From Renaissance ideal cities to modernist plans, from speculative megastructures to smart-city visions, from fictional urban worlds to the cities of video games, the anticipated city forms a vast archive of collective urban imagination. And the form of this archive has just emancipated and evolved into something that demands our attention. Multimodal foundation models are now no longer only tools for analyzing that archive, but have themselves become one of its contemporary forms. Trained on immense corpora of images and texts, they absorb and reorganize real, imagined, projected, and simulated cities into new constellations of resemblance, plausibility, and value.
Moreover, I ask how models reshape the relation between accumulated urban experience and anticipated urban futures. What happens when centuries of idealization, projection, fear, and desire are compressed into the operative memory of the model? On the one hand, this process risks narrowing the horizon of expectation, reducing novelty to recombination and pulling diverse urban imaginaries toward what is already legible and familiar. On the other hand, it may also reveal unexpected affinities, dormant lineages, and strange proximities across times, media, and ideologies.
