Researchers at the University of Pennsylvania unveiled ApexGO, an AI method designed to optimize peptide antibiotic candidates starting from a small number of imperfect molecules. Rather than screening massive libraries, ApexGO iteratively proposes modifications with a predictive algorithm to guide the next design round. In the Nature Machine Intelligence work, the team reported real-world validation: 85% of AI-generated molecules halted bacterial growth, and 72% outperformed the peptides they were derived from. In mice, two ApexGO-produced antimicrobial peptides reduced bacterial counts at levels comparable to polymyxin B. The group framed ApexGO as a navigational tool through antibiotic chemical space, emphasizing that lab performance largely matched model predictions and suggesting a path to faster peptide optimization cycles for antimicrobial discovery.