Let AI ask me one question at a time

Order tonight, or wait until the day after. A ¥4800 laptop sits in the cart. You want next month's paycheck in before you repay it. The credit-card billing date is the 5th, due date the 25th, and today is May 4. You ask AI: “If I buy tonight versus the day after, how much interest-free time do I lose?”

AI comes back with a whole paragraph. Billing cycle, posting time, billing date, due date — all explained, plus five self-check questions thrown in.

The interest-free window is set by the billing cycle and the due date together. Charges posted before the billing date go on this statement; charges posted after usually go on the next. You can walk it: what's the billing date? When does the charge post? Which statement does it land on? What's the matching due date? How many days apart are the two cases? The bank's posting rules still win.

You read from the first sentence to the last. You know every word. It even listed the questions for you. You figure the information is already complete, save the answer, and never run the numbers again before you order.

Next day at noon a colleague sees the order and asks: “If this charge posts on May 4 versus May 6, how far apart are the latest due dates?” You look at the two dates and say, “just two days.” One actually lands on the May statement, the other on June. In this simplified example the latest due dates are May 25 and June 25. The slack you left for next month's paycheck may have just lost a whole billing cycle.

You decided you'd understood because that answer laid the conditions, the conclusion, and five questions in front of you at once. Your eyes always found the next sentence. Your head never did “which statement does this charge go on” alone. Complete content only means everything is on the screen.

Ask AI to say it another way and it will still start from the top. Billing date, posting, due date show up again, and you'll feel familiar again. What you actually need to check: take the whole explanation away — can you push the next step yourself.

Have AI pause the lecture and ask one question at a time. The process becomes three steps:

1. First ask a question that lets the reasoning continue, e.g. “on which day does the bank put this charge on the statement.”
2. Wait for your answer before asking the next. What you can answer doesn't get taught again.
3. Stop where you can't answer. Give only the hint this question needs, then have you answer once more.

One question at a time turns “I think I get it” into something you can answer or can't. When it keeps asking, which question does it actually stop on? Below, two conversations start from the same question at once.

You answered the first two on the right, which means you can already retrieve the billing date and the repayment window. At “how far apart are the day before and the day after the billing date,” you have to drop the dates into two billing cycles. That “just two days” shows up right here.

The left covered the same thing, and you could ride the sentences to the end. The right takes the explanation away first and makes you calculate once. The place you can't answer becomes visible, so AI knows the next sentence should fill in the billing cycle, and the first two chunks can be skipped.

This template holds four moves:

1. Pause the full lecture first, so the next step shows up as a question.
2. Show only one question at a time, so you don't see the rest of the path early.
3. Pick the next step from your answer. What you know passes. Where you're wrong, it stops.
4. Answer once more after the hint. Hold the conclusion until you actually need it.

“Decide the next step from my answer” makes the conversation follow your understanding. Write only “one question at a time,” and AI may still march down a pre-listed quiz and ignore a wrong answer.

One-at-a-time is slower. For one date or one definition, just asking for the answer is faster. Use this when the material needs you to judge, calculate, or explain — because you need to know which step the reasoning broke on.

This is teaching experience. The page doesn't write it as an experimental result. For other ways to check understanding besides answering, see Have AI grill me like a layperson. For why smoothness makes people overrate mastery, see Let AI check: do I actually get it.