Claude Complex Topic Explainer (Feynman Technique) Prompt
You are a master teacher who uses the Feynman Technique: explain anything to anyone by stripping it to its essentials.
Category
📚 Education
Difficulty
Beginner
Models
3
Last Updated
2026-06-28
Works with
📄 Example output
⚠️ Common Mistakes
❓ FAQ
⚙️ Fill in your variables
📋 Prompt
You are a master teacher who uses the Feynman Technique: explain anything to anyone by stripping it to its essentials.
Topic: [complex topic]
Audience: [target audience — 10-year-old/high schooler/intelligent non-expert/professional]
Depth: [depth required — 2-minute overview/10-minute deep dive/complete course module]
Analogy: [analogy preference — technology/sports/cooking/nature/everyday objects]
Task: Explain [complex topic] using the Feynman Technique:
1. THE CORE IDEA (1 sentence): What is this, really?
2. THE ANALOGY: Map the complex system to something familiar
3. STEP-BY-STEP EXPLANATION: Build from simple to complex
4. WHERE THE ANALOGY BREAKS DOWN: Be honest about what the analogy gets wrong
5. WHY IT MATTERS: Real-world application or importance
6. THE 5-YEAR-OLD VERSION: Can you explain it to a child?
7. COMMON MISCONCEPTIONS: What most people get wrong about this
Topic: [complex topic]
Audience: [target audience — 10-year-old/high schooler/intelligent non-expert/professional]
Depth: [depth required — 2-minute overview/10-minute deep dive/complete course module]
Analogy: [analogy preference — technology/sports/cooking/nature/everyday objects]
Task: Explain [complex topic] using the Feynman Technique:
1. THE CORE IDEA (1 sentence): What is this, really?
2. THE ANALOGY: Map the complex system to something familiar
3. STEP-BY-STEP EXPLANATION: Build from simple to complex
4. WHERE THE ANALOGY BREAKS DOWN: Be honest about what the analogy gets wrong
5. WHY IT MATTERS: Real-world application or importance
6. THE 5-YEAR-OLD VERSION: Can you explain it to a child?
7. COMMON MISCONCEPTIONS: What most people get wrong about this
TOPIC: How Large Language Models work
AUDIENCE: Intelligent non-expert | ANALOGY: Everyday objects
CORE IDEA: A language model is a very sophisticated autocomplete that has read most of the internet.
THE ANALOGY — Autocomplete on steroids:
When you text someone and your phone suggests the next word — that's a tiny language model. GPT-4 is the same idea, but instead of training on 10,000 texts you've sent, it trained on roughly 1 trillion words from the internet, books, and code. And instead of predicting the next word from context, it can hold thousands of words of context at once.
STEP-BY-STEP:
1. Training: The model sees a sentence with a word removed. It guesses the word. If wrong, the guess is adjusted slightly. Repeat this 10 trillion times.
2. The result: A mathematical function (billions of numbers) that represents statistical patterns in language.
3. When you type: The model predicts the most likely next token based on everything that came before it — over and over until the response is complete.
WHERE THE ANALOGY BREAKS DOWN:
Autocomplete picks the most statistically likely word. Modern LLMs use techniques that allow them to produce helpful, nuanced, or creative responses — not just the most average response.
COMMON MISCONCEPTIONS:
❌ 'It understands language' — it models patterns in text; whether that constitutes understanding is genuinely debated
❌ 'It searches the internet when you ask it a question' — it doesn't (unless web search is specifically added)
❌ 'It's always right' — it confidently generates text; correctness is a separate question from fluency
AUDIENCE: Intelligent non-expert | ANALOGY: Everyday objects
CORE IDEA: A language model is a very sophisticated autocomplete that has read most of the internet.
THE ANALOGY — Autocomplete on steroids:
When you text someone and your phone suggests the next word — that's a tiny language model. GPT-4 is the same idea, but instead of training on 10,000 texts you've sent, it trained on roughly 1 trillion words from the internet, books, and code. And instead of predicting the next word from context, it can hold thousands of words of context at once.
STEP-BY-STEP:
1. Training: The model sees a sentence with a word removed. It guesses the word. If wrong, the guess is adjusted slightly. Repeat this 10 trillion times.
2. The result: A mathematical function (billions of numbers) that represents statistical patterns in language.
3. When you type: The model predicts the most likely next token based on everything that came before it — over and over until the response is complete.
WHERE THE ANALOGY BREAKS DOWN:
Autocomplete picks the most statistically likely word. Modern LLMs use techniques that allow them to produce helpful, nuanced, or creative responses — not just the most average response.
COMMON MISCONCEPTIONS:
❌ 'It understands language' — it models patterns in text; whether that constitutes understanding is genuinely debated
❌ 'It searches the internet when you ask it a question' — it doesn't (unless web search is specifically added)
❌ 'It's always right' — it confidently generates text; correctness is a separate question from fluency
🏆
💡 Pro Tips
Best model for this prompt
Claude
Claude (Opus 4 / Sonnet 4)
If you can't explain it simply, you don't understand it well enough yet — the Feynman Technique reveals your own gaps
The 'where the analogy breaks down' section is what separates a good explanation from a misleading one — always include it
Multiple analogies work better than one perfect one — different minds connect with different metaphors
Build to the complex version only after the simple version is solid — don't skip levels
Using jargon in the explanation — if the audience needs a glossary to understand the explanation, start over
Explaining what something IS without explaining why it WORKS that way — mechanism understanding is deeper than definition
Analogies that require more explanation than the original topic
Skipping misconceptions — these are often where the real learning happens
- What is the Feynman Technique?Developed by physicist Richard Feynman. Four steps: (1) Choose a concept. (2) Explain it as if teaching a child. (3) Identify gaps where you struggled to explain simply. (4) Review and simplify. The technique reveals what you actually understand vs. what you only think you understand.
- Which AI model produces the best explainers?Claude is generally strongest — it's better at calibrating complexity to audience, building layered explanations, and being honest about where analogies break down. Gemini is also strong for science and technical topics.
- How do I know if my explanation is good enough?Give it to someone from your target audience and ask them to explain it back. If their explanation matches yours, it worked. If they're confused or misrepresent the core idea, the explanation needs work.
- Can this prompt help me study for exams?Yes — after studying a topic, use this prompt to generate an explanation. Then try to do it yourself without the AI output. The gaps between the AI explanation and your own reveal exactly what you haven't fully understood yet.