Researchers have demonstrated de novo antibody design with atomic-level accuracy using RFdiffusion and an integrated hypermutation system, validating binding poses by cryo-electron microscopy. The study, led by Bennett, Watson and colleagues in the Baker lab and published in Nature, reports designed single-domain VHHs that bind user-specified epitopes across therapeutically relevant targets. The work confirms that generative AI models can construct antibody loops and binding interfaces previously accessible only through labor‑intensive screening. Cryo‑EM structural confirmation across multiple targets strengthens the claim that in silico design can shorten discovery timelines for monoclonal therapeutics. Xaira Therapeutics and other industry actors are already building on the method, positioning AI antibody design as a pipeline acceleration tool. Nobel laureate David Baker and coauthors caution the field remains nascent, but the data provide a tangible step toward AI-driven antibody drug discovery.