🔗 Thinking out loud
Ask the AI a tricky problem in two ways: “⚡ Immediately” - the model gives a bare answer, and “🔗 Reason” - lays out the solution step by step. Compare: reasoning can be checked step by step, but the bare answer has to be taken on faith.
🧩 Mark has 4 sisters, and each of the sisters has exactly one brother. How many children are there in the family?
⚡ Instant response
🔗 With reasoning
How it works: chain-of-thought
The model responds by predicting words one at a time. If you immediately ask for an answer, she “blurs out” the most likely one - and for trap tasks, the most likely answer is often incorrect (like a person’s intuition). She simply has no room to think.
When you ask to reason step by step, the model first writes intermediate steps - and each next one builds on the previous one. It's like counting on paper, not in your head: there is room for calculations, and there are fewer mistakes. This technique is called chain-of-thought, “chain of reasoning.”
Therefore, a simple “let’s think step by step” in the prompt noticeably increases accuracy in logic, mathematics and multi-step problems. Many models now do this themselves, but it is useful to know how to manage it. This is part of the prompting skill - you can practice it in a free lesson.
Start for free