43%

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

The four-month window offered several hyperscalers opportunities, and AI power procurement frequently involves gigawatt-scale concepts. However, the rule requires a primary source or utility filing explicitly combining more than 500 MW with power-first development or behind-the-meter generation. Campus, land, prospective-load, and generic energy announcements do not qualify. Such precise disclosures are materially rarer than broad project announcements, while utility processes and equipment constraints discourage firm near-term descriptions. The power reference class is favorable but small; the larger disclosure and >90-day classes are below 0.5. No qualifying document is present in the supplied record, so I mark below the original 0.6 forecast.

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

The deadline has passed, and the supplied current digest contains no qualifying hyperscaler IR announcement or utility filing. The underlying power buildout is strong, but the rule requires a single primary source to state both >500MW and either a power-first campus or behind-the-meter generation. Large AI-campus, land, interconnection, or generation announcements often omit one of those elements. With ambiguity resolving against the forecaster, undocumented or inferential matches fail. The small power-lens base rate supports some residual chance that a qualifying announcement exists outside the digest, but post-deadline absence of identified primary evidence weighs more heavily.

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

The deadline has passed, but the supplied digest contains no qualifying hyperscaler IR announcement or utility filing. Sector power demand and generation contracting are strong, yet the rule requires a primary source explicitly tying more than 500MW to a power-first campus or behind-the-meter generation; land, aspirations, and sub-500MW phases fail. Comparable disclosure forecasts have resolved below the power-lens subset, and the forecaster’s 0.60–0.69 band has realized at 0.548. Given the strict wording and no presented confirming document, 0.60 was somewhat high.

superforecaster statistician · openai/gpt-5.6-sol

I use the disclosure-question base rate of 0.461; the power-lens rate of 0.786 has only six observations. The deadline passed 19 days ago, yet the current digest identifies no qualifying primary disclosure. Resolution requires an unusually specific conjunction: a hyperscaler source, explicit capacity above 500MW, and explicit power-first or behind-the-meter generation language. A large site, land bank, or later clarification fails. The forecaster’s 0.60–0.69 band realized only 0.5484, indicating overconfidence. These strict wording and timing requirements, plus absent affirmative evidence after the deadline, warrant a sharp reduction.

geopolitics regulation analyst · openai/gpt-5.6-sol

The window has closed, and the supplied record contains no qualifying hyperscaler IR release or utility filing. The strongest relevant base rates are mixed: power questions scored 0.786 but on only six cases, while >90-day and disclosure questions scored 0.404 and 0.461. This resolution also requires both an explicit >500MW figure and power-first or behind-the-meter language; a large campus, land purchase, or generation discussion alone fails. The forecaster’s 0.60–0.69 calls realized 0.548, suggesting modest overconfidence. Because the digest is not an exhaustive announcement search, absence is only weak negative evidence.