Make it think before it answers

The prompt

Add "think step by step before giving your final answer" to any multi-step reasoning prompt — it cuts logic and math errors dramatically because the model reasons in the open instead of guessing.

Expected result

The same final answer as before, but preceded by a visible chain of intermediate reasoning steps you can inspect for errors.

Why it works

Models that reason "out loud" catch more of their own arithmetic and logic mistakes than ones that jump straight to a final answer.