Let AI Teach Me · Self-check

Have AI grill me like a layperson

Have it grill you with only junior-high knowledge. After this page you can run one concept and see which question you stall on.

“Twenty years later it's that much more because… because it grows exponentially.” You stop there.

Saturday, 4 p.m. You're at the table explaining compound interest to your eighth-grade niece. Ten minutes ago you finished a six-minute explainer. You can recite the definition, and on paper you got 100 becoming 105 in year one, 110.25 in year two. Both numbers are right. You feel like you've got it.

She asks first: “Who pays the extra twenty-five cents in year two?” You point at the first year's 5 and say it stays on the account and earns interest too. That one goes smoothly.

Then she slides the calculator over. At 5% a year: about 163 after 10 years, about 265 after 20. She asks: “Why did the second decade add 102, and the first only 63?” You say that half-sentence from a moment ago.

“Compound interest is interest on interest. Principal and old interest make new interest together, so the longer it runs, the stronger it gets — because that's exponential growth.”

She follows at once: “What does exponential growth mean? Why does it make the second decade grow faster?” You reopen the article. Your last two sentences are almost the article's own words. The real question still has no answer. You can recite the definition and compute two years from the example — that's why you judged yourself done.

Why you stall even after you can recite the definition

Reciting uses the sentence you just stored. The article said “the longer it runs, the stronger compound interest gets,” and you can say the same. As long as the original words are still there, the sentence keeps going.

Explaining is another job: you have to know which process each word points to. “Exponential growth” here has to land on that larger total each year: year two's 5% grows from 105, year three from 110.25. Each year's starting line already carries what grew before. If one word only fetches another word, the listener still doesn't get the process.

Reciting means you remembered the sentence. Explaining means you know what each word in it points to.

The AI can hold the question it doesn't get

A real listener often nods with your tone, or feels rude interrupting. The AI can stay inside the boundary you set, chase one unheard bit at a time, and point at the term you tossed back — as-is.

  1. Let it listen with only junior-high knowledge, and press down the background both sides would usually share. “Exponential growth” cannot count as a ready-made answer.
  2. Have it interrupt the moment it doesn't get it, so the gap lands on the sentence you just said — not piled up into a vague “please explain that again.”
  3. Ban it from finishing your sentence. A fluent explanation has to come from you. If it ghostwrites, you're checking whether it can talk, not you.

Question three already blocked “exponential growth.” If you toss that word back, versus walking through a year-by-year thickening — what different reaction does the layperson give? We start from that question.

Try it · let the layperson keep asking until it clicks
I'm explaining compound interest to someone with only junior-high knowledge
Question three is here. How will you answer
Switch the answer, then walk question three again.
1 · Compound vs ordinary interest
2 · Who pays those twenty-five cents
3 · Why twenty years is such a bigger gap
The follow-up it just asked
After play, the layperson's three follow-ups show up here in order.
How you answer, and whether the layperson got it (sketch)
The sentence you gave — and whether it counts as plain talk.
Pick an answer style, then play to see how the layperson follows up.
Why the first two questions were answerableYou can point at the numbers and the first two years.
Which gap question three exposesSwitch the answer and see if the layperson can keep computing.
Teaching sketch: the follow-ups and verdicts are simplified demo text, not a real model output.
Paste this whole block to the AI

How to study so it actually goes in already covered teaching a layperson. The block below writes the layperson's knowledge line, when to interrupt, and whether it may finish your words. Swap 【concept】 for what you just learned, then start from sentence one.

Copy the whole block

I'm going to explain 【concept】 to you. Play a layperson with only junior-high knowledge, hearing this for the first time. After each sentence, ask about one thing you didn't get. Interrupt at once — don't wait till I finish. If my answer uses a specialist word you might not know, keep asking about that word. Don't fill in, rewrite, or finish my sentences, and don't hint. When you're sure a junior-high student could follow my words and invent an example, tell me this part is clear. Let me start now.

What each of these three limits is blocking
  1. Junior-high knowledge only blocks a default background that's too rich. The AI knows “exponential growth” and the compound-interest formula. If you don't press the line down, it walks right over the layer you never explained.
  2. Interrupt when it doesn't get it blocks questions stacking up. A follow-up after every sentence shows you which word the gap starts at.
  3. No finishing your sentence blocks the AI polishing a half-sentence into a full answer. It can flag that “exponential growth” is still unexplained. The year-by-year thickening has to come from you.

Whoever finishes the explanation is whose understanding you're checking.Let the AI only listen, interrupt, and point at terms. The output stays on your side.

This is a workable check, not a controlled-experiment result. To have the AI ask one question at a time, see Have AI ask me only one question. If you stall because a prior concept is missing, go back to the four-step breakdown.

Being able to explain is only the first gate. If the conditions change and you can still reassemble the steps, see Can you adapt: solve a variant.

Wherever the layperson stops, that's the word you restart from.