Running 108 Natural Language Processing before LLMs ๐ 108 From n-gram counts to attention, learning by doing