Researchers at Nanjing Medical University reported an AI-generated T-cell engager for gastric cancer designed to recruit immune cells, but durable antitumor responses required added dendritic-cell support. The study used RFdiffusion to build a CLDN18.2-targeting T-cell engager and then combined it with IL-12 and chemokine signalling (CXCL9/CXCL10) to sustain efficacy. The study maps a practical immunotherapy design concept: engineered targeting molecules may be insufficient without correcting upstream antigen presentation and immune activation bottlenecks. For drug developers, the results reinforce dendritic-cell orchestration as a lever for improving response depth. Next steps should clarify biomarker dependencies and whether the approach can be translated into combinations that fit safety constraints, especially for gastric cancer where the tumour microenvironment often suppresses effective T-cell responses.