I've always wanted to tackle doing cellular algorithms on a hexagonal grid, which has certain advantages over the square grid, insofar as the immediate neighbors of a cell in a hexagonal grid are all in fact equidistant, whereas the Moore neighborhood on a square grid has four near neighbors and four distant neighbors.
but it seems (from my very cursory examination of what's already been done) that it's tougher to get "Life-like" behavior from a simple 2D cellular automaton on a hexagonal grid.
this GIF at least shows a glider emerging from "soup". the rule is B4/S34H, with a star-shaped twelve-cell neighborhood being considered (like this:) 