This article describes my research on generating musical note sequences with the GraphLearner as presented in [2]. It is based on some research and a talk I gave at the Autonomus Systems Conference 2024 in Cala Millor, Mallorca. The GraphLearner is a model that resembles a higher-order Markov chain. This model is used here to generate musical note sequences. For this purpose, a directed graph is constructed from the training data, which determines the probability of the next symbol or word based on the previous symbols. In this case, musical notes with a pitch and a duration are used as symbols. Notation is therefore generated, which can also be the result of a composition process. This experiment thus approaches music from a different angle than current generative AI models that produce acoustic artifacts. It can be shown that the GraphLearner is also suitable for generating musical note sequences in this way and delivers promising results with small corpora. In the beginning the essential differences between music and natural language are discussed. The GraphLearner is then introduced and the training and generation of musical note sequences are explained. The results of the experiments are presented and discussed. Finally, the limitations of the model are discussed and possible extensions are outlined.