Originally posted some time around 2016 or so, and just recently dug up from old web archives. I thought it was neat though. Even if it does clearly have some problems.

These models work by dropping “grains” on random spots on the grid. They stack up, and once a spot has 4 grains it “collapses” and distributes its 4 grains to its cardinal neighbors.

One very interesting property of this particular model is that it will always wind up the same given a certain starting state, regardless of what order you simulate the individual spots in. As long as you hit all of them until all of them have been hit without anything happening, the result will be the same no matter how you go about deciding how to order your simulation steps.

See Wikipedia: Abelian sandpile model

Running model

sandpile static display