Researchers from Insilico Medicine and Eli Lilly published a vision for a fully autonomous “prompt‑to‑drug” R&D pipeline that stitches together generative AI, automated synthesis and iterative experimental validation to compress discovery timelines. The perspective argues for integrated systems that translate natural‑language prompts into candidate molecules with minimal human intervention. Complementing that vision, the MULTI‑evolve lab‑in‑the‑loop framework demonstrated rapid directed evolution of multi‑mutant proteins by combining protein language models with compact experimental screens of double mutants to model epistatic interactions—delivering functional high‑fitness variants with fewer rounds. Taken together, these reports signal two converging paths: high‑level AI orchestration for end‑to‑end drug design and tightly coupled ML‑guided experimental loops that materially reduce iteration time. Both approaches emphasize the need for robust validation, data standards and transparency to translate models into reliable leads. Next steps: companies and labs will pilot integrated workflows, establish validation benchmarks, and invest in infrastructure to move from proof‑of‑concept to reproducible, scalable discovery.