How to Master Hard Books with AI and Obsidian
I want to describe a way of using AI that runs against the grain of how most people use it. The common approach is to let AI do the reading for you. Feed it the book and it will summarize the chapters, draft the notes, produce the outline, and generate the flashcards. In a few…
I want to describe a way of using AI that runs against the grain of how most people use it.
The common approach is to let AI do the reading for you. Feed it the book and it will summarize the chapters, draft the notes, produce the outline, and generate the flashcards. In a few seconds you have a tidy knowledge base and the pleasant sensation of having learned something. Most of the time, though, what you have actually done is watch the machine process the material on your behalf. The folder is full. Your head is not.
That may be fine for some purposes. If I need the gist of a report before a meeting, a summary is exactly the right tool. But it is not mastery, and it will not survive the few weeks it takes to forget everything you did not truly own.
In the newest Squared Away Life episode I walk through a different approach. I use AI as a tutor and study partner inside Obsidian, so that difficult reading becomes more active, more demanding, and more fruitful rather than less. The goal is not to outsource thought. The goal is to build scaffolding for better thought.
The Problem with Passive Learning
Almost everyone has had this experience. You read a serious book, or work through a course, or sit under a good lecture, and the whole time it feels like the material is going in. You are nodding. It makes sense. Then a month later someone asks you what the book argued and you can produce a vague gesture and a sentence or two.
The problem usually is not the quality of the material. It is the posture of the learner. Passive exposure feels like learning because comprehension in the moment is easy to mistake for retention over time. But recognizing an idea on the page is not the same as being able to summon it, state it, and use it when the page is closed.
We learn more deeply when we do something harder than absorbing. We learn when we retrieve ideas from memory, explain them, question them, argue with them, connect them to what we already know, and restate them in our own words. That work is effortful, and the effort is the point. It is also exactly the kind of work that a summary quietly removes.
This is where AI can help, if it is used carefully. The trouble is that the easy way to use it removes the effort, and the effort is what builds the understanding.
AI as Tutor, Not Substitute
For this experiment I built what I call a mastery kit inside Obsidian: a self-contained study system for working through a single difficult text.
The kit included a map of the whole work, so I could see the shape of the argument before diving in. It included atomic note stubs for the major ideas, each one titled as a claim I would have to defend rather than a heading I could copy. It included active recall prompts, a dated reading and review schedule, a deck of flashcards on a spacing schedule, source lists and background reading for the places I was thin, and space throughout for my own dictated or written responses.
Here is the part that matters most. The AI did not write the learning for me. It built the structure that forced me to do the learning myself.
It asked better questions than I would have thought to ask. It surfaced background issues I had glossed over. It identified the load-bearing ideas in the argument, the ones everything else rested on. It pressed me for clarification and pushed counterarguments at me when I was getting comfortable. But at no point did it hand me the understanding. I still had to read the text, close it, recall what I could, answer the prompts, judge what mattered, verify what I was unsure of, and write the notes in my own words.
That distinction is the whole thing. A tutor who does your homework is not a tutor. A tutor who makes you do the homework, and makes the homework sharper, is worth a great deal.
Build Your Own Mastery Kit
You do not have to build this from scratch. I packaged the same prompt I used as a free download.
It is a self-contained AI skill file that turns any book, paper, or course into a complete active-learning system: a map of the work, guided note stubs, a flashcard deck, and a dated read-recall-write-review schedule. You drop it into Claude or any capable assistant, name the resource you want to study, and it builds the scaffolding. You do the learning. The templates are plain Markdown, so they work in Obsidian, Logseq, or plain files.
What Makes the Approach Work
None of this rests on anything novel. The kit is built on a handful of learning practices that are well established and easy to neglect.
Active recall is the habit of closing the book and retrieving what you remember before you look. The retrieval itself is what strengthens the memory, far more than rereading does.
Spaced repetition means returning to the material over expanding intervals instead of cramming it once. Forgetting a little between sessions is not a failure. It is what makes the next retrieval do its work.
The generation effect is the finding that you remember what you produce far better than what you merely take in. This is why the notes in the kit are stubs you write, never prose the AI hands you.
Elaboration is connecting a new idea to what you already know, so it lands in a web of other ideas rather than floating alone.
Dialogue is arguing with the text instead of receiving it. Where do you agree, where do you diverge, and where would you press if the author were in the room?
AI is unusually good at building scaffolding for every one of these. It can generate prompts, schedules, questions, and review structures in seconds. But the value only shows up when the learner stays responsible for the actual cognitive work. The scaffolding is not the building.
The Guardrail
The danger in all of this is outsourced judgment.
If you let AI write the notes, decide what matters, supply the interpretation, and produce the conclusions, you can end up with a polished and impressive knowledge base and a genuinely shallow understanding of the thing you meant to master. Everything looks done. Nothing is owned. That is the exact opposite of the goal, and it is easy to slide into because the outsourced version is faster and feels almost as good in the moment.
The better use is to let AI become an always-available tutor: a partner that helps you ask better questions, tests whether you actually understand what you think you understand, and points out what you missed. Kept in that role, it makes the reading harder in precisely the way that makes it stick.
Use it to make yourself think harder and better. Not to think less.
It is free, and it comes with a short guide to running your first session.