Learning to wield large language models