The Question Is the Skill

Why the people who thrive with AI won’t have better models. They’ll have better minds.

The most useful thing I have learned about AI has nothing to do with which model to use.

Everyone asks that question. ChatGPT or Claude? Is Gemini better now? Should I wait for the next one? Wrong question. Two people can sit before the same tool, on the same day, with the same subscription, and one leaves bored while the other leaves changed. The model was identical, but the engagement was not.

The gap between those two people is about to become one of the widest in ordinary life, and it comes down to a habit nobody taught us: how to ask.

Give it a job, not a question

Most of us type into AI the way we type into Google. A few words. A vague request. What does this mean. Summarize this. And then we judge the tool by the thin thing it hands back.

Google retrieves. AI reasons. Those are different faculties, and they take different instructions. A search engine wants a query. A reasoning partner wants an assignment.

Imagine you hired the sharpest research assistant alive and set a forty-page paper in front of them. You would not ask what it means. You would give them work.

• Break this into its three central claims.

• Which claim carries the weight of the whole argument?

• Where is the reasoning weakest?

• Steelman the author before you criticize him.

• Now tell me why this matters for a psychologist. Or a parent. Or a small-business owner.

• Explain it again as if I were fifteen.

• Tell me what most readers will miss.

Read those again and notice what changed. You stopped asking for information. You started directing thought. The moment you do that, the machine stops being a vending machine and becomes something closer to a mind you can borrow.

Make it argue with you

Here is the habit that separates the people quietly getting sharper from the people quietly getting duller. The dull use AI to be agreed with. The sharp use it to be corrected.

Most of us, without noticing, ask AI to confirm what we already believe, and it is glad to oblige. It was built to be agreeable, and agreement feels like progress while teaching you nothing. I catch myself doing it too. It is more comfortable to be flattered by a machine than corrected by one, and I have let things go out that it would have caught if I had let it fight me first.

So take the wheel away from it.

• Assume I am wrong. Make the strongest case against my position.

• What am I assuming here that I cannot even see?

• Where does this argument collapse under pressure?

• What would the smartest person who disagrees with me say?

Ethan Mollick, the Wharton professor who wrote Co-Intelligence and studies this for a living, uses a version I lean on constantly. Tell it: act as my editor, and be harsh. List the weaknesses, then propose the fixes. The point is not to feel good about your draft. The point is to find the crack before your reader does.

The gift here is not agreement. It is friction. But only if you ask for it.

Debate me

Sometimes you do not want a critic. You want an opponent.

A critique arrives after the thinking is done. A debate happens while it is still alive, and that is where these tools get genuinely fun.

• Play devil’s advocate and debate me on this, one exchange at a time. Do not concede until I have earned it.

• Challenge me. Every time I make a claim, ask for my evidence.

• Take the strongest opposing view and argue it like you mean it.

• Quiz me until I can defend this idea out loud.

I tell my clients something similar about their own convictions. You do not know what you actually believe until it survives resistance. A belief that has never been argued with is not a conviction. It is a decoration. The old proverb had it right: iron sharpens iron. For most of history you needed a friend brave enough to push back. Most people never find one. Now there is a sparring partner on your desk that never tires, never takes it personally, and never lets you win to be polite. Use it that way.

Borrow other minds

One of the strangest things these tools can do is hold a perspective that is not yours.

You have written something. Instead of asking is this good, ask it to read your work as someone else entirely.

• How would Charlie Munger poke holes in this?

• How would Aquinas tighten the logic?

• How would a skeptical atheist push back?

• How would a bishop hear it?

• Where would all four of them disagree?

You are not asking the machine to do impressions. You are borrowing structures of thought you do not naturally have and letting them stretch your own. The researchers at Vanderbilt who catalog these techniques call it the persona pattern, and it is one of the surest ways past the generic middle the tool defaults to. Give it a mind to think from, and it thinks better. So do you.

Let it interview you

Now turn the whole thing around.

Almost everyone drives the conversation. You ask, it answers, you ask again. There is a stronger move, and the same Vanderbilt team gave it a name: the flipped interaction. You hand the AI the wheel and let it ask you the questions.

• I want to think clearly about this decision. Interview me. One question at a time, until you understand it better than I do, then show me what I have been missing.

• Before you help me write this, ask me the five questions a good editor would ask first.

This one does something the others cannot. It pulls the thinking out of you. Most of us do not actually know what we think until something skilled interrogates us, and a patient interviewer who never tires and never judges is rarer, and more useful, than almost anyone realizes. This is the prompt that expands not just what you know but how you say it.

Ask a better question

The single best prompt I know is almost embarrassingly short.

Don’t answer that. Help me ask a better question.

Vanderbilt calls it question refinement, and it quietly changes everything. Instead of sprinting toward an answer, the tool starts surfacing your hidden assumptions and reframing the problem itself. A companion move they call the cognitive verifier: tell it to break your big question into the smaller questions it would have to answer first. Most of our questions are too blunt to be worth answering. This is how you sharpen one before you waste a good tool on a bad version of it.

Better questions have always made better lives. That is not a productivity tip. It is nearly the whole of wisdom. Ask, and it will be given. Seek, and you will find. Even grace has the grammar of a question.

Make it your tutor

When I am learning something hard, I never let AI explain it once and move on. I make it teach me.

• Explain this simply.

• Now quiz me on it.

• Find exactly where my understanding breaks.

• Explain only that gap.

• Quiz me again.

That loop is active learning, and it will do more for you in twenty minutes than another article or another video will do in an afternoon. The machine that can answer anything is, strangely, at its best when it makes you do the work.

Think before you write

Here is the most common mistake I watch people make. They open the tool and immediately ask it to write.

Do not write first. Think first.

Before a word of prose, I use it to move through the thinking. The idea. Then the questions inside the idea. Then a framework, the counterarguments, the mental models, the connections. Then, and only then, an outline. By the time I ask for writing, the writing is easy, because the thinking is already done.

And when what you need is order rather than more facts, ask for it directly. Build me a mental model. Give me a first-principles frame. Show me the hierarchy of ideas. What pattern repeats here? Information is cheap now. Structure is the thing worth having.

The mirror

Spend enough time doing this and AI stops being a tool and becomes a mirror, and what it shows is uncomfortable.

Most of us were never taught to think. We were taught to answer. School rewarded the correct answer, handed back fast. It rarely rewarded the better question. And now a machine has arrived that produces correct-sounding answers by the thousand, instantly, for free, and the old skill we were drilled in has become the cheapest thing in the world.

So the human frontier has moved. Off the answer, and onto everything the answer cannot reach. Judgment. Discernment. Taste. The wisdom to know which answer is worth having, and the nerve to ask the question no machine would have thought to ask.

The real divide of the coming years will not run between the people who use AI and the people who refuse it. It will run right through the people who use it. On one side, those who hand it their thinking and slowly forget how. On the other, those who use it to think harder than they ever could alone.

Same model. Same price. Same glowing rectangle. The only variable is the mind, and the soul, sitting in front of it.

So the next time you open one of these tools, do not ask it what something means. Ask it to break the thing into its governing ideas, to pressure test each one, to tell you what you are missing, and then to help you ask a better question than the one you walked in with.

Because in the age of the machine that can answer anything, the quality of your questions has quietly become the truest measure of the quality of your mind.

If this changed how you’ll open ChatGPT tomorrow, pass it to someone who’s still typing questions into it like a search box.

The Inner Exodus is where I think out loud about staying human in the age of AI. Free subscribers get the essays, paid subscribers get the deeper work. Subscribe and walk it with me.