How to Get Claude to Teach You a New Topic

To get Claude to teach you a topic, describe what you want to be able to do, give your starting level, and ask for a short explanation followed by a practice question. Answer before requesting the solution. Then check the explanation against a reliable source and use your mistakes to choose the next exercise.

You can request this routine in an ordinary conversation; the learning goal matters more than a special prompt. fenbs can keep your progress, recurring misconceptions, and next lesson outside the chat so a later session starts with the work you actually completed.

How should you begin the first lesson?

1. Choose a narrow goal you can demonstrate, such as explaining a spreadsheet formula or interpreting a simple chart.
2. Tell Claude what you already know and provide a course outline, textbook excerpt, or other trusted starting material.
3. Ask for a few diagnostic questions before it decides where to start.
4. Request one short lesson, one worked example, and one exercise you must attempt before seeing the answer.

Replace teach me statistics with help me choose between mean and median when summarizing a small set of salaries. The second request creates a clear test of understanding. You should eventually be able to explain the choice and apply it to a different dataset without relying on the assistant.

Anthropic’s education introduction↗ (opens in a new tab) describes guided questions and attention to core concepts as part of its learning approach. You can ask for those teaching behaviors explicitly. The introduction to Claude AI explains the assistant if you are starting from scratch.

What should your teaching prompt include?

Give the topic, starting point, practical purpose, available time, and source material. Tell Claude how much help you want before the answer. Ask it to avoid moving forward merely because you say an explanation makes sense; require a demonstration of the idea instead.

Help me learn to choose mean or median for a workplace report. I can calculate an average but do not understand when it misleads. Use the source excerpt below. Ask two diagnostic questions, then teach one idea at a time. Give me an example and wait for my attempt at a new problem. Offer a hint before a full answer. Flag anything the source does not establish. End by recording what I can do and what I should practice next.

Ask for unfamiliar terms to be explained in one sentence. If the explanation is still confusing, request a concrete example or a different representation, such as a small table. A simpler explanation should preserve the important conditions rather than remove every qualification.

Change the prompt when you discover a useful learning preference. You might need shorter examples or more contrast between similar ideas. Keep a tested instruction instead of collecting long prompts you have never used. This is a practical routine, not a promise that any single prompt can teach every subject.

How do you make practice do more than repeat the example?

Attempt the exercise before opening the explanation. Write your answer and your reason, even if you are uncertain. Ask Claude to respond first to the reasoning that needs work, then give a hint. That makes the exchange more useful than receiving a complete worked solution immediately.

• Ask for a new example with different surface details but the same underlying idea.
• Explain the concept in your own words, then have Claude point out a specific gap.
• Compare a correct and an incorrect answer and explain why they differ.
• Try a short problem without assistance at the start of the next session.

For the salary example, try a small group with one unusually high value, then a group without it. Explain why the summary changes. If you can only reproduce the exact wording of the lesson, ask for another task that requires using the concept in a different setting.

Save the specific misconception you corrected, rather than labeling yourself bad at math. fenbs keeps Lessons learned and learning progress, which gives the next connected assistant a useful starting point for practice.

How can you check whether the explanation is accurate?

Choose a trusted reference for the subject and compare important explanations with it. Use your course materials, a textbook, or the documentation from the organization responsible for the tool you are learning. Open a cited source and check that it supports the precise claim.

Anthropic warns that Claude’s answers can be misleading or incorrect↗ (opens in a new tab). Confidence and fluent wording do not establish accuracy. If Claude and your course disagree, show the exact passage and ask it to identify the difference before you accept a correction.

You can also ask Claude to separate what the supplied source says from its own example or inference. Anthropic’s guidance on reducing hallucinations↗ (opens in a new tab) recommends allowing uncertainty and checking claims against sources. This supports your review; it does not replace it.

What should you do when the lesson is too hard or too easy?

If you are stuck, identify the earliest step you cannot explain. Ask for the prerequisite behind that step and a smaller exercise. Repeating a long explanation more slowly is less useful when the real gap is an unfamiliar term or operation the explanation assumes you know.

If you can solve the exercises easily, change one condition and explain what happens. Ask for a case where the simple rule does not apply. Keep the difficulty connected to your goal: a workplace report may require sound interpretation before it requires a formal mathematical proof.

This page covers the general learning routine. For a specialized goal, use the guides to language practice or learning to code with Claude. The distinction from teaching Claude matters: here you are developing your understanding. To change how the assistant performs your work, use the guide to teaching Claude a task.

How to do this with fenbs

1. Create a free account and add a project for a specific learning goal.
2. Add tasks for demonstrable skills, such as choosing a summary statistic and explaining the choice.
3. Keep the trusted reference links and useful context notes with the project, and save the teaching prompt in Skills and prompts.
4. Record a lesson when an exercise reveals a misconception: what happened, what you learned, and what to do differently.
5. Save Where we left off with what you demonstrated independently, what still needs practice, and the first exercise for next time.

fenbs does not grade your knowledge or certify competence. For a one-off explanation, a conversation and your own notes may be enough. Use a project record when you want progress and corrections to stay available across sessions or assistants.

What counts as progress at the end of a session?

Record what you can now do without help, with an example. Then record the remaining gap and the next check. Completed the lesson is less useful than correctly chose median for one uneven dataset; still need practice explaining why the mean changes.

Keep learning alongside real work when it helps. A designer exploring technical tasks can use the introduction to Claude Code to understand that separate tool. Browse AI at work when you want another practical problem to learn through.