The Feynman Technique: How to Know If You Actually Understand Something

Richard Feynman had a specific test for whether he actually understood something. He would pick a technical concept — a mathematical proof, a physical mechanism, a derivation — and try to explain it to a class of freshmen, or sometimes to himself out loud as if to a freshman. If he got through the explanation cleanly, he'd internalised it. If he got stuck — if at some point he had to retreat behind jargon, hand-wave over a step, or simply repeat a textbook phrase without owning it — he knew he hadn't. He'd go back to the source and work on that specific gap until the explanation held together. He called this "the first principle" of learning, though it's been widely popularised as "the Feynman Technique."

Most of the online treatment of this technique makes it sound simpler than it is, and misses what actually makes it work. The method isn't "pretend you're teaching a child." It's "use the act of explanation as a diagnostic instrument for gaps in your own understanding." The child is incidental. What matters is the ruthless specificity about what you can and cannot explain.

The Four Steps, Done Properly

Step 1: Write the concept's name at the top of a blank page

Not in a notebook you'll archive. A fresh page, one concept per page, nothing else on the page. The physical act of starting blank forces you to generate the explanation from memory, not reference.

A concept can be: a single idea (Bayesian updating, opportunity cost, Conway's Law), a process (how an internal combustion engine converts fuel to motion, how a peer-review process works), or a specific claim you're trying to understand (why small teams ship faster than large ones, why interest rates affect asset prices in a specific direction).

The name at the top is your commitment. You're going to try to explain this specific thing.

Step 2: Explain it in plain language, as if to someone with no prior knowledge

Write the explanation in full sentences. Complete paragraphs. No notation, no jargon, no technical shortcuts. If you need to use a technical term, define it immediately in plain language before using it.

The audience is a specific person — usually Feynman suggested a bright 12-year-old, but what you really want is "anyone intelligent who has no domain knowledge." This is the critical calibration. If you imagine a peer, you'll fall back on shared vocabulary that lets you gloss over gaps. If you imagine a child, you might oversimplify to the point of distortion. The sweet spot: an intelligent adult who simply hasn't encountered this field.

Write until you either complete the explanation or get stuck.

Step 3: Identify where you got stuck — the honest part

This is the step most people skim over and it's the step that does the work. When you get stuck, you're going to feel one of several things:

• A specific word or concept you reached for and realised you couldn't define in plain English.
• A step in the explanation where you jumped — there's a logical move between two propositions that you know the end of but can't derive.
• A hand-wave where you found yourself writing "and then basically" or "so the idea is" or other filler that bridges a gap you haven't actually closed.
• A point where you started quoting the textbook verbatim — a tell that you memorised the words but don't own the claim.

At each stuck point, write down what specifically you don't understand. Not "I'm stuck on Bayesian updating." That's too vague. "I can't explain why the prior matters when the likelihood is high, without just restating Bayes' theorem" — that's a real gap.

The specificity is the diagnostic value. Most people know they don't fully understand a topic. Almost nobody can state, at the level of a specific missing step, what exactly they don't understand. The Feynman Technique produces that specificity, which is what makes the next step possible.

Step 4: Go to the source with a specific question and fill the gap

Not "read the textbook chapter again." Go to the specific explanation of the specific thing you're stuck on. Read the one page. Understand the one move. Try to articulate it in your own words.

Then — and this is crucial — go back to step 2. Write the explanation again, with the new piece included. Do not skip straight to verifying that you've learned it. Write the full explanation again. More often than not, plugging one gap reveals another one, deeper in the explanation, that you didn't see the first time because you hadn't got to it. The second pass is often where the real understanding assembles.

Why This Works — the Mechanism

The technique works because of the specific asymmetry between recognition and understanding. When you read something, your brain does pattern-matching — it sees familiar words in a familiar structure and returns the judgement "yes, I'm familiar with this." Familiarity feels like understanding. It isn't.

