A HN post about using LLMs to learn complex topics just hit 240 points and 127 comments on Hacker News — and the discussion reveals a fundamental tension in how people are actually using AI for learning.
The original post describes a workflow: use an LLM to break down complex topics, generate summaries, create mind maps, and build a personalized curriculum. It's a compelling idea. But the comments tell a more nuanced story.
The Problem with LLM-Generated Learning Material
One commenter put it perfectly:
"I get exhausted reading LLM prose. I really don't want to read anything generated by something like Opus 5 at this point."
This is a real problem. LLM-generated explanations tend to be:
- Dense but superficial — they cover everything but explain nothing deeply
- Uniformly structured — every section has the same rhythm, making it hard to identify what matters most
- Missing the "why" — they tell you what but often skip why it matters or how it connects to other ideas
As one HN commenter noted: "As I dive deeper, I need a way to organize the information in a useful way as I begin to branch out in many different directions."
What Actually Works: Feed the Book, Ask Questions
The highest-upvoted comment in the thread was remarkably simple:
"It's much better to feed the book to the LLM and ask questions as you read along, instead of asking the LLM to basically write a custom book for you from scratch."
This aligns with what learning science tells us: active recall and elaborative interrogation (asking "why does this work?") are far more effective than passive reading. When you feed a trusted source to an LLM and interrogate it, you're doing active learning with an infinitely patient tutor.
The Socratic Method: Let the Model Guide You
Another commenter introduced a powerful technique:
"I came across the Socratic method recently, and have used it to learn a couple of topics that I was having trouble getting to stick. It works for concepts as opposed to facts, and causes the model to guide you to answers through your own reasoning, which is both much more engaging than reading a wall of LLM text and helps the information stick."
This is brilliant. Instead of asking the LLM to explain, you ask it to question you. The LLM becomes a Socratic tutor, probing your understanding and forcing you to articulate concepts in your own words. This approach:
- Forces active recall — you can't passively read; you have to generate answers
- Reveals gaps — when you can't answer a question, you know exactly what to study next
- Builds mental models — by reasoning through questions, you construct understanding rather than memorizing facts
You can implement this with a simple system prompt like:
You are a Socratic tutor for [topic]. Do not give me direct answers. Instead, ask me questions that guide me to discover the answer myself. Start by asking what I already know about [topic].
The Quiz-and-Feedback Loop
Another practical suggestion from the thread:
"Ask it to quiz you if you are feeling it."
This combines the best of both worlds:
- Read a trusted source (book, documentation, paper)
- Ask the LLM to generate quiz questions based on what you just read
- Answer the questions — active recall
- Ask the LLM to evaluate your answers and identify gaps
- Go back to the source to fill those gaps
This is essentially the Feynman Technique with an AI partner.
When to Use LLMs vs Books
The thread converged on a practical division of labor:
Use books for:
- Initial exposure to a topic (structured, vetted, coherent narrative)
- Deep dives where you need a single author's coherent mental model
- Topics where accuracy is critical (security, medicine, law)
Use LLMs for:
- Clarifying specific questions or concepts you didn't understand from the book
- Generating practice problems and quizzes
- Socratic dialogue to test your understanding
- Connecting ideas across different domains ("How does X relate to Y?")
- Translating between different levels of abstraction
The Real Insight
The deepest insight from the thread is this: LLMs are better as learning partners than as learning material generators. When you ask an LLM to generate content, you get dense, uniform prose that's hard to learn from. When you ask an LLM to interact with you — questioning, quizzing, clarifying — you get a personalized learning experience that adapts to your gaps.
The best learning happens when you combine a trusted, structured source (a book, a course) with an LLM as an interactive tutor. The book gives you the foundation. The LLM helps you build on it.
As one commenter succinctly put it: "I am back to books and mostly try to use LLMs to clarify certain questions or ideas I have."
That might sound like a retreat from AI, but it's actually the most effective way to use it.
What's your experience using LLMs for learning? Have you tried the Socratic method or quiz-based approach? Share your workflow in the comments.
Top comments (0)