...The core algorithm is iterative and “hill-climbing”: given a target image, it repeatedly finds the best single shape to add that will reduce the error between the current approximation and the target image, then draws that shape. Over time (e.g., adding 50-200 shapes) the output becomes an abstracted version of the input image—recognizable, artistic, and often visually striking. The tool supports input resizing, different shape modes (triangle, rectangle, circle, bezier, polygon), alpha blending, and can output PNG, JPG, SVG, or even animated GIFs that show the incremental build process. Because the Go code is self-contained and efficient, users can run it locally on images, script batch runs, or use it to generate “primitive art” or stylized thumbnails.