The Original Feynman Technique, and Its Limitation
Physicist Richard Feynman’s famous learning method is simple: pick a concept, explain it in plain language as if teaching a total beginner, find the gaps where your explanation gets vague or you reach for jargon, and go back to the source material to fill those gaps. It works because teaching forces you to notice what you don’t actually understand. The catch is that it needs an audience, and most people don’t have a patient beginner on standby every time they want to learn something. So the technique quietly gets skipped, and people go back to just re-reading notes, which feels productive but barely improves retention.
Turning AI Into a Practice Student
An AI chatbot solves the audience problem, but only if you set it up correctly. The trick is to stop asking the AI to explain things to you and instead assign it the role of a curious, slightly skeptical student who asks follow-up questions. Give it a prompt like: “I’m going to explain [topic] to you. You’re a smart beginner who doesn’t know this subject. After I explain, ask me one clarifying question that a real beginner would ask, and point out anywhere my explanation used a term I never defined.” This flips the interaction from AI-teaches-you to you-teach-AI, which is where the actual learning happens.
A Simple 4-Step Workflow
First, pick one specific concept, not a whole subject; “how compound interest works” beats “finance.” Second, write or speak a two-to-three paragraph explanation without looking anything up, then paste it to the AI with the student prompt above. Third, read its follow-up question and try to answer it from memory before searching for the answer; if you can’t, that’s your gap, and it’s exactly what you needed to find. Fourth, go find the real answer, rewrite your explanation to include it, and run the loop again until the AI’s questions stop exposing new holes.
Why This Beats Studying Alone
Re-reading and highlighting create a false sense of fluency because recognizing information feels like knowing it. Explaining out loud to an AI that pushes back removes that illusion immediately, since vague spots become obvious the moment you try to phrase them precisely. It also costs nothing, works at 11 p.m. when no study partner is awake, and scales to any subject from tax brackets to guitar scales. The goal isn’t to let AI hand you answers; it’s to use AI’s patience and follow-up questions to do what a good study partner would do, so the gaps in your understanding surface while you can still do something about them.
