According to @nic__carter, multiple sections in Dwarkesh’s book assert that models cannot do certain tasks, while footnotes note that by publication time the models already can, signaling rapid capability progression and shrinking research-to-reality timelines (source: @nic__carter on X, Oct 17, 2025). According to @nic__carter, traders should re-validate AI-dependent theses and backtests against current model performance to avoid stale assumptions when positioning in AI-exposed equities and AI-linked crypto narratives (source: @nic__carter on X, Oct 17, 2025).