Root

A personal knowledge companion that keeps every idea anchored to its source

921 commits · last on Sep 27, 2026

TanStackFastAPIReact NativePostgrespgvectorLearningKnowledge ManagementBetter Auth
GitHub

I spend a lot of time with books, podcasts, and long-form interviews — and for years, most of it evaporated. I’d remember that I heard something interesting, but not where it came from, why it mattered, or what I actually thought about it at the time. The same frustrations kept surfacing in conversation:

  • “I remember hearing something like X in a podcast once.”
  • “What book was that stat from?”
  • “What were my actual takeaways from that source?”

Root is the fix: a personal knowledge companion for capturing what I consume, keeping every idea anchored to its source, and building my own thinking on top.

Having spent much of my life learning about investing, I think in terms of compounding — small, consistent deposits that accumulate into something outsized. Learning works the same way, but only if each insight lands somewhere it can build on the last. Root is my attempt to compound learning itself.

How It Works

The easiest way to explain the pieces is to follow one real thread through my library.

Everything starts with a source — a book, podcast, interview, or article. Sources act as anchors, giving every idea a clear origin and preventing insights from becoming detached or misremembered.

Steve Jobs Book

Walter Isaacson

As I read, highlights become citations, and comments capture what I thought in the moment. Citations are concrete facts and quotes exactly as they appear in the source — precise, attributable, and free of interpretation. Comments are the margin notes attached to them: my own in-the-moment thinking, kept deliberately separate from the source material. A single citation can accumulate multiple comments over time as my thinking evolves.

citation
“

“You son of a bitch!” Jobs said. But it was not in anger. There was a big grin on his face. As soon as he realized what Belleville and the other engineers had done behind his back, said Hertzfeld, “Steve swallowed his pride and thanked them for disobeying him and doing the right thing.”

”
— Steve Jobs · p. 147

Each source gets distilled into at most five takeaways. A source can hold dozens of citations, but Root caps takeaways at five — the limit is the point. It forces me to synthesize patterns and arguments instead of hoarding highlights.

Key takeaway 1/5

Charm, Belief, and Getting Shit Done

Jobs learned to use his charisma and his attitude to motivate the people around him. His belief that he could do anything was placed in the other people he worked with, which allowed him to get things done. Not only that, but he was a fantastic storyteller. His ability to paint a picture and narrative to motivate, coerce or convince was incredible.

9 quotes

Notes sit above individual sources. Where a takeaway distills a single source, a note weaves citations and themes from many — books, podcasts, interviews — into a longer, coherent argument. When the same idea starts appearing across multiple sources, that’s when a note begins. Isaacson describes Jobs’s obsession with “whole widgets” — products designed end-to-end, software tailored to hardware — and when the Cerebras CEO made the same argument about wafer-scale chips on a podcast months later, the overlap became this note:

note

Strengths of Vertical Integration

I actually really like the advantages of owning the entire stack (vertical integration). I think it makes your life really easy, as you don’t need to over adapt and defensively program since you can’t make that many assumptions. On the other hand, when you own the entire stack, it’s quite clear what can happen, and what can’t. This cohesiveness also allows you to create better, more integrated products…

weaves 2 sources: Steve Jobs20VC: Andrew Feldman (Cerebras)

Search & Retrieval

As the library grows, I can ask questions across my takeaways, citations, and comments. Answers cite the specific citations and takeaways they draw on, and the agent’s searches are visible as it works, so I can check where an answer came from.

Graph

The graph projects sources, takeaways, and tags into a 2D embedding space and draws edges between semantically related items. Tag clusters surface as labeled regions — Business, AI, Venture, and so on — making it easy to see where my reading is concentrated and where adjacent threads connect. Sliders tune how much the layout weighs shared meaning versus shared tags.

Guiding Principles

Root uses AI to assist, not to think for me. Real learning comes from friction, and AI can easily rob us of it. So the rule is: remove friction from capturing, keep it in thinking.

Easy in, source always attached — Getting an idea into Root should take seconds, but its origin is always preserved. A citation without its source is just a floating quote.

AI proposes, I decide — AI can draft citations and suggest takeaways, but nothing enters my library until I approve it. Drafts from voice notes are cleanups of what I said, and anything that drifts into the speaker’s claims is rejected.

Compounding value — Every highlight, comment, and note makes future retrieval and synthesis richer.

Incremental refinement — Insights aren’t fixed. A takeaway you write today might look different after you’ve read three more books on the same topic. Root is designed to support that evolution rather than freeze your thinking at a single point in time.

How I Use Root

Reading

I upload PDFs directly into Root and highlight text as I read, the same way you’d mark up a physical book. Highlights become citations I can reference later, search across, and pull into notes.

Podcasts & Interviews

Most of my listening happens in the Root mobile app, which has its own player, offline downloads, and a car mode. Podcasts and YouTube interviews get interactive transcripts, so I can skim a long discussion and see my comments and citations pinned to the exact moment they refer to.

The Walk While Talk flow. Capturing ideas while listening on the go used to be high-friction. Now it looks like this:

  1. 1

    Tap voice capture

    The player pauses automatically and starts recording.

  2. 2

    Dictate the thought

    A quick voice memo — no typing while walking.

  3. 3

    Queued in the offline outbox

    Memos wait locally until connectivity returns.

  4. 4

    Uploaded & processed

    Audio is transcribed and drafted as comments or citations anchored to the right timestamp.

  5. 5

    Reviewed at my desk

    I step through the drafts inline on the transcript, editing, approving, or dismissing each one.

Synthesis

As I’m finishing a source, I’ll often notice a recurring theme and start drafting a takeaway. I’ll give it a rough title and attach citations as I go. For example, while reading the Steve Jobs biography, a pattern that stood out early was his ability to persuade through charm and conviction; I created a takeaway around it — “Charm, Belief, and Getting Shit Done” — and linked supporting quotes as they surfaced (nine, by the end). When the same idea starts appearing across multiple sources, that’s when I’ll open a note and start writing something longer.

Roadmap

  • planned

    Physical Book Highlights (OCR) — Quick photo capture of printed highlights, run through the suggestions review queue.

  • planned

    Guided Knowledge Synthesis — Proactive AI prodding on the dashboard — guided questions that help synthesize margin notes into takeaways via interactive chat, keeping the friction of active learning.

  • planned

    Global Search — Fuzzy search across citations, comments, and takeaways with filters.

  • planned

    Kindle & Web Clipper Extensions — Browser integrations for Kindle highlights and general article clips.