Does the Feynman technique work?

It’s a form of self-explanation and retrieval practice, two of the best-supported study habits: explaining an idea without notes shows you exactly what you don’t understand yet.

What’s a good AI prompt for the Feynman technique?

Tell it you’ll explain the idea as if to a beginner, that it shouldn’t correct you until you finish, and that it should then ask a beginner’s questions and point out what you skipped, without explaining it all for you.

Should I do the Feynman technique out loud or in writing?

Either works; out loud is faster and harder to fake, because every step has to be put into words.

How to Use the Feynman Technique with AI, Out Loud

The Feynman technique, named after the physicist Richard Feynman, is a way to test whether you understand something: explain it in plain words, as if to a beginner. Where you can’t, you’ve found a gap. AI makes a good listener for this, as long as you do the explaining and it does the asking.

The four steps

Pick one idea. Small enough to explain in a few minutes: the chain rule, how a hash map handles collisions, why enzymes speed up reactions.
Explain it simply. Out loud, without notes, in words a beginner would follow. No jargon you couldn’t define.
Find the gaps. Notice where you hesitate, wave your hands or reach for a term you can’t unpack.
Go back and simplify. Return to the source for those parts only, then explain the whole thing again.

Why out loud works better

In your head, a half-understood step feels fine. Spoken, it falls apart. Explaining to yourself is one of the better-supported study habits in learning research (it’s called self-explanation), and speaking forces you to put every step into words.

Using AI as the listener

The trap is letting the AI explain it to you instead: that feels like learning but skips the test. Set it up so you talk first:

I’m going to explain [the idea] as if to a beginner. Don’t correct me until I finish. Then ask me the two or three questions a curious beginner would ask, and tell me which step I skipped or got wrong. Don’t give me a full explanation unless I ask.

Then answer its questions, out loud again. Repeat until you can explain it start to finish without a stumble.

Common mistakes

Reading the AI’s explanation and calling it understood. Close it and explain it back.
Picking an idea that’s too big. “Thermodynamics” isn’t one idea; “why entropy increases” is.
Only doing it once. Explain it again a day later, without notes.

Doing it with Pen Buddy

Pen Buddy is built around talking it through. Press one button and explain your step: Buddy listens, follows the page you’re on, and asks what you’d try next, answering in a sentence or two. On iPad, work it out by hand and ask “is this right?”. Each session leaves review questions in your notes, so you can explain it again tomorrow. See studying with Buddy.

How to use the Feynman technique with AI