded because a question is not reporting","documents":6291,"outlets":704,"activeDays":176,"firstSeen":"2026-02-25","firstSeenStatus":"observed","firstSeenNote":null,"lastSeen":null,"lastSeenStatus":"withheld-future-dated","lastSeenNote":"unmeasured — the lastSeen stamp on this subject read 2026-11-01, which is after today (2026-09-05). A document we have not reached yet cannot be one we have seen, so the stamp is withheld rather than clamped: a clamped date would be a fabricated observation. The cause is a forward-dated publisher field (a grant's period-of-performance start, a journal's ahead-of-print issue date) reaching our \"when we saw it\" axis. This is neither a real date nor a measured zero.","homeDomains":["auto"],"asserts":null,"assertionCoverage":0.191,"movesWith":[{"name":"nvidia","together":566,"sharedDomains":["auto"],"verbs":{"acquisition":4,"meeting-visit":2}},{"name":"china","together":555,"sharedDomains":["auto"],"verbs":{"meeting-visit":5,"acquisition":2}},{"name":"abstract","together":545,"sharedDomains":["auto"]},{"name":"computer science","together":505,"sharedDomains":["auto"]},{"name":"openai","together":985,"verbs":{"acquisition":3}},{"name":"view pdf html","together":440,"sharedDomains":["auto"]},{"name":"meta","together":427,"sharedDomains":["auto"]},{"name":"anthropic","together":911,"verbs":{"acquisition":8,"meeting-visit":2}}],"movesWithLabels":["nvidia","china","abstract","computer science","openai","view pdf html","meta","anthropic"],"provenance":"curated"},"coverageMetric":{"label":"Coverage state","basis":"our own measurement — our attention, not the event's state","direction":"steady","baselineRelativePct":6,"reading":"Our coverage of The AI model race is 6% above its 3-week baseline. This measures our attention — how much our tracked set is writing about it — not the event itself.","trackedDays":118,"eventMentions":17969,"periods":[{"period":"2028","mentions":14},{"period":"2027-12","mentions":997},{"period":"2027-11","mentions":10},{"period":"2027-10","mentions":10},{"period":"2027-09","mentions":52},{"period":"2027-08","mentions":10},{"period":"2027-07","mentions":10},{"period":"2027-06","mentions":101}]},"market":{"label":"Market expectation","basis":"attributed — markets price collective belief about defined outcomes; belief, not fact, and never proof that an event occurred","rowsStatus":"none-captured","rowsNote":"Measured zero: no priced capture rides \"ai\" in this build.","current":[],"arc":[],"eventArc":{"name":"The AI model race","slug":"ai-model-race","questionCount":76,"questions":["OpenAI announces it has achieved AGI before 2027?","Will DeepSeek have a #1 AI model by December 31, 2026?","Will DeepSeek have the best AI model at the end of December 2026?","Will OpenAI launch a new consumer hardware product by December 31, 2026?","Will OpenAI have a #1 AI model by December 31, 2026?","Will the next GPT model released by OpenAI debut at a score of at least 1500 by December 31, 2026?"],"mentions":17969,"firstSeen":"2026-03-29","firstSeenStatus":"observed","lastSeen":"2026-09-04","lastSeenStatus":"observed","stampNote":null}},"judgment":{"label":"Our reading","basis":"derived judgment — our reading over the layers above; ours, and labeled ours","asOf":"2026-09-04","text":"A standard-tier record keeps its layers apart and does not collapse them. Lead subject \"ai\" — named in this record's own subject line (\"The AI model race\"). The reporting, record, market and state-history layers below are that subject's own; none is borrowed from a neighbouring entity. Not bound: china — the registry attaches it to this event, but the subject line does not name it. Our coverage metric measures our attention; the reporting is attributed, not verified against first-party issuers; the market layer prices belief, not fact. Reading any single layer as \"the answer\" is how a machine gets this wrong — and this record's honest limit is stated in its tier: no primary-source layer exists here yet."},"consensus":{"label":"Reporting consensus","basis":"retroactive claims ledger — kernel facts scored by independent-chain agreement","pending":true,"note":"No claims ledger has been computed for this subject yet — consensus scoring is full-tier machinery.","retired":[]},"stateHistory":{"label":"State history","basis":"recorded transitions — never edited after the fact","status":"recorded","note":null,"transitions":[{"field":"coverage state","from":"steady","to":"fading","at":"2026-09-01T15:11:18.454578+00:00","heldSince":"2026-08-30T15:08:25.442508+00:00"},{"field":"coverage state","from":"fading","to":"steady","at":"2026-09-02T15:12:31.676793+00:00","heldSince":"2026-09-01T15:11:18.454578+00:00"}]}}}