We describe work in progress on the development of a new hierarchical model of machine creativity operating in the domain of music. Similar to the way human brains work, our system separates low-level components associated with pattern recognition and analysis from the high-level creative components in two extensible layers. Separating this functionality in different layers of our system provides better visibility into the behavior of the creative component. This increased visibility has led to many improvements over previous iterations including the reward calculation for the creative component. Additionally, the design of an abstract input feature layer allows for greater flexibility in the number and combination of low-level features that can be used within our system.