In 2023, the Antimicrobial Peptide Database (currently available at https://aps.unmc.edu) is 20‐years‐old. The timeline for the APD expansion in peptide entries, classification methods, search functions, post‐translational modifications, binding targets, and mechanisms of action of antimicrobial peptides (AMPs) has been summarized in our previous Protein Science paper. This article highlights new database additions and findings. To facilitate antimicrobial development to combat drug‐resistant pathogens, the APD has been re‐annotating the data for antibacterial activity (active, inactive, and uncertain), toxicity (hemolytic and nonhemolytic AMPs), and salt tolerance (salt sensitive and insensitive). Comparison of the respective desired and undesired AMP groups produces new knowledge for peptide design. Our unification of AMPs from the six life kingdoms into “natural AMPs” enabled the first comparison with globular or transmembrane proteins. Due to the dominance of amphipathic helical and disulfide‐linked peptides, cysteine, glycine, and lysine in natural AMPs are much more abundant than those in globular proteins. To include peptides predicted by machine learning, a new “predicted” group has been created. Remarkably, the averaged amino acid composition of predicted peptides is located between the lower bound of natural AMPs and the upper bound of synthetic peptides. Synthetic peptides in the current APD, with the highest cationic and hydrophobic amino acid percentages, are mostly designed with varying degrees of optimization. Hence, natural AMPs accumulated in the APD over 20 years have laid the foundation for machine learning prediction. We discuss future directions for peptide discovery. It is anticipated that the APD will continue to play a role in research and education.