Science needs a new operating system.

The Protein Data Bank transformed biology by making protein structure a shared foundation. The Diffuse Project aims to do the same for protein dynamics: developing the experimental tooling, computational methods, and data systems needed to make molecular motion visible, measurable and usable.

Diffuse is a collaboration that we fund and help coordinate across full-time Radial team members, multiple universities, and national labs.

Scientific publishing wasn't built for machines, open collaboration, or the pace of modern research. The Stacks is an experimental platform rebuilding it from scratch, helping nano publications, open and continuous peer feedback, and machine-readable outputs to become standard practice.

Among other avenues for experimenting with research sharing, gauging downstream utility, and developing machine-interpretable solutions, we are building The Stacks in house and dogfooding with early beta testers and our own scientists.

Predicting how a small molecule will behave in the body is fundamental to drug development, but models are built on limited proprietary data, resulting in fragmented progress and gatekeeping. OpenADMET is building open, high-quality datasets characterizing small molecule properties that the broader field can train on for real-world utility.

We support not only data generation but also the infrastructure to make that data useful: a platform for running AI model competitions on open source datasets, so the field can learn from the work rather than disappear into closed pipelines.