Using AIs For Learning, tutoring and creating quizzes

For most of my life, "studying" meant reading a book, taking some notes, and hoping the material stuck. Over the past year, as I've built out a self-directed Learning and Thinking Bootcamp — memory, learning methodology, thinking frameworks, writing, psychology, and emotional intelligence, all stacked on top of each other — I've come to rely on AI models as something closer to a tutor than a search engine. That shift has changed how I study more than any single book on the reading list.

From Answer Machine to Tutor

The lazy way to use an AI is to ask it a question and copy the answer. The useful way is to treat it like a patient, well-read tutor who is available at 11pm on a Tuesday and never gets tired of your third follow-up question.

A few ways this plays out in practice:

• Explaining a concept in a different way. If I read a chapter of Deep Work or Hyperfocus and a section doesn't click, I can ask the AI to re-explain it using a different analogy, a worked example, or simpler language. Nine times out of ten, the second explanation is the one that lands.
• Socratic questioning. Instead of asking the AI to summarize a chapter for me, I can ask it to quiz me first, challenge my interpretation, and explain anything I get wrong. This forces me to reconstruct the idea myself before I get the "correct" version handed to me.
• Acting as a sounding board. When I'm working through something like Annie Duke's expected-value thinking, or trying to apply a thinking framework to a real decision, I can talk it through with the AI much as I might with a study partner. The difference is that the AI can keep asking questions, propose alternative interpretations, and help me connect the discussion to the underlying ideas and theories.
• Adjusting to my actual level. A tutor that has access to the context of what I've already studied can meet me where I am rather than defaulting to either a beginner's overview or a graduate seminar. I can say, "Explain this assuming I've already read Adler's How to Read a Book," and get a response calibrated to that.
• Following the weak point. This may be one of the most useful differences between an AI tutor and a static study guide. If I give a vague or partially correct answer, the AI doesn't necessarily have to move on to the next question. It can ask a narrower follow-up question and keep probing until the source of the problem becomes clearer.

That last point matters. Sometimes the problem is that I genuinely don't know the material. Sometimes I understand the general idea but cannot explain it precisely. Sometimes I know the concept but cannot apply it to a new situation. Those are different learning problems, and they call for different kinds of practice.

None of this replaces the primary source material. The book is still the book. What the AI adds is a second pass — a way to test whether I actually absorbed what I read, work through confusion interactively, and fill gaps without derailing my whole afternoon chasing down a tangent.

Retrieval Practice: Using AI-Generated Quizzes

The single biggest change to my study routine has been using AI-generated quizzes for retrieval practice, sometimes called active recall — pulling information back out of your head rather than simply re-reading it.

Re-reading feels like learning because the material becomes fluent and familiar. You look at a paragraph and think, "Yes, I know this." But familiarity is not the same thing as being able to retrieve, explain, or apply the idea later without looking at the page.

Quizzing forces a different test: Can you produce the information yourself?

My current process is simple:

1. Finish a section or chapter of whatever I'm studying or consolidating.
2. Ask the AI to generate a short quiz on that material — mostly short-answer and "explain this concept" questions rather than multiple choice, since multiple choice often lets you recognize the answer instead of retrieving it.
3. Answer without looking back at the text.
4. Have the AI evaluate my answers and flag anything shaky, vague, incomplete, or flat-out wrong.
5. Use the results to identify what needs further attention rather than automatically re-reading the entire chapter.

This has been especially useful during the "consolidation" phase of my bootcamp, where the goal isn't necessarily to create new content but to get existing material — such as my identity and psychology-of-belief article series — genuinely "down cold" rather than simply filed away as something I once read.

It has also pushed me toward thinking more seriously about spaced repetition. Retrieval practice is useful once, but the real advantage comes from returning to important material over days and weeks rather than successfully answering a question once and assuming it is permanently learned.

Ask for Retrieval, Not Recognition

"Quiz me" is a weak prompt. A better request specifies what kind of cognitive work you want to do.

For example:

"Ask me short-answer questions that require me to explain these ideas in my own words. Do not give me multiple-choice options. Ask follow-up questions when my answer is vague or incomplete, but do not immediately give me the correct answer."

