Running 127 Natural Language Processing before LLMs ๐ 127 From n-gram counts to attention, learning by doing