Nobel laureate John Jumper leaves DeepMind for Anthropic, plus open models and AI tutors deliver real wins

John Jumper, who shared the 2024 Nobel Prize in Chemistry for creating AlphaFold, left Google DeepMind after nearly nine years to join Anthropic. This move signals where serious AI-for-science talent is heading. Anthropic also launched Claude Science, a research workbench that orchestrates across scientific databases and runs on a lab's own infrastructure, alongside grants for up to 50 research projects. Meanwhile, the open-weight frontier saw two major moves: Tencent released Hunyuan Hy3, a 295-billion-parameter model with 21 billion active parameters, under the permissive Apache 2.0 license—making it viable for startups and labs worldwide. Mistral's CEO confirmed a new open-weight model coming this summer, with early access in July, as the French lab's ARR surpassed $400 million.

Beyond models, AI delivered tangible real-world results. A randomized trial published in Nature's Scientific Reports found an AI tutor outperformed in-class active learning, providing the strongest evidence yet that a well-designed AI tutor can beat traditional lectures. Another multicenter study in Nature Cancer showed AI-supported mammography caught more clinically relevant cancers without increasing false-positive rates. These developments, quieter than data-center megadeals, represent the concrete payoff behind the billions being invested. The noise is loud, but the real wins—open models, scientific breakthroughs, and improved healthcare—are building the foundation for AI's lasting impact.