Talus Bioscience released Ptarmigan-1, a “structure-free” AI model intended to predict small-molecule binding sites across the human proteome, including proteins that are difficult to model with traditional 3D structural approaches. The company says the platform is powered by its MARMOT assay, which measures proteins in their native cellular environment using snapshots rather than purified structures. The product release matters because it targets an enduring bottleneck in discovery: binding-site prediction for disordered or unstructured proteins. By claiming coverage beyond what standard modeling tools can resolve, Talus is positioning its approach as an alternative route to prioritize targets and chemotypes earlier in the discovery cycle. For biotech teams, the practical implication is a tighter design-build-test loop for compound screening and lead optimization, supported by proprietary datasets collected over multiple years. The move also signals how AI companies are shifting from “predictive demos” toward deployable model offerings tied to experimental assay infrastructure. The next scrutiny will likely involve how Ptarmigan-1’s predictions translate into hit rates in real screening programs and how reproducibility and validation are handled across collaboration partners.