MolecularMind has released the latest technical report on its AI de novo design platform, MMDesign. The report reveals that across 19 design projects involving 18 targets and two molecular modalities—nanobodies and macrocyclic peptides—the platform achieved a remarkable project success rate of 94.7%. On average, each project tested only 9 to 50 candidate molecules, with a candidate pool hit rate of approximately 30.7%. Notably, all six highly challenging GPCR targets were successfully designed, with some targets achieving a candidate hit rate of up to 100%. For five GPCR targets lacking prior examples of known therapeutic antibodies, the candidate hit rate still reached 59.6%. After conducting a new round of design for six targets with existing design outcomes, five projects demonstrated improvements in hit rate, best affinity, or both, while both macrocyclic peptide de novo design projects successfully yielded binding molecules. The molecules designed by the platform exhibited excellent developability and sequence novelty. Its underlying self-developed structural prediction model, MMFold, demonstrated exceptional performance in predicting antibody-antigen interfaces, providing crucial support for candidate evaluation and selection. These findings indicate that AI de novo design is transitioning from Demonstration Application (demonstrative applications on individual targets) to a reproducible cross-target, cross-modal research and development workflow. This shift is driving the early-stage drug discovery paradigm from reliance on massive screening to precision design, marking a significant step forward in advancing AI de novo design into an industrial-grade research and development tool.