EPISODE 04 · Obsidian second brain · 2026.05.17

A second brain
is not a bookmark folder.

This is a talk about Obsidian.
More accurately — it's a talk about not mistaking "hoarding notes" for "having a second brain."

AGENT 101 SERIES · EP 04 · NO-BULLSHIT EDITION
§ 01 · Start with the word

"Knowledge base" and "bookmark folder" are two completely different things

How many "read later" articles have you saved? How many tips have you bookmarked? Now answer one question: when did you last actually reuse something you saved three months ago? This is where most people are stuck — mistaking "saved it" for "learned it."

  • CONCEPT · 01 · SAVING

    Saving = hoarding, not owning

    The "collector's fallacy": you think dropping it in a folder means you've put the knowledge in your head. But the moment you save it, your brain relaxes — "I own it now" — and never looks again. Saving produces the feeling of learning, the same comfort the "1000-prompt bundle" from EP01 sells.

  • CONCEPT · 02 · NOTES

    A note = a byproduct of thinking, not a warehouse

    A good note isn't copying someone's words down — it's chewing them up, rewriting in your own words, wiring them to what you already know. A note's value isn't "what got stored," it's "what you were forced to think through to write it." Copying produces no thought; rewriting does.

  • CONCEPT · 03 · SECOND BRAIN

    A second brain = a thinking system that can be retrieved and recombined

    Its only test isn't "how much is stored," it's "can it, at the exact moment you need it, surface the relevant past and recombine it into new output." A second brain is the system that, while you write or decide, lets you follow a link back to a note from three years ago — not a folder that grows forever and can never be found.

§ 02 · Audit your note system

"I started taking notes" is an empty phrase

It sounds like discipline, like progress. But it fails the three simplest checks. If it fails, what you built isn't a second brain — it's a digital landfill that grows forever and never gets read back.

  • SYSTEM TEST · 01

    When did you last actually reuse an old note?

    Not "I've taken a lot" — "while writing this, I went back to a note from three months ago and it genuinely helped." If you can't name a specific moment, your note base is intake-only — storing equals not storing.

  • SYSTEM TEST · 02

    Can you start from one idea and follow links to three related old notes?

    A second brain's value is in connection, not count. If your notes are isolated islands with no links, they're just harder-to-search documents. Only when you can walk from one to the related next does knowledge start compounding.

  • SYSTEM TEST · 03

    If you deleted the whole base, would your output get worse?

    The brutal one. If deleting it changes nothing, it never entered your production — it was just a warehouse for peace of mind. A note base that hurts to delete is a real second brain.

The first has no finish line — "the perfect base" is a bottomless pit. The second has an action, an output, and a failure signal. A second brain isn't built — it's grown, one note a day.

§ 03 · After the word

Why Obsidian: local, plain text, backlinks

There are many note apps; only a few structures can hold a second brain. This is the only contrast you need: cloud SaaS notes vs local Obsidian. The difference isn't features — it's whether this brain actually belongs to you.

CLOUD SAAS NOTES · what you might use

Convenient, but not yours

Data on someone's server, in their proprietary format, on their terms.

  • Notes live on someone's cloud; exports come out broken
  • Proprietary format lock-in, brutal to migrate
  • Links are mostly decoration, not real structure
  • Shutdown / price hike / redesign — you just take it
LOCAL OBSIDIAN · what you're installing

Plain, but yours forever

A pile of local markdown files; plain text; links are structure; AI can read it directly.

  • Local .md files — always openable, always portable
  • Plain text never goes obsolete, readable in 20 years
  • Backlinks are the skeleton — your web of thought
  • AI / scripts can read and write the whole vault directly
§ 04 · How a second brain turns

Four steps: Capture → Connect → Distill → Express

A living second brain is these four steps turning. Most people only do the first (hoard) and stall at the second (never link), so they never reach the fourth (output). Remember the four and you'll know exactly where your system breaks.

SECOND BRAIN CAPTURE CONNECT DISTILL EXPRESS
  • 01

    CaptureCAPTURE

    Record it with zero friction: jot in a daily note, one-click external material in via an importer. The only requirement is "fast enough that you won't skip it" — if capturing one thing takes five steps, you won't.

  • 02

    ConnectCONNECT

    Link every new note to two old ones (backlinks [[ ]]). This is the step most people skip, and the root difference between a second brain and a folder — links turn islands into a web, and knowledge starts to compound.

