An MIT‑led team unveiled BoltzGen, a supercomputer‑trained AI model that moves beyond high‑accuracy structure prediction to generalizable therapeutic design across modalities — nanobodies, mini‑binders, peptides and small molecules. The model is released under an MIT license and several academic and industry collaborators report early wet‑lab validation showing nanomolar affinities. BoltzGen’s open‑source availability aims to democratize design workflows and accelerate discovery for targets previously considered undruggable; commercialization and IP strategies will be a near‑term focus for industry adopters integrating the model with proprietary datasets.