Code synthesis and self-play with reinforcement learning - View it on GitHub