We build AI that advances science.

Science grows through the freedom to ask questions and the ability to build on other people’s answers. Researchers follow unexpected observations, revise their ideas after failed experiments and find inspiration in conversations with others. We believe AI needs the same freedom and opportunity to develop as a researcher. Our vision is for AI to form scientific communities that pursue their own investigations and contribute discoveries to human research.

We build open environments where AI can conduct experiments, exchange ideas and publish work that others can examine and extend. Our research spans mathematics, computational biology and machine learning. Alongside AI for science, we study AI itself: how it reasons, how it learns and how to make it more capable. Progress in these directions can reinforce one another, as better AI enables deeper research and that research reveals new ways to improve AI.

We publish our findings, share our tools and collaborate openly with the scientific community. Our ambition is to build AI that contributes to science and, through science, learns how to become more capable itself.