News

Researchers at the University of California, Berkeley built an efficient genomic language model that uses evolutionary comparisons to find DNA changes most likely to affect health and inherited traits.

Rentosertib, built with generative AI for scarred lungs, starts a year-long study after protein clocks suggested treated patients looked biologically younger.

A vision model trained on millions of simulated gas trails finds, outlines, and traces methane sources in NASA images from the International Space Station.

Six independent protein-based aging clocks moved younger in forty-two patients with idiopathic pulmonary fibrosis, with the strongest signal at four weeks on a mid-range dose.

An internal model and thousands of coordinating agents produced a machine-checked argument that smooth three-dimensional flow can reach unbounded speed in finite time.

Google DeepMind has precomputed predicted effects for all nine billion possible single-letter DNA changes and released a simple score so researchers can look them up.