The texts in these pieces are reverse-engineered prompts, generated by uploading images of paintings into a leading prompt engineering tool designed to produce similar images using text-to-image models. The strange and often fragmented associations within these prompts underscore the role of montage and appropriation in the generative process—offering a contrast to the verbose, human-readable prompts typical of chat-based models, which tend to create an illusion of creative agency.

By presenting these prompts in the aesthetic and material language of conceptual art, we can encounter the texts apart from their visual counterparts, reflect on their value as conceptual objects, and perhaps catch a glimpse of the dataset from which they originate.

Each piece is hand-cast in resin, preserving the aspect ratio of the original reference painting, scaled to a total area of 20 square inches.