Every year I am setting "an educational guessing game based on a public API with a permissive licence" as a final year project and so far no one has tackled the interesting question of finding ways to automatically classify the difficulty of the generated questions. 😩