
Anthropic mathematician claims he used Claude to prove 19th-century Riemann hypothesis
THE SO WHAT
Whether or not this Riemann claim survives scrutiny, using Claude over six weeks as a collaborator on frontier math is a preview of how AI will enter domains that once felt verification-limited. If you run research-heavy teams, the question isn’t “can the model solve it” but “how do we structure review, credit, and IP when it helps attack problems at this level.”
READ THE SOURCE
MORE FROM THE WIRE
Applied AICursor may drop Chinese AI models as India’s Sarvam starts selling them
Model provenance is becoming a board-level geopolitical risk variable, not just a quality or cost choice. If your core product depends on foreign open-weight models, you need a contingency plan for abrupt deprecation driven by regulators or internal risk committees, not just by performance.
Applied AIWhy do math when AI can?
If AI handles more of the grind, the edge shifts to people who can pose the right problems and sanity-check the answers. For technical teams, that means training for mathematical taste and verification, not just tool fluency.
Applied AIChina’s tech giants double AI spending to $139bn as labs earn less than 8%
China’s big platforms are pouring $139B into AI capex while labs contribute under 8% of that revenue—this is a long-duration infrastructure bet, not a near-term P&L story. Western operators should assume a world where Chinese AI capacity is abundant but monetization models are distorted by strategic, not purely commercial, incentives.
Applied AIIs selling an AI version of your face to this microdrama startup the future of Hollywood?
Licensing your face as an AI asset turns performers into perpetual IP portfolios and compresses the marginal cost of content. Studios and platforms will need new contract templates, residual structures, and risk controls around identity misuse long before this hits true scale.