AI / Machine

Visualizations of what neural networks see — feature maps, loss landscapes, attention heads.

A two-sentence note on the aesthetic: ML visualizations turn opaque computation into legible geometry — mountains of loss, constellations of embedding, and rivers of attention — rendered as glowing cartography of how a model thinks. Each card is a placeholder shaped after the kind of artifact a researcher would pin to the wall: gradients as pigment, layers as strata, vectors as constellations.