Following from a shared interest in livecoding and real-time algorithmic performance, Joana Chicau and Jonathan Reus begin a research project into techniques for in-situ dissections of machine learning algorithms. We seek to better understand the habitual and fixed objects of machine learning as well as their terminologies, and provide counter-techniques for conditions of emergence and movement. In our processual approach, we aim to develop an online repository of terminology and techniques for a critical examination of the “anatomy” of learning and prediction processes, data corpus and models of machine learning algorithms. And explore, through performance practice, how such a toolkit can confront the idealized bodies of artificial intelligence.