The gap is exposed only when you try to generate the explanation yourself, from memory, in novel words. At that point, the pattern-matching can't help you — you need the underlying model, and the absence of the model becomes immediately visible. The friction you feel at the stuck point is your brain reporting that there's nothing there to draw on.

Cognitive psychology research on this is known as the illusion of explanatory depth — work originally by Rozenblit and Keil at Yale in 2002. They showed that people consistently overestimate how well they understand everyday objects and concepts until they're asked to explain them. A typical subject, asked to rate their understanding of how a zipper works, gives themselves a seven out of ten. After being asked to write an explanation of how a zipper works, they rate their understanding at three or four. Nothing changed except that they tried to explain. The attempt revealed the gap.

The technique is a structured way to deliberately surface this gap, and then close it. It turns the illusion of understanding into actual understanding, one specific gap at a time.

The Specific Topics Where This Is Most Useful

The technique pays its highest dividend on material where you suspect you've been nodding along without really owning it. In practice, this is most of what we consume.

• Technical claims you repeat but can't derive. You say "compound interest is the eighth wonder of the world" but can you derive the actual math that makes compounding asymmetric? If not, explain it on paper.
• Business frameworks. You use terms like "unit economics," "CAC/LTV," "operating leverage" in meetings. Can you explain each one without using other jargon? If not, the concept is borrowed, not owned.
• Intellectual positions you hold. You have views on monetary policy, immigration, AI safety. Can you explain the strongest version of your view to a thoughtful person who disagrees, without straw-manning the other side? If not, the view is a tribal signal, not a conviction.
• Core concepts in your own professional domain. This is the most uncomfortable one. You'd be surprised how many senior professionals can't cleanly explain the central claims of their own field. Try with yours.

The Mistake of Using It as a Performance

A specific failure mode: treating the Feynman Technique as a content-creation exercise. You "Feynman" a topic, post the explanation online, and treat the posting as evidence you understood it.

This is subtly different from the real practice. In the real practice, nobody but you sees the explanation. The audience exists only in your head. When you know the explanation is going to be published, the writing gets more polished — more jargon, cleaner transitions, better structure — and in that polish, the gaps get papered over rather than exposed.

If you want to use the technique as a teaching exercise, do it privately first. Write the private version where you're allowed to get stuck, to show your working, to admit in writing "I don't know what happens here." Then, if you want, clean it up for publication. The private version is where the learning happens.

The Boring Version of "Teach to Learn"

The technique is sometimes sold as "the best way to learn something is to teach it." This is true but misleading. Teaching someone else is useful mostly because they'll ask questions you didn't anticipate, which exposes gaps you didn't know you had. The Feynman Technique captures most of this benefit without needing another person — you're pre-asking yourself the questions they would have asked.

The advantage: it's scalable. You can't find a student every time you want to test your understanding. You can always find a blank page. The private version of teaching, done with discipline, produces about 80% of the learning benefit at zero scheduling cost. It becomes a habit, like exercise — something you do for 30 minutes a week as part of the normal maintenance of a professional mind.

The Final Test

When you've genuinely understood something using this technique, two things should be true. You can explain it cleanly without reference material. And — equally important — you can explain it at multiple levels of abstraction: a one-sentence version, a one-paragraph version, a 2,000-word version. The ability to compress and expand the explanation at will is the signature of real understanding. Someone who has memorised a textbook explanation can deliver the 2,000-word version and nothing else. Someone who owns the concept can move between lengths fluidly, because they're drawing on the underlying model rather than the stored words.

Feynman used to demonstrate this in lectures. He could explain quantum electrodynamics in one sentence to a journalist, in a five-minute version to a physics undergrad, and in full detail to graduate students, and each version was genuinely about the same underlying physics. This is what understanding looks like. Most of us, most of the time, don't have this for most of what we claim to know. The technique is a tool for closing the gap, one concept at a time.