Running 131 Natural Language Processing before LLMs π 131 From n-gram counts to attention, learning by doing