Chain of thought
Asking a model to work through its reasoning before answering, which measurably improves accuracy on multi-step problems.
Also written: chain-of-thought, CoT, step by step.
Telling a model to work through a problem step by step before answering reliably improves accuracy on anything requiring more than one inference (Running a trained model to get an answer. Everything you pay for by the token is inference, not training.). The working-out gives the model something to condition its final answer on, rather than committing in one shot.
Newer reasoning models (A model trained to produce a long internal working-out before its answer. More accurate on hard problems, slower and dearer.) do this internally whether you ask or not, so the explicit instruction matters less than it did. It is still worth asking for on conventional models, and still worth asking to see when you need to check the logic rather than trust the conclusion.