KNN Cerulean 46

Scatter points at random across a square, then let each one draw a line to the eight points closest to it. That is the whole rule.

What comes out isn't an even mesh but a kind of lace. Where the points happen to crowd together, short lines pile up into dense knots. Where they thin out, open rounded holes appear, crossed by long single lines reaching for a neighbour far away. Being close isn't mutual: a lone point reaches across the gap, but the crowd on the other side rarely reaches back. So each piece maps the small accidents in its own random scatter.

The idea comes from k-nearest neighbors, one of the oldest methods in machine learning. Here only the part that finds the neighbours is kept, and drawn. Each image is generated in Processing from a recorded seed and set of settings, so the composition is the same at any size, line for line.