5-Level Technical Explainer: Make Any Concept Land with Any Audience

The Prompt

Explain [CONCEPT] at exactly five levels:

**Level 1 — For an 8-year-old:** Use simple words, a relatable analogy from everyday life, and no jargon. One short paragraph.

**Level 2 — For a curious high schooler:** Introduce the real vocabulary but still anchor it to intuition. Can reference basic science or math. Two paragraphs max.

**Level 3 — For a college student in a related field:** Use proper terminology. Assume they know fundamentals but not specifics. Cover the key mechanism or idea in depth. Three paragraphs.

**Level 4 — For a working professional in tech:** Skip the basics. Focus on practical implications, trade-offs, and where this concept fits into real systems. Be direct and specific.

**Level 5 — For a domain expert:** Discuss nuance, edge cases, current open problems, and what the expert community debates about this topic. Reference specific papers, implementations, or known limitations if relevant.

Topic: [YOUR TOPIC HERE]

How to Use It

Replace [CONCEPT] and [YOUR TOPIC HERE] with the same topic. The repetition in the template is intentional — it reinforces to the model that all five levels should address the same subject. Examples:

• Explain how neural networks learn
• Explain what a database index does
• Explain how HTTPS encryption works
• Explain what venture capital is
• Explain quantum superposition
• Explain how a compiler works

Why the Five-Level Structure Works

Each level forces a different cognitive operation: Level 1 requires finding the best analogy, Level 2 requires identifying which vocabulary is load-bearing, Level 3 requires explaining mechanism without assuming prior exposure, Level 4 requires distilling to implications only, and Level 5 requires engaging with what isn't fully understood yet.

Models like Claude and GPT-4 handle this structure well because the level labels give explicit anchors rather than vague instructions like "explain this simply." The more specific the audience description, the more calibrated the output.

Practical Use Cases

• Technical documentation: Pull Level 3 for your main docs, Level 2 for the overview page, Level 1 for the marketing site.
• Onboarding new engineers: Use Level 3 or 4 to generate first-pass explanations of system components, then edit for accuracy.
• Stakeholder communication: Executives get Level 2 or 3. Engineers get Level 4. Academic collaborators get Level 5.
• Teaching: Paste Level 1 into a presentation intro slide, then build toward Level 3 across the session.
• Self-learning: If Level 5 contains terms you don't recognize, run the prompt again with those terms as the new topic.

Tips for Best Results

• Be specific in the topic: "database indexes" works better than "databases." The more constrained the topic, the sharper each level.
• Claude performs best on highly technical Level 5s for software and science topics. GPT-4o tends to produce better Level 1 analogies for abstract concepts.
• If Level 5 is too shallow: Add "cite specific papers or implementations" to the Level 5 instruction, or follow up with: Go deeper on Level 5 — include the names of key researchers, papers, or open problems in this area.
• Chain this prompt: Take the Level 3 output and run it through a "now write this as a blog post introduction" prompt to get a ready-to-use paragraph.

Example Output (Partial)

Topic: How a database index works

Level 1: Imagine you have a giant book with 1,000 pages and you want to find every page that mentions "dragon." You could read the whole book — that takes forever. Or you could check the index at the back, which lists "dragon" and points you straight to pages 14, 237, and 891. A database index does the same thing: it keeps a sorted shortcut list so the computer doesn't have to check every row when you search.

Level 4: An index is a separate data structure — typically a B-tree or hash table — maintained alongside the table. It trades write overhead and storage for O(log n) or O(1) read performance on indexed columns. The query planner decides whether to use an index based on cardinality estimates and selectivity; low-cardinality columns (like a boolean flag) often see full table scans even with an index because the overhead of the index lookup exceeds sequential scan cost. Covering indexes eliminate table heap access entirely by storing all queried columns in the index leaf nodes.