This essay examines the LightFM recommender system through Deleuze and Guattari’s concept of micropolitics. It argues that machine learning models like LightFM don’t just reflect user preferences—they help produce subjectivity by embedding users in global infrastructures of control. Drawing on theories of the machinic unconscious and proto-subjectivity, the piece explores how prediction and vectorization reshape agency, identity, and power in subtle, material ways.