The synthetic generation of data is appropriate in many cases. The data generated can be of different nature: program texts, tables, XML documents, etc. The application area of such data comprises not only the scientific research and analysis of the real program system features, but the teaching activity, too. The generation of texts of different problems, belonging to the same type, represents such a case. The problems could be used for self-preparation as after-class training, for solving exercises in class, or for assessing the students? knowledge in a given discipline. This paper discusses the philosophy and development of a system generated sets of synthetic program texts of the same type. The examples considered are for teaching purpose.