We have just published a new paper in Journal of Cheminformatics. This paper presents C2PO, a machine-learning tool designed to help scientists improve how well cyclic peptides can cross cell membranes. Poor membrane permeability is a major reason many peptide drugs fail to work when taken by mouth. C2PO predicts how easily a peptide can cross a membrane, then suggests small chemical changes to make it more permeable. Because AI-generated structures can sometimes be unrealistic, the system also includes an automatic correction step to ensure the resulting molecules are chemically sensible. Overall, the approach helps researchers explore better drug designs more efficiently. Many thanks to Roy Aerts, Joris Tavernier (Open Analytics), Alan Kerstjens (Hyle), and the people from Janssen Pharmaceutica (Mazen Ahmad, Jose Carlos Gómez-Tamayo, Gary Tresadern) for co-developing this work.