Field Guide · Tools for Thought
A note-taking app that quietly turns into a personal knowledge network — and why people pair it with an AI like Claude to build a “second brain.”
Obsidian is a note-taking app built on plain text files that live on your own computer, where notes can link to each other — and those links turn a pile of separate notes into a connected web of knowledge.
That’s really the whole idea. The two words that make Obsidian different from a normal notes app are local (the files are yours, sitting in a folder) and linked (notes point at each other, like pages on Wikipedia).
Apple Notes is a drawer: you drop notes in and dig them out later. Obsidian is a city map: every note is a place, and the roads between them are the point.
Most apps make you choose one folder for each note. But real ideas belong in many places at once. A note on “battery thermal runaway” is part diagnostics, part chemistry, part safety. Folders force you to pick one shelf. Links let the note sit on all of them.
Plain .md text files. No special format, no lock-in — readable in any editor, forever.
Type [[Note name]] to connect notes. Obsidian also shows backlinks: who links to this note.
Just a normal folder on your disk holding all the notes. That folder is your knowledge base.
An auto-drawn map of every note and link. Clusters reveal what you actually think about.
You don’t have to plan a structure up front. Just write notes and link them when it feels natural. The structure emerges — this is why people call it a “bottom-up” or organic system instead of a rigid filing cabinet.
This is the part that’s hard to feel from a description, so here it is live. Each circle is a note; lines are links. Tap any note to see what it connects to — the same move you make in Obsidian when you follow a trail of thought.
Tap a note to light up its connections · tap empty space to reset
Notice how tapping one note reveals neighbors you might have forgotten were related. That’s the whole value of links over folders, made visible.
Obsidian has both, and beginners mix them up. The screenshot that started this even shows a “Tags” icon. The simple rule:
[[Project Atlas]] says “this note is about or connected to that specific thing.” Use for real, meaningful ties.
#idea or #todo says “this belongs to a broad category.” Use for status and filtering across many notes.
Rule of thumb: if you could write a sentence about why two notes relate, use a link. If you just want to group lots of notes by type or status, use a tag. Links build the map; tags help you sweep across it.
At the very top of a note you can add a small block of properties (also called frontmatter, written in YAML). It’s a handful of key–value fields that describe the note — its status, date, type, related people. Humans skim it; software can read it.
--- (this block sits at the top of the note)
title: Bearing failure detection strategy
type: decision
status: in-review
date: 2026-06-07
tags: [diagnostics, prognostics]
related: ["[[Vibration features]]", "[[RUL estimation]]"]
---Properties turn a free-text note into something an agent can filter and sort: “find every note where status: in-review and type: decision.” The same three-layer idea you’d use in a well-structured skill — discoverable metadata up top, content below.
The most reliable way to actually use Obsidian is the daily note: one note per day, named by date, where you dump whatever happens — tasks, calls, half-ideas. Because you link out from it as you write, each day becomes a hub that quietly wires everything together.
[[link]] you type threads it into the wider web — no extra filing required.Don’t treat daily notes as a diary to keep forever. Treat them as a landing strip: capture fast, then move anything worth keeping into its own linked note. The daily note can stay messy.
Your real brain is great at thinking but terrible at remembering details reliably. A second brain is an outside store that holds the facts, links, and context so your real brain is freed up to reason, not memorize.
The trick is that a second brain only works if retrieving something is as easy as writing it down. Folders make retrieval hard (which drawer did I put that in?). Links make it easy (follow the trail). That’s the entire promise.
Search only finds what you remember to ask for. Links surface things you forgot you connected — the note about a meeting reminds you of the decision, which reminds you of the open risk. Serendipity, on purpose.
On its own, Obsidian is a great memory but it can’t think — it just stores. An AI like Claude is great at reasoning but has no memory of your world. Connect the two and each fixes the other’s weakness:
The dotted return arrow is the part most people miss. If you only ask the AI questions, you get a chatbot. If you save its useful answers back as linked notes, your memory grows smarter every time. That feedback loop is what the original post meant by “that’s when the system finally works.”
The AI can’t see your whole vault at once, so how does it use it? The common pattern is retrieval — fetch only the few notes relevant to your question, then hand those to the model as context. In plain terms: find the right pages, then read them, then answer.
You don’t need fancy tooling to start. Think of it as a ladder — climb only as far as you need.
Begin with copy-paste for a week or two. You’ll quickly learn which questions are worth asking and which notes are worth feeding in — then graduate to a plugin or an agent that reads the vault folder for you.
This is the catch nobody mentions in the hype posts. Obsidian itself is local-first: the notes sit on your disk and never go anywhere unless you send them. But the moment you pipe a note into a cloud AI, that text travels to the AI provider to be processed. Your second brain stops being purely private at exactly that step.
Keep secrets, passwords, regulated or confidential material out of any note you’ll send to a cloud model — or redact it first. If the content is sensitive, prefer a local model running on your own machine, or paste only the specific excerpt the AI needs rather than the whole note.
Abstract diagrams are easy to nod at. Here’s the actual loop with real text. Say you have three rough notes from the week — a meeting, a test result, and an open question. You select them and prompt:
# Prompt to Claude
Here are three notes from this week (pasted below).
Summarize them into a single decision note: what we
decided, what's still open, and the next action. Add
property fields and link back to the source topics.Claude reasons over them and hands back a clean, structured note. You paste it into the vault as a new file — that’s the dotted arrow from Figure 3 made real:
Decided. Default sampling stays low; the [[Edge triage]] node bumps to 2 kHz only when an anomaly score crosses threshold, to control upload cost.
Still open. Threshold value isn’t set — depends on the [[Data acquisition]] campaign results due next week.
Next action. Sara to run the campaign and report the false-trigger rate. See [[Sara]].
Now that note is linked into the web. Next month, opening [[Edge triage]] shows this decision as a backlink — the reasoning is permanently wired into your memory instead of lost in a chat window.
[[ ]].Over-organizing: people spend weeks on folder structures and plugins instead of writing. Links matter more than folders. Hoarding: saving without linking just rebuilds the messy drawer. A note with zero links is barely in your second brain at all.
After writing any note, ask: “What two existing notes should this connect to?” Add those links. Do only that, consistently, and the graph builds itself.
Obsidian isn’t always the right pick. A quick orientation against the tools it’s usually compared to:
| Obsidian | Notion | Apple Notes | Roam | |
|---|---|---|---|---|
| Where notes live | Local plain-text files (yours) | Notion’s cloud | Apple’s cloud | Roam’s cloud |
| Linking | First-class — [[ ]] + backlinks | Possible, more manual | None to speak of | First-class, core idea |
| Works offline | Fully | Partial | Fully | Limited |
| Databases / tables | Basic (via plugins) | Excellent | No | Basic |
| Team collaboration | Weak (built for solo) | Strong | Light sharing | Moderate |
| Best for | Linked personal knowledge, long-term thinking | Team docs & structured databases | Fast everyday capture | Networked daily thought |
| Cost | Free for personal use | Freemium | Free with Apple device | Paid |
An honest field guide says where the tool fits poorly:
Obsidian shines when you’re building knowledge over time and want to own it. For one-off capture or team docs, simpler or cloud-native tools win. Pick the tool you’ll actually keep open.
Obsidian remembers. Claude reasons. The link between them — and the habit of saving results back — is what turns a notes app into a second brain.
**bold**, # headings). Readable anywhere, forever.[[Note name]]. The thing that makes Obsidian a network.#idea. Good for grouping; weaker than a link for real relationships.Links point to primary/official sources. Concepts like the Zettelkasten and “second brain” are widely documented; these are starting points rather than an exhaustive bibliography.