The Feynman Technique: The Learning Strategy That Actually Works
By AI Anyone Team · 2026-08-02 · 7 min read · Learning Tips
Explain it simply, find where the explanation breaks, then go back to the source. The four steps, a worked example, and the one most people skip.
Say out loud, right now, how a bicycle stays upright. Not the short version: the whole mechanism, in plain words, for somebody who has never ridden one.
Most people last about eight seconds. "The wheels spin, spinning wheels resist tipping, and you sort of correct with the handlebars." So what is the correction doing? Why does the bike fall the moment it stops? You have ridden one for years and cannot produce the explanation. That gap, between the feeling of understanding and the ability to generate it on demand, is what the Feynman technique is built to catch.
The four steps
1. Pick one concept and put it on a blank page
One concept, not a chapter. "How a bicycle stays upright," not "the physics of motion." Write the name at the top of an empty page and close everything else. The blank page does real work: anything you can see, you will lean on, and leaning on it is what hid the gap in the first place.
2. Explain it as if the listener is a beginner
Say it out loud, as though the listener is a curious twelve year old, interested and completely uninformed. Two rules: no term you cannot define on the spot in ordinary words, and no skipping. If you say "gyroscopic," you owe a plain definition of gyroscopic. If you say "and then it balances," you owe the mechanism by which it balances.
Speaking beats writing, because speech has a tempo you cannot pause. On the page you can hover over a hard sentence and never notice you stalled. Out loud, the stall is audible.
3. Mark exactly where the explanation breaks
This is the step that does the work, and the one almost everybody skips. Go back through what you produced and mark every place it went soft. There are four kinds of soft:
- The vague verb. "It stabilizes." "The forces balance out." A verb doing the job of a mechanism.
- The borrowed word. Using the name of the thing as the explanation for the thing. A bike stays up because of gyroscopic stability.
- The jump. Two sentences that are both true, with an unstated step between them.
- The trail-off. "And so on." "You get the idea." "Something like that."
Mark them, do not fix them yet, and do not talk yourself past them. A gap you argue with is a gap you keep.
4. Go back to the source, then simplify
Now open the book, the paper, the lecture. Not to reread from the top: to answer the specific questions you marked. Then rewrite the explanation from scratch, shorter than the first, with the repairs folded in so cleanly that a listener could not find the seams. Repeat until nothing is marked.
Why re-reading feels productive and changes nothing
Re-reading and highlighting feel like learning because they produce a real sensation: recognition. The sentence looks familiar, it parses easily, and the brain reads that ease as competence. That is the illusion of fluency, and it imitates understanding well enough that few people audit it.
Recognition and recall are different abilities. Recognition needs the material in front of you. Recall has to rebuild it from nothing. Every exam, interview, and actual use of what you know is a recall task, and re-reading rehearses the other one. Highlighting is worse than neutral, because it turns a decision into a gesture: you mark a sentence important, feel the satisfaction of having dealt with it, and what you did was move a pen.
The Feynman technique reverses the flow. Nothing goes in. Something has to come out, in a form that survives contact with a listener who does not already know the answer. That is harder than reading, it feels worse, and the discomfort is the signal that it is working.
Worked example: how a bicycle stays upright
First pass, the honest version most people produce:
"A bike stays up because the wheels are spinning. A spinning wheel wants to keep spinning in the same plane, so it resists tipping over. That is why it falls when you stop. And if you start to lean, you correct with the handlebars."
Step 3, marking it up. Three breaks.
- "Wants to keep spinning in the same plane" is a borrowed word doing no work. What force is that, and how strong is it?
- "Resists tipping over" hides the question inside the answer. Resists it how, and by enough?
- "You correct with the handlebars" is the trail-off. Correct how? Five words are carrying the entire mechanism.
Back to the source, and two things come back. The gyroscopic effect of a bicycle wheel is real but modest at riding speed, nowhere near strong enough on its own to hold up a rider. And engineers have built test bicycles with counter-rotating wheels to cancel that effect out. Those bikes still balance. The answer you have carried since childhood is not the answer.
The mechanism is steering. When the bike leans left, the front wheel turns left. Sometimes the rider does that, sometimes the bike does it unaided: the front tire meets the ground slightly behind the point where the steering axis would, so a lean makes the wheel flop into the lean by itself. Turning left curves the path left, which swings the tires back under the center of mass. The bike is not resisting the fall. It is steering underneath it, several times a second.
Simplified pass:
"A moving bike stays up by steering into its own fall. Lean left, the wheel turns left, the path curves left, the tires come back under you. It is a broom balanced on your palm, except the palm is the whole bicycle. Stop moving and you cannot steer, so you cannot balance."
Four sentences, no jargon, and it predicts what the first version could not: why balancing at a dead stop is nearly impossible, why a shoved riderless bike coasts upright, why steering into a skid is the right instinct.
The hard part is that step 3 wants a second person. Read your own explanation back and you will nod straight through the sentence doing nothing, because you know what you meant. A listener does not, and asks. If nobody in your life wants to hear about bicycle geometry at 11:00 PM, hand your explanation to Richard Feynman and let him ask the follow up. He was famously unwilling to let a vague word slide past.
The step everybody skips
Steps 2 and 4 are pleasant. You know things, you say them, they sound reasonable, and rewriting is craft. Step 3 is where you sit with a sentence you wrote and admit it means nothing.
The reflex at that moment is to patch instead of return. You swap in a heavier word, add a qualifier, gesture at difficulty ("it is more complicated than that, obviously"), or decide the gap is a detail. Every one of those leaves the hole where it was and paints over the top of it.
The test is mechanical. Could you defend the sentence against someone asking "why?" three times in a row? If the third why is where you would say "that is just how it works," go back to the source. Not the summary, not your notes, the thing that actually explains it.
It works on abstractions too
Ask what an integral is and you usually get "the area under a curve." That is a picture, not a mechanism, and it collapses the moment the thing being integrated is not a shape. The rebuilt version: you are adding up an enormous number of thin slices, each one a value times a tiny width, and the integral is the number those running sums settle toward as the slices get thinner. Area is one thing that happens to compute, not the definition. Better by a simple test: it tells you what to do, where the first only told you what to picture.
Where to start
Pick something you use constantly and have never had to say out loud. How the tax withholding on your paycheck is calculated. What your job's core metric measures. Give it eight sentences before you allow yourself to look anything up, and watch for the exact word where your voice slows down. That word is your lesson plan.
Then do the part you cannot do alone: try the explanation on Richard Feynman and let him push on the soft spots you were about to walk past.
