Genetic Stepping Stone

Algorithm

Original implementation Hill-climb baseline: many random rectangle proposals each frame and accepts the single best immediate improvement. Auto-improved Genetic generated with Codex 5.3 Population of rectangle proposals per frame with tournament selection, crossover, mutation, and elitism. Primitive — github.com/fogleman/primitive Adapted for this scaffold (rewritten with Claude Sonnet 4.6). Random shape candidates, best-first selection, hill climbing with analytically optimal colors; triangles, rectangles, ellipses, and circles.

Choose an algorithm, load an image, then press Start.

Tap to upload or drop an image here

Source

Evolving

Elapsed
0 ms

Last step
0 ms

Gen / sec
0

Shapes / sec
0

Error
0

Shapes
0

Generation
0

Canvas
—

Exports use the same pixel size as the evolving canvas (uniform scale; aspect ratio unchanged).