A generative art engine that creates one artwork per day, shaped entirely by @Noahbolanowski's tweets about art and technology history. Each piece encodes the themes, images, and emotional register of a day's worth of curatorial thought — then breeds with its ancestors to evolve over time.
Tweets are scraped nightly and analyzed for visual and thematic signals — palettes, motifs, dominant concepts. A genome of ~25 parameters is biased toward the day's content. 12 candidates render; the best is selected by an aesthetic fitness function (with occasional curatorial overrides when theme and fitness disagree). The winning genome enters an ancestry archive and breeds future generations.
Every card is a day. The condensed reason describes what the system heard in the tweets and why it made the choices it made — an algorithmic catalogue note. Dashed lines between cards show lineage: when a winner's genome influenced a later generation. ★ marks Hall of Fame entries that spawned three or more descendants.
Restraint is the method. Each piece reduces its source to a single visual signature and renders that alone — one dominant mark, everything else near silence.
A generative engine's natural pull is toward the average. The discipline here is to resist it: every constraint — one signature per piece, a small fixed vocabulary of marks, a deterministic seed pinned and certified so a work lives at one exact point rather than drifting through endless variation — is a deliberate stand against the mean. The source is translated, never illustrated. What is left out is the work.
After Lialina — speak in the medium's own language; after Manovich — an AI left unchecked drifts to the statistical center. Constraint is where authorship lives.