You finished a book in March. By August, all that's left is the cover, a vague positive feeling, and one quote you keep half-remembering. Someone asks you what it was about and you produce three sentences that sound like they could describe nine different books.
Most people read more than they remember. The fix isn't reading slower or more carefully — it's writing a small amount down at the end, in a place where you'll find it again. Not a book report. Just enough that the next time you reach for what you learned, it's there.
A vault that holds one summary per book and one running reading list, with the agent quietly doing the searching, makes that small practice actually stick. (Readers who run a book club on top of this should also see Book Club Notes: Discussion Threads, Highlights, and Per-Member Context.)
One page per book, and the easiest possible template
In a markdown vault like Docapybara, each book gets its own page, named with the title and author. The template can be as light as you want it. A useful starting shape:
- One-paragraph summary — what the book is actually about, in your own words.
- Three things I want to remember — the ideas, scenes, or arguments that mattered most.
- A favourite passage or quote — written out, not paraphrased.
- Who I'd recommend it to — if anyone.
- Rating, loose — a number out of ten or a sentence on how you felt about it.
That's it. Five fields, five minutes. The point is that this is something you'll actually do — anything more involved becomes a chore and you stop doing it after three books.
For deeper books, add child pages: Highlights, Notes per chapter, Open questions, Things I want to look into. Pages nest with no depth limit, so a book that warrants a real engagement can grow without forcing every other book to match.
Highlights and Kindle exports, without the spreadsheet ritual
Most e-readers can export your highlights — Kindle as a CSV or text file, Kobo as a text file, Apple Books as a clipboard dump. Some apps export PDFs.
Drop the file on the book's Highlights child page. PDFs are auto-converted to markdown, so the highlights become searchable text the agent can actually read. Text and CSV land directly.
For the books where you didn't highlight as you went — and for physical books — voice is the right tool after the fact. Narrate the parts that stuck. "The chapter on slow productivity argued that having three projects at a time is sustainable, four is the breaking point. The author tied it to research on cognitive load. Worth coming back to." Transcribed, on the page, searchable.
When you want to find a specific idea later, ask the agent: "What did the slow productivity book say about cognitive load and the number of concurrent projects?" The answer comes back with a citation pointing to the highlight. (For active learners turning books into practiced skills, AI Notes for Learning a New Skill is the next step.)
A reading list that survives — not a wishlist that rots
Reading lists collapse for a familiar reason: they grow faster than you read, and at some point you stop opening them. The list becomes an aspiration shrine instead of a tool.
The fix is to give the list two states: To read soon (a short list of three to five candidates) and Considering (everything else). An inline :::database::: on a Reading list page handles both — columns for title, author, recommended by, why I'm interested, status, and a date added.
The agent maintains it. "Add Slow Productivity by Cal Newport — recommended by Sarah, interested in the deep work follow-up, status considering." Row appears. "Move The Three-Body Problem to To read soon — I just finished the previous book." Done.
When you're picking the next book, ask: "What's on my To read soon list, sorted by how long it's been there, and what was the reason I added each one?" You get a focused decision, not a paralysis list.
For books recommended to you in conversation, the same shape works. Voice it in: "Marcus just recommended Wind/Pinball by Murakami, said it's a good entry point to him. Add to considering." Thirty seconds, on the list.
A library that you can actually search
Once a few dozen books are in the vault, the agent becomes useful in a different way: it can read across all of them.
Some questions that get useful answers:
- "Which books I've read in the last two years had the strongest takes on attention and focus?"
- "Find the book where someone made the argument that institutions outlast individuals — I think it was a non-fiction one I read last summer."
- "What did three different books I've read say about the great resignation? Compare their framings."
- "List every book I gave an 8 or higher to in the last three years."
The agent reads the summaries, the highlights, the notes — whatever you've written — and assembles an answer grounded in your own library. It's not searching the internet for what those books say in general; it's searching what you wrote about them.
For book club hosts, this is genuinely useful when you're picking the next book — you can pull patterns across what the group has already read. For solo readers, it makes your library something more than a shelf.
The recommendations that come at you sideways
A friend recommends a book in a text. A podcast guest mentions one in passing. An article quotes a passage and you make a note to find the book. Most of these never make it onto the reading list because there's friction between hearing and capturing.
Voice removes that friction. "Add to considering — the book James mentioned on the call, something about the history of pencils. Author starts with P maybe? Worth looking up." That's enough — vague captures are fine because the agent can resolve them later.
For the half-remembered ones, the agent's web_search tool can fill in the blanks: "Find the book about the history of the pencil — I think the author starts with P." Comes back with a likely match (Henry Petroski, The Pencil), which you can confirm and add to the list properly.
For book newsletters, podcast notes, and longreads that recommend books in passing, drop the article PDF on a Sources child page. The agent can read across them: "Look at the last three articles I saved about reading lists and pull every book they recommended that I haven't already added to my list." (The same "let your inbox become a knowledge base" pattern is in Turning Your Inbox Into a Searchable Knowledge Base.)
A "year in books" page that builds itself
End-of-year reading reflections are a small ritual a lot of people enjoy and most can't actually pull off because the data is scattered. With one summary per book, the year-in-review writes most of itself.
A Year in books page (one per year) can hold the year's reading, sorted however you want. The agent assembles it: "List every book I finished in 2026, with a one-line description of what it was about and what I rated it." You get the list with the kind of detail that's actually fun to share.
For more reflective summaries, you can ask: "Look at my reading from this year. What were the threads — recurring topics, authors I returned to, books that pushed against each other?" The agent reads across the summaries and finds patterns. Sometimes the patterns are obvious (you read four books on attention this year). Sometimes they surprise you.
A starter shape that works on day one
If you're starting today and you've got two or three books fresh in mind, here's where to start:
- Books — one parent page.
- For each book — one child page using the five-field template above.
- Reading list — one page with the inline database.
- Year in books — one page per year, built up over time.
That's the whole shape. No template required, no metadata to fill in, no taxonomy to maintain.
For the books you've already read but never wrote anything down for, voice is a good way to backfill. "Just want to add a quick note on Project Hail Mary — finished it last August, loved the science, the friendship is the heart of it. Three things I want to remember: the rocky language puzzle, the way the author handled the time gap, the ending I genuinely didn't see coming. Recommend to anyone who liked The Martian." Sixty seconds, page filled.
The point isn't to systematize reading. It's that the books you've actually engaged with stay engaged with — findable, searchable, and quietly forming a personal library that gets more useful the longer you keep it.
Try Docapybara free. Start with the last book you finished and a five-minute summary, and see how much more present that book feels in a month.