Running 140 Natural Language Processing before LLMs π 140 From n-gram counts to attention, learning by doing