  • 03

    DistillDISTILL

    Boil scattered excerpts into "atomic notes" — one note holds one idea, said clearly in your own words. Then a MOC (map of content) strings related atomic notes into a theme. Distilling is the step that turns information into your knowledge.

  • 04

    ExpressEXPRESS

    Recombine notes into output: an article, a PRD, a content angle. The whole point of a second brain cashes out here — if notes never become output, the first three steps were wasted. This step is also where AI adds the most.

§ 05 · The plugins running in my own vault

More isn't better — each of these is wired into one workflow

Obsidian has thousands of plugins; installing a wall of them is the same disease as the "tool collector" (see EP03). Below are the few actually running in my own vault, ordered by the four-step loop: importers handle capture, Git handles safety, Claudian handles express. Each gets one minimal workflow.

  • CAPTURE · MATERIAL INTAKE · Douyin Capture + Xiaohongshu Importer manual hauling → one-click intake

    Get the good content you scroll past into your brain, friction-free

    Contrarian take: the topic bottleneck isn't "no material," it's "saw it, didn't save it; saved it, now it's scattered everywhere." One-click the good Douyin / Xiaohongshu content into your vault — no-watermark video and images, local Whisper transcription to Simplified Chinese, data stays on-device — and capture stops dropping the ball.

    WORKFLOWscroll a gem → intake → feed topic pipeline
    01PASTEpaste a Douyin / Xiaohongshu share link into Obsidian
    02IMPORTno-watermark video / images + title / topic / tags / frontmatter
    03WHISPERlocal Whisper → Simplified Chinese · data on-device
    04FEEDthe note → feed Claude to mine angles / hooks
    SETUPinstall + local Python backend (see README) · ROImaterial stops scattering · capture is frictionless
    Douyin Capture · lyxdream Xiaohongshu Importer · bnchiang96 local Whisper
  • SAFETY · VERSIONING & BACKUP · Obsidian Git fear of deleting → revert anytime

    Your second brain can't live as a single copy

    Contrarian take: many people won't heavily edit their notes for fear of breaking them — so the brain stiffens. Add Git auto-backup: every change has version history, any mistaken delete or botched edit reverts instantly. Only with a safety net will you dare to recombine, cut, and distill.

    WORKFLOWauto-backup + version history
    01AUTOevery N minutes · auto-commit all changes
    02PUSHpush to a private remote · an off-site copy
    03HISTORYfull version history per note · diff any version
    04REVERTmistaken delete / botched edit · revert to any point
    SETUP10 min to a private repo · ROIdare to edit · never lose a note
    Obsidian Git · Vinzent private Git remote
  • EXPRESS · AI COLLABORATOR · Claudian dead note base → a thinking partner

    Let Claude treat your whole vault as its working directory

    Contrarian take: no matter how many notes, unrecombined they're dead. The truly valuable step of a second brain is AI reading your entire vault. Claudian embeds Claude Code / Codex into the vault, making the vault its working directory — file reads and writes, search, bash, multi-step workflows. It synthesizes across dozens of old notes and writes a new note you couldn't have produced alone.

    WORKFLOWsynthesize across notes → new output
    01ASK"synthesize all my notes on Agents into an outline"
    02READClaudian · search + read relevant notes across the vault
    03SYNTHsynthesize across N notes · surface links you missed
    04WRITEwrite a new note / article draft · back into the vault
    SETUPinstall + connect Claude Code · ROIvault goes from warehouse to thinking partner
    Claudian · Yishen Tu Claude Code Codex
§ 06 · Run it tonight

Three steps — wire the second-brain loop tonight

Everything above is theory until the first note gets reused. The three below install in under half an hour and run tonight — one each for capture, safety, express. Don't aim for a perfect base in one shot; first get the loop turning.

DEMO · 01 10 min setup · 2 min daily use

Install Obsidian + open a daily note, write one and link one tonight

A second brain starts with one daily note. Install Obsidian tonight, open the daily note, write one thing you actually learned today, then use [[ ]] to link it to any existing note. Capture + connect, both done.

