John Loeber says his custom Lean theorem prover failed to close the Hadwiger Conjecture, then an unreleased OpenAI frontier model disproved it. He reports that model ran 4,000 major problems at about 3.5 hours of compute each, producing hundreds of claimed results. Loeber calls his elaborate engineering outweighed by base-model gains.
Google, Google DeepMind and MIT FutureTech report that AI saves researchers about seven hours a week, yet 44 percent of 637 surveyed US and UK scientists say their main bottleneck moved downstream and 41 percent face growing backlogs of untested hypotheses. Co-lead Mihai Codreanu says 88 percent spend more than 10 percent of saved time verifying outputs.
Anthropic says its internal research model designed 17 protein binders that attach to scorpion neurotoxins, working mostly autonomously over 64 hours with four follow-up prompts. Of 211 designs tested by Adaptyv Bio and Twist Bioscience, 17 met the company's binding criteria. None has been tested in cells or animals, and Anthropic releases all 211 designs and their data.
Researchers at MD Anderson and Texas collaborators report SPACER, an AI model using spatial transcriptomics, decodes signals that draw or repel immune cells. Tested on 37 cancer datasets, the model performed four to 21 times better than existing methods at finding recruitment signals. They say mucins may block T cells, linked to poorer checkpoint inhibitor responses.
Hugging Face's HuggingFaceBio team releases Carbon-A, a 1.2-billion-parameter model that predicts protein-coding genes from DNA, and the Carbon Annotation Database of 566 million gene candidates across 22,617 species. The team reports a 0.944 nucleotide F1 across 42 benchmark genomes. Iso-Seq tests in cat, hamster, chicken and Arabidopsis support some predictions missing from RefSeq.
CZ Biohub, the Department of Energy, and the NIH commit $1.8 billion to generate open, AI-ready biological data for predicting and treating disease. The DOE plans to spend more than $500 million over five years through its Genesis Mission, while Google DeepMind, Isomorphic Labs, and Meta together invest $300 million in the Virtual Biology Initiative. Biohub will standardize datasets from prior NIH-funded repositories for AI model training.
Zuckerberg and Chan's Biohub is expanding its $500M Virtual Biology Initiative into a $1.8B effort with NIH, the Energy Department, Google DeepMind, Meta, and Isomorphic Labs to train AI that simulates cells. The Energy Department commits $500M+ over five years. Science head Alex Rives says accurate models will need data on trillions of cells, versus hundreds of millions today.
OpenAI says an unreleased AI model solves the Navier-Stokes Millennium Prize problem using 10,000 concurrent agents and creates an independent mathematician advisory group to vet that claim and scores of other reported results.
Google DeepMind unveils and open-sources SynthID Bio, which embeds detectable watermarks in AI-designed protein sequences and structures without reducing function, with wet-lab tests spanning VEGF-A, SARS-CoV-2 spike RBD and PD-L1 binders.
Anthropic says 950 Claude agents make a molecular-biology discovery in 21 hours by identifying an uncatalogued repeating pattern around a known enzyme, but University of Copenhagen biologist Mario RodrÃguez Mestre says his team had already found it and raises concerns about whether Claude learned from their chats, which Anthropic denies.