In this repository we provide the necessary code to complete our paper. The use of reinforcement learning to discover new mathematical algorithms. You can find two folders, one of them related to a reinforcement learning problem (rockets) and the other (alphatensor), based on an example of a faster algorithm to multiply matrices.