AI Newsletter

John Jumper, AlphaFold inventor and Nobel laureate, is leaving Google DeepMind and joining Anthropic – the third high-profile departure in a short time. In parallel, multiple Reddit threads report that DeepMind is struggling to keep pace with Anthropic and OpenAI, while Gemini 3.5 Pro is considered no decisive leap. This brain drain is a strategic alarm signal for Google's competitiveness in the frontier AI race.

A viral Reddit thread warns: AI services are currently heavily cross-subsidized by VC capital – analogous to early Uber – and hardly anyone has priced in what happens when prices adjust. The most-liked comment points to local models (Qwen3.6) as a hedge, while others point to massive, unstructured AI usage ('we'll see what sticks'). For companies building their cost calculations on today's API prices, a significant business model risk emerges.

According to Grok sources, Andrej Karpathy bolstered Anthropic's pre-training team in May after recently commenting publicly on Claude Fable 5 and Mythos. Combined with Jumper's departure from DeepMind and the Karpathy move, a massive talent pull toward Anthropic is emerging, which should fundamentally reorder the balance of power in frontier models in the medium term. Polymarket currently prices Anthropic at 95% probability as the leading model by end of June.

Official US Census data (as of May 2026) show: Only 37% of firms with 250+ employees operationally deploy AI, for mid-size companies it's 32%. In parallel, an MIT analysis confirms that around 95% of enterprise GenAI pilots generate zero measurable P&L impact – the gap between demo culture and actual deployment is the central unsolved problem. This opens a massive market window for providers of productivity and agent tools.

Immediately after the US export control ban against Anthropic's Fable 5/Mythos, Mistral CEO Arthur Mensch positions his company offensively as a state-independent open-source alternative: 'We exist outside state control.' The incident catapulted Mistral into public focus and underscores that regulatory interventions in proprietary frontier models directly fuel demand for open models. A new geopolitical differentiation factor in the model market is emerging.