In a landscape where the technologies governing our lives are increasingly obscured, this publication aims to reveal the underlying processes of a machine learning program. It seeks to mediate machine “sight” and human comprehension. The book documents the technical capabilities of an algorithm trained to analyze the formal patterns of the alphabet and imagine other letters from that knowledge. The resulting outputs — letters without semantic value — embraces and emphasizes the meaningless results of the machine: it is the human input and framework that imbues “sense” to these forms. Combining expository exercises with surreal typographic characters, Latent Space: Notes on Seeing Letters Like a Machine is both a grounding text on the mechanics of machine learning and a visual celebration of formal possibility.