The Shape of Randomness in Generative Art
Randomness is rarely the point. Constraint is.
Generative art is often described as “letting randomness create beauty,” but that description leaves out the most important part. Randomness by itself is mostly noise. The interesting thing is how a system channels randomness through structure.
A seeded process creates a family resemblance between outputs. The rules are fixed, but the realization varies. This is close to how style works in human work as well: not identical results, but recognizable habits.
I like systems where the constraint is mathematical and the output is visual. Angle restrictions, graph rewrites, nearest-neighbor growth, and energy minimization all produce artifacts that feel intentional without being literal.
There is also something satisfying about reproducibility. If a piece emerges from code and seed, then the artwork is both an image and a proof that the image could happen. That dual identity still feels magical to me.
When I get stuck, I usually do not add more randomness. I add better geometry. Disorder is abundant. Interesting boundaries are rare.