Genetic programming has now been successfully used to automatically synthesize human-competitive designs in a number of fields, including analog electrical circuits, optical lens systems, antennas, controllers, mechanical systems, and quantum computing circuits. Despite considerable variation in representation and technique, the above results share a number of common features. This paper identifies five common features and suggests possible explanations and interpretations for them. Some of these cross-domain observations clearly could not have been anticipated prior to the work being done.