Hit the flag and watch the simulation.
The AI purposely starts out with bad habits, but when it does something good, it gets rewarded. If it does something bad it gets punished. Touching the red is bad, touching the yellow is good, and touching the green is really good. By far the most advanced AI I've built. (I've made some on my other account too) When it gets to the blue it must perform 1 of 4 actions, jump, big jump, turn, and keep going.