The technologies of generative AI are increasingly penetrating the social and political world. Not merely in the direct sense that AI models are being deployed to govern difficult problems – from decisions on the battlefield to responses to pandemic – but because generative AI is shaping and delimiting the politics of what can be known and actioned in the world. The machine learning algorithm is leaving the trace of its technical worldview in the political spaces of people, communities, objects, and scenes. This is a new and distinctive political logic: it remodels and aligns social and political relationships to make them fit the abstractions of machine learning models. The latent spaces and embedding spaces of AI are becoming present in the actual lived spaces of our world. In this zero-shot world, the machine’s capacity to act on unencountered data and tasks structures human societies of experimental probability, foreclosing the spaces of social organisation and resistance.