Tutoring and explaining a concept

The task

You want a concept explained at a level that helps you use it. That might mean a technical topic, a codebase pattern, a mathematical idea, or a domain you are new to.

Why the old approach is outdated

The old default was “explain step by step like I’m five.” That asks for maximum simplification, not useful calibration. It often removes the parts an adult learner actually needs.

Current models do better when you state your background, the task you need to do, and the depth you want. OpenAI and Anthropic both emphasize clear instructions, context, and examples over vague persona-style framing.

The current approach

Tell the model what you already know and what the explanation should help you do.

I'm a backend engineer with no frontend experience. Explain how React reconciliation
decides what to re-render.

Aim the explanation at debugging performance problems in an existing React app. Do not
start from "what is JavaScript" or turn this into a beginner React tutorial.

For nontechnical topics, the same pattern holds.

I understand basic personal budgeting but not bond funds. Explain why rising interest
rates can make an existing bond fund lose value. Keep the explanation practical enough
that I can read a retirement-account allocation page without getting lost.

For deeper understanding

Ask for a self-check instead of a staged lesson plan.

After the explanation, give me one question that would reveal whether I understood
the key mechanism. Do not give away the answer until I try.

For code, provide the code and the reason you care.

Explain what this function does and why it is structured this way. Assume I can read
TypeScript but do not know this codebase's conventions.

{{code}}

If the explanation itself is the deliverable, chain of thought and reasoning models explains how to ask for a useful visible explanation without demanding private scratch work.

What to avoid

• Using “like I’m five” as the default. State your real background instead.
• Asking for a rigid lesson plan before the model knows where you are stuck.
• Asking for code explanation without giving the code.
• Asking for hidden reasoning. Ask for the key idea, a worked example, and a self-check question.