An unreleased Anthropic Claude model drove recent trader attention by raising the proven lower bound on nontrivial zeros of the Riemann zeta function lying on the critical line from 41.6% to 67.2% in August 2026. The large language model synthesized prior work from Bombieri, Baluyot, Goldston and others, coordinated dozens of subagents over multiple days, and produced a Lean-verified proof validated by external number theorists. Follow-up human papers in September pushed related bounds higher, including over 83% distinct zeros, while unverified preprints claiming full proofs continue to surface. These AI-assisted gains highlight accelerating machine capabilities in number theory without resolving the hypothesis itself, keeping focus on upcoming formal verifications, conference announcements, and any new model releases that could extend the approach.