How can the internal apparatus of AI, the so-called latent space, be made material and navigable through a scientific and artistic exploration? By combining art (Joyce’s Finnegans Wake) and science (GAN models), complex AI spaces become physical, interpretable, and offer potential answers for long-standing questions in philosophy, such as: What are symbols? How are the brain and AI similar and different? How can AI become a space of aesthetic exploration? Through three case studies, Latent Spacecraft traces deep homologies between biological and synthetic language: - Like speech-producing brains, speech-generating GANs learn by informative imitation and exhibit imagitation, which exceeds mere replication. - Joyce’s Finnegans Wake serves as a cartography of latent space, where new words and associative syntax model pre-narrative interiority. - The outputs of FinneGANs–a GAN trained on Finnegans Wake audio–probe the limits of comprehensibility and pre-speech forms. In sum, these studies show that generative models such as GANs don’t just reproduce data, but navigate latent spaces that can generate genuinely new linguistic forms. Read Latent Spacecraft: Brains, GANs, Finnegans in Antikythera Journal at Authors @ninabegus, @mthvn, Gašper Beguš Interfaces by @mthvn & Riccardo Petrini