38%

grid and power economist · openai/gpt-5.6-sol

The relevant software and disclosure base rates are only 0.446 and 0.461. This question is harder than either class because the deadline has passed and success requires two distinct frontier labs, each with a publicly documented, separately priced SKU—not merely a cyber-capable model, pilot, integration, or research preview. The supplied digest contains no qualifying pricing page, blog disclosure, or SEC filing. Given the ambiguity-against-forecaster rule, plausible but loosely branded security tools would not suffice. The forecaster’s 0.6 also exceeds the realized 0.548 rate for its confidence band.

semiconductor supply chain analyst · openai/gpt-5.6-sol

The criterion is stricter than general cyber-model availability: two distinct frontier labs must each have a cyber-specific SKU, gated access, separate pricing, and qualifying public documentation dated by September 30. Labs commonly announce cyber capabilities or private previews without publishing standalone pricing, while security features are often bundled into enterprise or developer plans. The supplied digest contains no evidence of even one qualifying disclosure. Given the software/disclosure base rates, the two-lab requirement, and ambiguity resolving against the forecaster, 0.6 was too high.

capital markets credit analyst · openai/gpt-5.6-sol

The deadline has passed, and the supplied public digest contains no qualifying pricing page, blog, or filing. A literal hit requires two independent frontier labs each to document a cyber-specific product, gated access, and separate commercial packaging by September 30. Labs commonly announce cyber agents as research previews, private pilots, or features bundled into enterprise/API plans; those would fail. One qualifying launch was plausible, but obtaining a second with explicit SKU or pricing within five months was materially harder. The relevant software, disclosure, and long-horizon base rates cluster around 0.40–0.46; the two-lab threshold and ambiguity policy justify a discount below them.

superforecaster statistician · openai/gpt-5.6-sol

I use the software-ledger base rate of 0.446, cross-checked against disclosure (0.461) and >90-day (0.404) rates. I adjust downward because resolution requires two labs, not one, and each must satisfy three strict elements: cyber-specific, access-gated, and separately priced/SKU’d in an approved primary source by September 30. Security agents, limited previews, or cyber capabilities bundled into general enterprise/API pricing would fail. The supplied digest contains no qualifying disclosure evidence. The forecaster’s 0.60–0.69 calibration band realized only 0.548, supporting an overconfidence correction, though the strict two-lab conjunction drives most of the reduction.

geopolitics regulation analyst · openai/gpt-5.6-sol

The strict criterion requires two distinct frontier labs, each with a public, separately priced SKU that is also gated and explicitly cyber-focused by September 30. Labs commonly disclose cyber safeguards, controlled-access programs, or security agents without giving them standalone pricing; general enterprise/API pricing would not qualify. Requiring two qualifying disclosures compounds this documentation risk, and ambiguity resolves against a hit. The software, disclosure, and >90-day ledger base rates cluster around 0.40–0.46, below the original 0.60 forecast. The supplied digest provides no direct evidence of either qualifying launch.