The Feynman technique, one bullet at a time

Richard Feynman kept a notebook titled Notebook of things I don't know about. The technique that carries his name is not something he wrote down as a method — it is reconstructed from how he worked, and from a habit he was blunt about: if he could not explain something in plain language, he did not consider that he understood it.

That is the whole idea. The four steps are just a way of forcing the test.

The four steps

1. Explain it plainly. Write out the thing as if to a bright twelve-year-old. No jargon, no borrowed phrasing, no "as we know". Full sentences.

2. Find where you stumbled. You will stall somewhere. You will also catch yourself using a word you cannot define, or writing a sentence that sounds like the textbook rather than like you. Those are the gaps.

3. Go back to the source. Not to re-read everything — to answer exactly the gaps from step 2.

4. Say it again, shorter. If the second version is shorter and clearer, you have learnt something. If it got longer, you have not; you have added qualifications to hide the hole.

Why plain language is such a good test

Jargon compresses. That is what it is for, and among people who share it, it works. But compression also lets you carry an idea around without ever having unpacked it — you can use the phrase correctly, in the right contexts, and have no model underneath.

Writing for a twelve-year-old removes the compression. Every borrowed term has to be replaced with the thing it stands for, and any place where there is nothing behind the term becomes a sentence you cannot finish. You have not suddenly got worse at the subject; you have found out where you already were.

This is also why it is uncomfortable, and why most people do step 1 and stop. Step 2 is the method.

Mapped onto an outline

The template is one branch per thing you are trying to understand, with the four steps as children. Two of them are documents rather than lists:

• 1. Explain it plainly is a markdown page — zoom in and write prose. Bullets let you skip the connective tissue between ideas, and the connective tissue is precisely what this step is testing.
• 2. Where I stumbled is a list, because gaps are a list.
• 3. Back to the source holds what you looked up and what it said.
• 4. Say it again, shorter — and if it is genuinely shorter, that bullet is the note worth keeping.

Keep the branches. Six months of them is a map of what you actually understand, which is a rarer thing to own than a pile of notes about what you have read.

• The Feynman techniqueNamed for Richard Feynman: explain it simply, and the gaps announce themselves.
  • The thing I am trying to understandName it in the bullet above — one subject per branch.
    • 1. Explain it plainlyA document — zoom in to write.
    • 2. Where I stumbledThe sentences you could not finish, and the words you used without meaning them.
      • The bit I hand-waved
    • 3. Back to the sourceNot to re-read it all — to answer exactly what is in step 2.
      • What I looked up, and what it said
    • 4. Say it again, shorterIf it got shorter and clearer, you have learnt something. If it got longer, you have not.

Folded bullets open with a click; every line is yours to rename or delete. Open it →

How to run it

Twenty minutes, one thing at a time. Write step 1 in one sitting, without looking anything up — looking things up mid-explanation is how you smuggle the source's phrasing back in.

Then read what you wrote, hunting specifically for two tells: a sentence you could not finish, and a sentence that sounds like the book. Both go into step 2.

Do it again a week later, from scratch, without reading the old version. The gap between the two attempts is the clearest feedback on learning that you will get anywhere.

Where people go wrong

Explaining to an expert instead of a child. The moment you imagine a knowledgeable reader, you start writing in shorthand again, and the test switches itself off.

Treating step 4 as tidying. The point of the second, shorter version is that it should be shorter. If you find yourself adding "of course, strictly speaking…", you have found another gap, not a nuance.

Only doing it for things you already know. It is a diagnostic. Point it at the thing you have been nodding along to in meetings for a year.