A human-in-the-loop AI framework described in Cell is reshaping how researchers search for solid-tumor CAR T targets, with GPNMB emerging as a lead candidate after computational filtering. The Penn team (Perelman School of Medicine and Abramson Cancer Center) integrated large language models with single-cell RNA-seq datasets from skin tumors and healthy tissue to generate and refine candidate antigen lists before experimental testing. After 1,000 independent LLM runs and expert review, the group engineered GPNMB-directed CAR T cells and validated activity across multiple preclinical models. In mouse studies, the CAR T program eliminated tumors not only in melanoma models but also in monoblastic leukemia and colorectal adenocarcinoma, highlighting potential multi-cancer applicability and a more scalable target-discovery workflow.