33%

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

The deadline has passed, but the supplied digest contains no qualifying model card, lab post, or AMD disclosure. AMD’s rapid investment supports greater accelerator adoption, yet it does not establish a second named frontier-class open-weights model trained end-to-end on AMD. The strict wording creates several failure modes: mixed-vendor training, inference-only AMD use, a model below frontier class, or disclosure that omits the training cluster. Relevant disclosure and longer-horizon reference rates are below one-half, and the forecaster’s 0.50–0.59 band has realized only 0.375 outcomes. Ambiguity resolving against the forecaster warrants a substantial discount from 0.55.

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

The deadline has passed, and the supplied public digest contains no qualifying model card, lab post, or AMD disclosure. The rule is unusually strict: this must be a second named open-weights frontier-class model, with explicit end-to-end AMD training—not inference, fine-tuning, partial training, or generic cluster access. AMD’s elevated capex supports ecosystem progress but does not establish the required disclosure. Disclosure questions and the forecaster’s 0.50–0.59 band also historically resolved below the stated confidence. Ambiguity resolving against the forecaster further lowers the chance of a hit.

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

The deadline has passed, and the supplied record contains no qualifying model card, lab post, or AMD disclosure. This is a strict conjunction: a second named model, open weights, frontier-class, end-to-end AMD training, and a named cluster. Explicit end-to-end hardware provenance is uncommon, while “frontier-class” creates additional ambiguity that resolves against the forecast. Relevant long-horizon and disclosure base rates are below 0.5, and the forecaster’s 0.50–0.59 band has realized only 0.3.

superforecaster statistician · openai/gpt-5.6-sol

I use the >90-day base rate of 0.404, cross-checked against the 0.461 disclosure rate. The forecaster’s 0.50–0.59 band realized only 0.30, indicating substantial overconfidence. This resolution additionally requires a second named model, open weights, frontier-class status, end-to-end AMD training, and a qualifying primary source by a hard deadline. The deadline has passed, and the supplied evidence contains no such disclosure; AMD capital spending does not establish any model’s training provenance. Ambiguity also resolves against a hit, leaving only residual probability that a qualifying announcement exists but was omitted.

geopolitics regulation analyst · openai/gpt-5.6-sol

The deadline has passed, but the supplied digest contains no qualifying model-card, lab-blog, or AMD disclosure. The hardware reference class is favorable, yet the more relevant long-horizon and disclosure base rates are only 0.404 and 0.461, while this forecaster’s 0.50–0.59 calls realized at 0.30. The resolution rule is unusually strict: a second named model must be open-weights, plausibly frontier-class, explicitly trained end-to-end on AMD, and documented by September 30. AMD adoption could have produced another model, but mixed-cluster training, inference-only use, unclear frontier status, or late disclosure all resolve as misses.