Problem Solving with Mazes and Crayon — LessWrong