1. A New Way of Learning
For a long time, learning something new meant finding a good book or course and following the path designed by someone else—from the fundamentals to the advanced topics.
That made sense when access to knowledge was expensive.
But AI changes this equation.
For people who already have substantial experience, the problem is often not that they know nothing about a new subject. They already have years of knowledge, mental models, and experience that can serve as a starting point.
The real challenge is figuring out what they actually need to learn, how the new concepts connect to what they already know, and where their understanding is incomplete or wrong.
This makes a different approach possible:
Instead of starting from page one, start with a question.
Tell the AI what you already know and what you are trying to understand. Let it identify the relevant concepts, explain the missing pieces, ask questions, challenge your assumptions, and adjust the depth based on your responses.
Matt Pocock's teach skill is an interesting example of this direction. Rather than simply asking an AI to explain something, it turns the interaction into an ongoing learning process, using goals, learning history, questions, exercises, and feedback to help determine what to learn next.
The underlying idea is quite different from simply asking AI to summarize a textbook:
Don't just consume a predefined curriculum. Let AI help organize the learning process around your questions, existing knowledge, and actual gaps.
Traditional learning often looks like:
Course → Knowledge → Exercises → Exam
AI-assisted learning can look more like:
Question → Dialogue → Understanding → Practice → Feedback → Correction → Deeper Exploration
The knowledge itself hasn't changed. The path through the knowledge has.
And this may be particularly valuable for experienced learners. Instead of spending weeks going through things they already understand, they can start from what they know and focus their time on the missing pieces and the connections they haven't made yet.
This doesn't mean traditional books, courses, or systematic study are obsolete. They remain important, especially when building foundational knowledge.
But they are no longer the only possible entry point.
We don't always have to learn everything first and solve problems later. We can start with a problem and learn our way through it.
2. Making AI a Learning Interface
I've been experimenting with this idea myself and created a learning resource for studying LLMs with AI:
Understanding LLMs for Software Engineers · Understanding LLMs for Software Engineers
A systems-first guide for software engineers to understand LLM principles, mechanisms, and boundaries.
kingson4wu.github.io
The idea is slightly different from a conventional learning website.
It is designed not only for humans to read, but also for AI tools to navigate and use as a knowledge base.
The material includes a structured knowledge map, relationships between topics, learning indexes, and different entry points based on questions. It can be used with different AI tools such as ChatGPT, Claude, or Codex.
Instead of:
Start from Chapter 1 → Read everything → Take notes → Move on
you can try:
Ask a real question → Let AI locate the relevant knowledge → Discuss it → Let AI challenge your understanding → Go deeper into the source material when necessary.
I think this is only an early example of what learning might look like in the AI era.
Learning materials don't have to be something we simply read anymore. They can become a knowledge base that AI and humans use together to explore, understand, and verify ideas.
Top comments (0)