autobliss 1.0 simulates a programmable 2-input logic gate that you can program using reinforcement learning, i.e. reward and punishment. It can learn any of the 16 possible 2-input logic functions, i.e.AND, OR, XOR, NAND, etc. You train it by rewarding it when it gives the correct output (0 or 1) for a given input (00, 01, 10, 11), and punishing it when it gives the wrong output.
I'd like to see implementations of this in other languages when I have time.