I highly recommend putting this into turbowarp and turning up the clone count slider above 150 this is just "machine learning I think" and "neural network learns the smat" but it works much better -A and D to move camera -Click camera button to toggle spectating the farthest cat -Colors are random, more brightness indicates complexity
Creates a bunch of randomized cats that go through a series of courses The most successful one makes it to the next trial and gets cloned a bunch of times with minor changes (The original is kept as well) More clones = more diversity = faster learning(but more lag)
