I read Lee Si-han’s How Smart People Think and Ask (Korean title 똑똑한 사람은 어떻게 생각하고 질문하는가, Book Pleasure). The subtitle calls questioning a survival skill in an age that has misplaced thinking. The marketing leans on Mensa and ChatGPT. The useful claim is smaller than that. Knowing more is no longer scarce. Asking well is. A search bar, and now a model, will furnish answers all day. They will not tell you which question was worth opening.
Lee has been a Korean Mensa member, a corporate lecturer, and a university instructor. He writes as someone who has watched “smart” people work and decided the difference is not the size of the storehouse. It is what they do before they open their mouth, and what they do when they are alone with a problem. The book is a popular how-to, not a treatise on cognition. That is useful to know before you open it. You will get stages, conversation tricks, and a last chapter on prompting. You will also get one stubborn instruction: do not start with the answer.
Table of contents
Open Table of contents
What the book actually argues
In one sentence: intelligence is the habit of making a question that has a core and a context.
Lee’s historical sketch is local and blunt. For a long stretch, Korea rewarded the person who followed the manual. Growth was high, the work was to copy a process already known, and the virtues on the poster were diligence, grit, responsibility, passion. After the late-1990s shock, growth slowed, firms had to lead rather than follow, and information became something you could search. The person who only executed the sheet became expensive. The person who asked why the sheet was the sheet became useful. AI, in the last chapter, is just that shift again, faster. When answers are generated, the scarce act is the prompt — and, more importantly, the human question behind it.
He splits the work into three rooms. Questions you ask other people. Questions you ask yourself. Questions you ask a machine. The first gathers knowledge. The second, if you stay with it, is closer to judgment. The third is a force multiplier only if the first two already exist. A vague prompt is the same failure as a vague meeting question: you outsourced the thinking and then blamed the reply.
Before any of that, he wants two words held in the head: core and context. If you lose the thread of the conversation, or of your own problem, the next question will be decoration. Smart people, in his account, keep asking Where are we? so they do not perform curiosity.
The misconceptions it breaks first
A good question is just being curious. Warmth is enough to get a real answer. The first why is enough. More information will make you smarter. Lee doubts all four.
Curiosity without aim is browsing. He would rather you name the purpose first. Do you want a fact, an opinion, or a fight that moves? If you cannot say which, you are not asking. You are making noise that happens to end with a question mark.
Warmth without analysis is the other popular mistake. Empathy, here, is not a soft voice over an empty sentence. It is an accurate reading of the other person’s stake, knowledge, and timing, then a question that can be received. The worst question is the one asked to display the asker. Nobody learns, and the room closes.
The first why is a start, not a method. He likes the old five-whys habit: stay with the same object until the cause stops being a slogan. One why gets you a convenient story. Several, aimed at the same point, get you nearer the mechanism.
The last misconception is the adult’s favorite hiding place. Collect articles, save threads, attend another talk. External questions can pile up knowledge. They do not, by themselves, become judgment. That turn happens when the same person asks inward, and stays for the answer they did not want.
The method, in ordinary language
The book is four chapters. The last one is a ChatGPT manual. The first three are the actual argument.
1. Think before you ask. Name the purpose. Check that the knowledge inside the question is not already wrong. Make the question specific enough that a useful answer is possible. Vague questions harvest vague replies; specific ones can be wrong, which is still better. Consider the other person’s position and the timing. Predict a reply so you know what you are steering toward. Keep a second question ready if the first one dies. This list is long on purpose. The point is that the expensive work happens before the sentence is spoken.
2. Design the conversation from the first word. Why looks backward, toward causes. How looks forward, toward a move. What if opens a counterfactual. For example forces a picture instead of a fog. Those four are not magic words. They set the pattern of the next ten minutes. Tone does the same job. A constructive frame pulls solutions. A negative frame, used on purpose, names a risk. Used as a habit, it only makes people defend themselves.
3. Ask yourself the converging question. Questions aimed outward fan out. Questions aimed inward should narrow. Knowledge, Lee says, becomes something closer to intelligence when you strip the shell: What is the claim? What would count as a reason? What would count as a counterexample? He borrows the usual costumes — Aristotle for argument, Holmes for inference, a comic-book engineer for synthesis — but the practice is ordinary. Close the tab. State the problem in one sentence. Ask the next why of yourself, not of the feed.
4. Treat the machine as a mouth, not a mind. If you cannot say what you want recognized, specified, and returned, the model will invent a fluent average. Better prompts are not a new personality. They are the same discipline as a good meeting question, written down. The chapter is dated in the way all 2024 AI chapters are dated. The rule underneath is not: the quality of the output tracks the quality of the constraint.
He also wants the brain kept slightly uncomfortable. Left alone, curiosity finishes its small tour and goes to sleep inside common sense. The “question-mark brain” is just the decision to keep poking the process, the room, and the job that everyone has already agreed is fine.
Why an adult should read this
School trained the answer. There was a range, a date, and a correct term. Work still rewards people who can finish a known task. It increasingly punishes people who cannot see when the known task is the wrong one. Lee’s book is useful because it names that gap without turning it into a talent myth. You do not need a Mensa card. You need a pause before the question, and a second question after the first reply.
It is also useful as a warning about performance. Many workplace questions are status moves. Many AI sessions are the same move, aimed at a machine. Both produce sentences. Neither produces a core. If you read the book as a list of clever openers, it will evaporate. If you read it as a demand to locate the problem before you speak, it sits closer to the rest of this blog than the cover suggests. Ultralearning asks you to name the skill. This book asks you to name the question that would prove you are working on it.
Skip or skim the last chapter if you already know how to talk to a model. The first three chapters are the ones that change Monday.
What remains after reading
The remaining idea is smaller than “become a smart person.” It is this: an answer is cheap, and a question that has a core and a context is not.
The next action does not require a new identity as a thinker. Take one stuck piece of work. Write the purpose of the next question in a clause. Ask why until the cause stops being a slogan, or how until a next move appears. If you can still say where you are in the problem after that, the book has started to work — not when you finished the last page, but when the next meeting opened with a sentence you had already paid for.