Language models happen to share human flaws, but like humans they can amplify their abilities and reliability by building and using reliable tools.