copied from https://observablehq.com/collection/@j-kan/reinforcement-learning and others