That last instruction can be particularly important. One of the dangers of using AI for learning is that it is almost too helpful. The moment you encounter difficulty, you can ask for the answer.

But difficulty is not always a problem to be eliminated.

There is a difference between:

• "I'm completely stuck. Help me understand this."

and:

• "This is becoming mentally uncomfortable. Please give me the answer so I don't have to struggle with it."

Those are not the same thing.

Sometimes the better use of AI is to tell it not to answer immediately.

For example:

"Do not give me the answer yet. Give me one hint and let me try again."

Or:

"Ask me questions that help me work this out myself. Only explain the answer after I have made several serious attempts."

Used this way, the AI helps preserve some of the productive struggle that can be part of genuine learning rather than constantly rescuing you from it.

Using AI at Different Levels of Learning

AI-generated quizzes do not have to stop at factual recall. A useful tutor can push the learner through progressively more demanding forms of thinking.

Level	What You Are Testing	Example AI Prompt
Recall	Can I retrieve the information?	"Define this concept without looking."
Understanding	Can I explain what it means?	"Ask me to explain this in my own words."
Application	Can I use it in a new situation?	"Give me a new scenario where I have to apply this framework."
Analysis	Can I break the idea apart and compare it with alternatives?	"Give me two competing explanations and ask me to analyze the differences."
Evaluation	Can I judge an argument or defend a position?	"Challenge my interpretation and make me defend it."
Creation	Can I use the ideas to build something new?	"Give me a problem that requires combining these ideas into a solution."

This is where AI becomes more interesting than simply asking it to generate flashcards. It can help move the learner from "Can I remember this?" toward questions such as "Do I really understand it?", "Can I apply it?", and "Can I defend or criticize it?"

Those are much harder tests of whether something has actually been learned.

Use AI to Find Blind Spots, Not Just Gaps

The most valuable quiz questions are not always the ones I cannot answer at all.

Sometimes I can answer a question confidently and still reveal that I understood the concept differently from what the author intended. Other times I know the definition but cannot explain why it matters. Or I understand the theory in the abstract but fail when asked to apply it to a realistic situation.

Those are different kinds of blind spots.

A good AI tutor can help expose them by asking questions such as:

• "What assumption is built into your answer?"
• "Can you give me an example that would challenge your interpretation?"
• "How would this concept apply if one of the important conditions changed?"
• "What is the strongest alternative explanation?"
• "You gave me the definition. Now explain why it matters."

That kind of questioning is often more valuable than simply discovering that I forgot a particular fact.

A Few Practical Notes

• Be specific about the kind of tutoring you want. "Help me study" leaves too much unspecified. Tell the AI whether you want explanation, questioning, retrieval practice, application problems, argument analysis, or some combination of them.
• Have it grade honestly. Explicitly tell the AI not to be generous. A tutor that always tells you "close enough!" isn't doing its job.
• Make an attempt before asking for the answer. If possible, retrieve, explain, or reason through the problem first. Use the AI to respond to your thinking rather than replacing it before it begins.
• Ask follow-up questions. If you get something partially right, don't settle for a simple "correct" or "incorrect." Ask the AI to identify exactly what was strong, what was vague, and what was missing.
• Use it to find blind spots, not just missing facts. The goal is not merely to identify information you forgot. It is also to discover misunderstandings, weak reasoning, and situations where you know a concept but cannot actually use it.
• Don't outsource the reading or the thinking. The AI is a study partner, not a substitute for doing the primary reading. Its value goes up in direct proportion to how much real material and genuine thought you have already put into the process.

The Basic Principle

Used this way, an AI tutor doesn't just save time. It changes the shape of the learning itself.

The important shift is from:

AI gives me information.

to:

AI gives me more opportunities to retrieve, explain, apply, analyze, defend, and receive feedback on what I am learning.

That distinction matters.

The best use of AI for learning is not to reduce the amount of thinking you do. It is to create more opportunities for you to do the kind of thinking that actually helps learning stick.