Household robots rely on computer vision to navigate their environment, but a camera does not know what it is looking at. In order to recognise and understand the spaces and objects it encounters, a robot’s vision technology needs to learn about its future home. To this end, large datasets of 3D files are assembled into model homes, which ultimately are unable to represent the complexity of life itself….This talk speaks of the absurd and precarious state of the training datasets of home robots, using them as a way of entering a wider discussion about cohabitation with technology and an obsession of sorting the world into categories.