Chain of thought

For math, logic, and multi-step problems, models often get the wrong answer if you ask for it directly — but the right one if you ask them to work through it first.

The magic phrase: "think step by step." Or "show your reasoning." The model writes out its intermediate thinking, which improves accuracy and lets you spot where it went wrong.

Modern models often do this automatically for hard questions. But for anything requiring careful reasoning, it's worth asking for explicitly. It's also useful when you're learning — reading the model's reasoning teaches you how to approach similar problems.