Procgen Image Grammar (PIG) is something I've worked on on the last two days

http://zara.verge-rpg.com/procgen/

It works like this: it reads in an image and interprets it as rules. each rule is a image it searches for in the screen, and a bunch of replacement images for that bit

Rules images that exist so far are:

rules.png (makes a grid)
rules2.png (test of mouse control: click on a blue box)
water.png (simulates water aka world of sand)
maze.png (random maze generation)
trees.png (random tree/hole generation)
city.png
castle.png

it might be useful to generate stuff procedurally.

> It might make more sense if you look at the images directly; for example look at http://zara.verge-rpg.com/procgen/rules.png -- you see each row has two images; the program searches the current screen for the stuff in the left column, and replaces it with the corresponding stuff in the right column, which leads to animations.  I think.