This article examines the features of algorithmic music performance systems and considers how these might apply to other generative content creation contexts. Based on the assumption that all generative processes are performative, the article draws from an analysis of live algorithmic music practices, it outlines lessons that may be helpful more generally for co-creation with algorithmic systems. In particular the article discusses algorithm selection and expression, the architecture of algorithmic system design, the effects of materiality on algorithmic performance, and how co-creative strategies openly embrace the influence of humans as agents in generative content systems.