You finished a really good book in February. By May, you remember three or four sentences from it, the cover, and the rough shape of one argument. By August, you remember the cover. The thing you actually wanted from reading the book — the ideas it gave you, the ways you thought differently after — has mostly evaporated. You'd consult it if you could find your notes, but you took maybe four pages of notes on a flight, and they're somewhere on your laptop or in a notebook you haven't opened in months.
Most readers run into this version of the same problem. The reading itself is fine. The retention isn't great. And without retention, the books mostly become trophies on a shelf rather than a working library you can pull from.
A vault that holds your book notes in a searchable shape — with the agent doing the cross-book pulling — turns the shelf into something you can actually use. Below is the shape we'd suggest.
One library, one parent page, child pages per book
In Docapybara, Library gets a top-level page, and each book becomes a child page underneath. Pages nest indefinitely, OneNote-style, so a long-running re-read of a series can have child pages per book, and a book you've read three times can have one page per reading.
The agent treats the whole tree as one searchable surface. "What did I write about decision-making frameworks across all my management books?" gets answered from across the vault — not from one folder at a time. (For the polymath version of this — reading widely across unrelated domains — Notes for Polymaths: Connecting Ideas Across Unrelated Domains takes the same vault shape further.)
The reading database — what you've finished, what you're reading, what's queued
The high-leverage move is a Reading list inline database via the :::database::: directive on the Library parent page. Columns: title, author, status (queued, reading, finished, abandoned), category, started, finished, rating, one-line takeaway, link to the book's child page.
The agent can update it. "I'm starting [book title] today — add it to the database, set status to reading, category to architecture." Row appears. "I just finished it — set status to finished, rate it four out of five, the one-line takeaway is that the building-as-system framing changed how I look at neighborhoods."
Once the database has even thirty or forty books in it, queries get useful. "What did I read in the design category last year, and what did I rate four or higher?" "Show me the books I marked as abandoned and why." "Which authors do I keep coming back to?"
The per-book page — the actual notes
Each book's child page is where the working notes live. Open the page, jot as you read. We'd suggest a loose structure but not enforce one — just a heading for Quotes I marked, a heading for Ideas in my own words, and a heading for Open questions. The agent searches across whatever you write regardless of structure.
For physical books, voice is the right tool. After a reading session, walk around the block and talk for two minutes about what stuck. Audio recording in-app gives you a transcript with speaker labels. The transcript lands on the book's page; you can clean it up later if you want, or leave it as raw thought-stream.
For ebooks, the highlights and notes from your reader can paste in. Kindle exports, Apple Books exports, Readwise exports — drop the text. The agent reads it as part of the page.
For PDFs and academic reading, drop the file directly. Docapybara converts uploaded PDFs to markdown via docstrange, so the agent can actually read the contents — "in the chapter I marked from the design book, what was the example about the airport that didn't work?" gets answered with the actual passage.
The cross-book questions — where the library starts to earn its keep
The reason to keep notes at all is the cross-book question. Without notes, every conversation about "have I read anything that bears on this?" starts from memory and ends in vague gestures. With notes, the agent can answer.
"Across the books I've read on cities, what do they collectively say about governance at the neighborhood scale?" The agent reads across the relevant book pages, pulls the relevant excerpts, and gives you a synthesis with citations. You can dig into any specific book the synthesis pointed to.
For writing — essays, talks, longer pieces — the same query gives you the bibliographic foundation in two minutes instead of a weekend of relooking through books. "For an essay on whether AI tooling is closer to industrial automation than to past computing waves, pull the most relevant five book excerpts from my library." Comes back with the excerpts and the books they came from. (The broader writing-grounded-in-your-own-material shape is in How to Draft Emails, Proposals, and Newsletters Inside Your Notes App — same agent, different writing surface.)
Reading that connects to your work — keeping the bridge alive
For working professionals, the most valuable book notes are the ones that bridge reading to current projects. A Reading-to-work page (or a tag in your database) tracks the connections explicitly.
When a book gives you an idea you want to apply at work, drop a quick note on the bridge. "The chapter on small-team coordination in [book] applies to how we're organizing the new pod — try the daily 15-minute design rather than the weekly review." Six months later, you'll have built a small library of book-to-work bridges that you can pull from. "What ideas from my recent reading have I applied or wanted to apply at work?" (The professional-reference-library extension of this is at Building a Reference Library for Your Profession — books are just one input source.)
Recommendations that don't disappear
Every reader gets recommendations and loses most of them. Books recommended (a sub-page or a tag in the database) catches them. When a friend mentions a book, drop it: title, who recommended it, why, your interest level. The agent can pull when you're book-shopping. "What's been recommended to me in the last year that I haven't read yet, sorted by who recommended it?"
For book clubs specifically, you can organize the recommendations by the friend or the club. The book-club discussion shape is in Book Club Notes: Discussion Threads, Highlights, and Per-Member Context if you want a fuller setup for the social-reading version.
The capture habit, not the cataloging project
A note on re-reads: some books deserve a second or third reading. The vault makes it visible — when you re-read, drop a new sub-page on the book's page (or a section), dated with the year of the re-read. Don't try to merge with old notes — just write fresh. The agent can show the evolution: "For [book], what did I write the first time I read it, and what did I write the second time? What changed about my reading?" That's a question that's hard to answer without the notes and easy with them. For books you'd like to re-read but haven't gotten around to, mark them in the database with a re-read queue status. The agent can pull the queue when you're picking what's next.
The risk with book-tracking is treating it as a cataloging project. It's not. The point is the working notes — the ideas in your own words, the open questions, the things you'll come back to — not a perfect record of what you've consumed.
So keep the bar low. A good book might end up with three pages of notes; a fine one with three sentences. An abandoned one might have one paragraph on why you bailed. The agent works with whatever's there. (The broader low-friction-capture-as-habit shape is at The Capture Habit: Remembering the Things That Actually Matter — book notes are just one capture surface.)
A starter shape
If you're moving from "I take notes on a flight and never see them again" to a real library, this is what we'd suggest starting with:
- Library — top-level parent
- Reading list — inline database on the parent page
- Per-book pages — one child per book, loosely structured
- Reading-to-work — a sub-page or tag for the work bridges
- Books recommended — a sub-page or tag for the recs
That's it. Add an Audiobook notes sub-page if you listen, a Quotes I keep coming back to page if that's how your brain works, a Translations page if you read across languages. The vault adapts to you.
The point of a personal library isn't the stack of books. It's the working memory you've built from reading them. The agent makes that memory accessible. You only have to do the dropping.
Try Docapybara free — start with the Reading list database and the page for whatever you're reading right now, and the notes will start compounding from there.