If your reading list this month covers behavioral economics, classical guitar technique, watershed restoration, and the history of the printing press — you already know what the problem is. The interesting part isn't reading any of those things. The interesting part is the moment when something in the printing-press essay reminds you of something you read about watershed governance two months ago, and you want to follow that thread.
Most note-taking apps don't help with that moment. They keep your watershed notes in one folder and your printing-press notes in another, and the connection lives only in your head — until it doesn't, because you forgot it. Polymaths need a vault that holds everything in one searchable surface, and an agent that can actually pull across topics.
One vault, no folders fighting over your topics
In Docapybara, you don't have to pick whether the watershed-governance notes go in Environment or Politics or Reading. They go on a page. Pages nest indefinitely, OneNote-style, so you can put them under Reading > 2026 > Watershed governance if you want, but the agent treats the whole vault as one searchable pile regardless of where you filed them.
That's the move. The structure becomes how you find a page; the agent finds across pages on its own.
If you've already got a setup for the more focused version of "tracking what you read," the How to Track Your Reading Notes and Build a Personal Library shape extends naturally — polymath reading is just reading-list discipline plus more domains.
Capture in plain English, file when convenient
The capture habit matters more than the filing system. Three working ideas a day, dropped in plain English, beats a perfect taxonomy you'll abandon in two weeks.
Voice is the right tool for most of it. Walking, driving, mid-shower-thought, post-conversation: tap record on the audio recording in-app and talk for thirty seconds. You get a transcript with speaker labels (useful when the idea came up in a conversation with someone). The transcript drops into a daily Inbox page or wherever's easiest. The agent makes the connections later — you don't have to file at the moment of capture.
For text capture, drop the half-thought wherever you happen to be. Plain English is the interface. "The thing the printing-press essay said about pamphlet networks reminds me of how the small mutual-aid networks during COVID re-formed local information flow. There's something about decentralized information infrastructure that's a recurring shape." Two sentences. The agent can find this six months later when the same shape comes up in a different reading.
The 27-tool agent does the cross-domain searching
Most cross-domain insight is hidden by the friction of finding what you already know. The agent (Capy) has 27 tools, including searches across pages, child pages, and uploaded files. Ask in plain English: "Have I written anything that connects pamphlet-era information networks to modern decentralized communication patterns?" It pulls the relevant pages from across the vault — the printing-press notes, the COVID mutual-aid notes, the things in between — and gives you a synthesis with citations to where each piece came from.
That's the cross-domain move that's hard to do by manually browsing folders. Easy when the agent reads across the whole vault for you. The same shape — agent-acts-on-documents instead of just chatting about them — is the broader Docapybara differentiator; we wrote about it at Claude Code for Documents if you want the full pitch.
Inline databases for the things that actually have structure
Some polymath material is genuinely list-shaped. Books read, ideas in progress, people whose work you're tracking, courses you're considering. An inline database via the :::database::: directive lets a small spreadsheet live alongside the prose on the same page.
A Reading database with title, author, domain, status (queued, reading, finished, abandoned), rating, and one-line takeaway is a polymath staple. The agent can pull subsets: "What did I read in the architecture domain last year, and what were the takeaways?" It also handles the cross-domain version: "Which of my finished books had takeaways that touched on systems-thinking, regardless of domain?"
For tracking people across domains — the economist whose podcast you follow, the gardener whose substack you read, the historian whose books you keep buying — a People database with name, domain, what they think about, where you find their work, and your notes makes it findable. (For collectors with a similar inventory shape, the Notes for Collectors: Inventory, Provenance, and Market Research in One Place approach uses the same database mechanic for different content.)
The synthesis page — where the connections turn into something
Cross-domain reading creates ideas that want to become essays, projects, talks, or just clearer thinking. The Synthesis page (or one per topic) is where ideas escape the bookmark folder and start to mean something.
Each synthesis page is a working document. A working title, the question it's trying to answer, the threads that feed into it, the open questions, the things you'd need to read or talk to someone about to make it real. The agent can pull related material from the vault when you're ready to write: "For the synthesis on decentralized information infrastructure, pull the most relevant five pages from the vault — book notes, conversation notes, anything that touches the topic."
Then the agent can help draft. "Using only the notes you just pulled, draft a 600-word post on the recurring shape — keep it specific to my actual reading, don't generalize beyond what's in the notes." The draft is grounded in your own material, with a clear audit trail back to where each idea came from. (The same agent-grounded-in-your-notes shape works for How First-Time Founders Use AI Notes to Move Faster, where the synthesis is an investor pitch instead of an essay — but the mechanics are identical.)
Conversations — the polymath gold mine
Polymath thinking is mostly forged in conversations across fields. The economist friend, the gardener neighbor, the museum-curator cousin, the engineer in your bookclub. Most of those conversations leave no record, which means the ideas they sparked die.
A Conversations page tree — one child page per person you talk with regularly — fixes the leak. After a conversation, drop a quick note. Voice works well; thirty seconds of "Talked to Maya about cooperative ownership models — she pointed me at two case studies, said the legal-structure question is the bottleneck, sent me a paper I should read" is enough.
The agent can pull across people: "Across all my conversations this year, what topics keep coming up, and who else is interested in them?" That's the kind of cross-cutting view that's hard to get by browsing notes one person at a time.
Daily tickle file — what you wanted to keep thinking about
A quick note on the medium: polymath reading lives partly in academic PDFs, conference papers, and long-form essays you saved. Dropping them on a page works because Docapybara converts uploaded PDFs to markdown via docstrange — the agent reads them as text, not as opaque blobs. When you ask "in the watershed governance paper I uploaded last month, what was the section on small-grant funding?", the agent can actually pull that section back to you, quote the relevant paragraph, and link to where it sits in the document. Same applies to chapter scans, conference proceedings, and dense essays you'll only read sections of.
A daily-ish Tickle file page (or a journal pattern of one entry per day) holds the open threads. Three or four lines a day: what you read, what you're chewing on, what you want to come back to.
The agent can summarize across time: "What were the recurring questions I kept noting in my tickle file across March and April?" That's the kind of pattern-spotting that's easy to miss in real time and obvious in summary.
The same daily-capture shape is in The Capture Habit: Remembering the Things That Actually Matter — the polymath version is just denser and more cross-cutting.
A starter shape that doesn't pretend to be tidy
Polymaths shouldn't try to impose a clean taxonomy up front. The taxonomy that emerges over a year of writing is the one that fits how you actually think. So start small:
- Inbox — daily voice and text capture
- Reading — page tree per book or paper, plus a database of titles
- Conversations — one child page per person, dated entries
- Synthesis — one page per ongoing essay or topic
- People — database of who you're tracking
- Tickle file — daily three-line entries
That's it. The agent does the searching across all of it. Folders can grow later if you want them; you don't need them to make the agent useful.
The polymath problem isn't input — it's connection. The vault solves the input side, and the agent solves the connection side. What's left is the actual thinking, which is the part you wanted to do anyway.
Try Docapybara free — start with the Inbox page and the first Synthesis page, and the next time you read something that reminds you of something else, you'll be able to find both.