Large language models (LLMs) are potentially powerful tools for automating educational tasks. In this paper, we observe two use cases of LLMs related to introductory programming education. In the first case, we created an LLM-based tool for creating variations of existing exercises. In the second case, we used LLM for generating the unit tests and good-quality feedback for students’ answers to programming exercises. Both approaches were studied by gathering data from two instances of a large introductory programming course. Our results indicate, that both approaches were successful. In addition to discussing the results, we discuss the insights gained, the identified use cases, and the significance of the rapid progress the LLMs have on programming education.