How it works
This is the Gray–Scott model. Two imaginary chemicals share a grid. A spreads out and is topped up from outside at the feed rate; B spreads out more slowly and is removed at the kill rate. The only thing coupling them is the reaction A + 2B → 3B, which consumes a unit of A and produces one more B wherever B is already present — so B eats A, and it eats fastest where there is most of it.
That is the entire model, and everything on the page is a consequence of two numbers. Move the feed and kill rates a few thousandths and the same three lines of arithmetic give spots, stripes, a maze, dividing cells or travelling waves. Most of the plane is dead — the pattern either dies out or floods — and the band that does anything is narrow, which is why the regimes are named rather than left to the sliders. B has to diffuse more slowly than A, too: at equal rates nothing happens at all, and that difference in rates is the whole of what Turing proposed in 1952.
The simulation is a raster, but nothing raster is drawn. The finished concentration field is contoured with marching squares at evenly spaced levels, so what lands on the sheet is line work and the print file is a vector. The grid wraps at the edges, which is why the pattern has no border to organise itself against, and it is a fixed grid rather than one the size of the canvas — an A2 is the same picture as the gallery thumbnail rather than a denser relative of it.