  • Download Obsidian (free), create a vault
  • Enable the core plugin Daily notes, set a template
  • Tonight, write one note: a specific thing you learned, in your own words
  • Use [[ to link it to an old note (create one if none)
  • Do it 7 days straight — a second brain is grown, not built
obsidian · daily note · 2026-05-17
## Learned today
A second brain's value is in "connection," not "count" —
isolated notes are just harder-to-search documents.
Related: [[collector's fallacy]] · [[atomic notes]]
✓ Captured + linked · loop day 1
DEMO · 02 10 min setup · then automatic

Install Obsidian Git, give your brain a safety net

Before you start heavily editing notes, set up auto-backup. With version history and an off-site copy, you'll dare to recombine, cut, and distill — without fear of breaking anything.

  • In community plugins, search Obsidian Git (by Vinzent), install and enable
  • Create a private Git remote (GitHub / self-hosted)
  • Set auto-commit interval (e.g. every 10 min) + auto-push
  • Deliberately edit a note, open history, confirm you can revert
  • Now edit freely — mistaken deletes are always recoverable
obsidian git · auto-backup
10:00auto-commit · 3 files changed
10:10auto-commit + push · synced to remote
— deleted an important note by mistake —
▌ open history → revert to the 10:00 version
✓ note recovered · 30 seconds total
DEMO · 03 20 min setup · instant payoff

Install Claudian, let AI read your vault and answer a question

This is the step where a second brain goes from "warehouse" to "partner." Install Claudian, let Claude treat your vault as its working directory, and ask it a question that requires synthesizing several notes — you'll see "the past you" surfaced for the first time.

  • Install Claudian (by Yishen Tu) from community plugins, connect Claude Code
  • Confirm its working directory points at your vault
  • Ask a cross-note question, e.g. "synthesize all my notes on X"
  • Watch it search + read + synthesize, then write a new note back into the vault
  • You only judge and cut — that's the "express" step
claudian · vault as working directory
YOUsynthesize all my "second brain" notes into an outline
CLAUDIANsearching the vault…
VAULTread 12 relevant notes
▌ Synthesized · found 3 threads you hadn't linked
· collector's fallacy ↔ missing distillation (same root)
✓ new note "Second Brain Outline" written to vault
§ 07 · The next 90 days

From collector to second brain

This 90-day path doesn't promise a "perfect knowledge base." It gives four specific actions, each with an explicit failure signal. If a week's signal lights up, go back to the previous week. Don't push forward.

  • WEEK 1 – 2 · Build capture

    One daily note a day — make "recording" a reflex first

    Every night, 5 minutes: one thing you actually learned, in your own words. Don't worry about links yet, just record every day. Failure signal: half the days in week 2 are blank. What to do: drop the bar to one sentence — shorter is easier to sustain.

  • WEEK 3 – 4 · Start linking

    Every new note must link to two old ones

    From now on, every new note must use [[ ]] to link two existing notes. This is the watershed between a second brain and a folder. Failure signal: by end of week 4 your notes are still islands. What to do: notes too messy to connect — go back to atomic notes, one idea per note.

  • WEEK 5 – 8 · Automate capture + safety

    Install Git + an importer, make intake frictionless

    Install Obsidian Git for auto-backup; install an importer (e.g. Douyin Capture / Xiaohongshu Importer) to one-click external material in. Failure signal: installed, but material still scatters in bookmarks. What to do: go back to weeks 1–2 — the capture habit isn't set, and tools can't save a missing habit.

  • WEEK 9 – 12 · Wire AI, start expressing

    Install Claudian, let the vault synthesize and write for you

    Install Claudian, let Claude synthesize across notes in your vault and recombine them into real output (article / PRD / topic). Failure signal: by week 12 your notes have never become a single piece of output. What to do: the problem isn't the tool — you never wired "taking notes" and "producing from notes" into one loop.

After 90 days you should be able to answer: "In the past week, which piece I wrote was synthesized by following links back to old notes?" — if you can answer, with output to show, you actually have a second brain. If you can't, you spent 90 days hoarding more notes, not growing a brain.

Real Agent Use Cases

Hoarding a thousand notes is worth less than reusing one

The "read later" you saved will mostly never be read. But a second brain wired into your output hands the past you back to the present you, again and again.
"Execution is always undervalued." If you don't write and link the first note tonight, you won't tomorrow either. That's not a pep talk. It's an observation.

← EP 03 · Tools & Workflows · EP 04 (you are here) · EP 05 (coming)