Researchers at MIT introduced BoltzGen, a generative AI model designed to propose candidate molecules for hard‑to‑drug targets, showcasing a push to speed early discovery by integrating advanced machine learning with chemical design. The seminar unveiling drew broad academic and industry attendance and signals renewed momentum for open scientific AI tools in small‑molecule discovery. Concurrently, a proteome‑wide deep generative model, popEVE, combined evolutionary and human population data to score missense variants across the proteome and improve rare‑disease diagnosis workflows. popEVE aims to prioritize damaging variants in clinical sequencing without parental samples, addressing bottlenecks in variant interpretation for rare disease clinics. These computational advances highlight two fronts where AI is reshaping biotech: de novo molecule generation to expand chemical space, and variant effect prediction to accelerate genetic diagnosis. Both are likely to influence partnering, investment and regulator expectations for AI‑driven tools in R&D and clinical translation.