Getting started with a second brain takes one decision and one habit: keep your notes as plain-text files in a single folder, and capture anything worth keeping the moment you meet it. Everything else, the structure, the tags, the AI on top, gets layered on later, and it all works better when the foundation is that simple. This guide walks the whole path: what a second brain actually is in 2026, the tools you need (most are free), the seven steps from empty folder to a system that compounds, and the honest first month, including the moment most people quit.
What is a second brain?
A second brain is an external system that holds your notes, ideas, sources, and decisions so your biological brain does not have to. The term was popularized by Tiago Forte's Building a Second Brain method. In 2026 the working definition has shifted: a second brain is a folder of plain-text notes that an AI assistant can read, search, and write back to, so it answers from your context instead of starting cold every session.
Why is everyone talking about second brains again?
The idea is a decade old, but this summer it caught a second wind, and the reason is not a new app. It is that AI finally gave the second brain a job.
For years the honest critique of note systems was that they were write-only. You captured diligently, organized elaborately, and retrieved almost never. The notes sat there. What changed is that AI assistants became good enough to be the retrieval layer: point one at your notes folder and it can read every note before it answers, connect the new to the old, and write new notes back when it learns something. The viral framing this summer, echoed everywhere from Andrej Karpathy's living-wiki idea to a thousand build threads, is blunt: stop using AI only as a chatbot, and let it think with you against everything you have ever captured.
Underneath the hype is a real shift worth internalizing. The people getting the most out of AI right now are not the ones with better prompts. They are the ones with better memory systems. Every insight you fail to capture is a question you will re-answer from zero, and every note you do capture is context your AI can use forever. That asymmetry is the entire case for starting a second brain this year.
What a second brain actually is (and is not)
Strip away the productivity theater and a second brain is three habits wrapped around one folder:
- Capture: get ideas, quotes, decisions, and sources out of your head and into the folder with as little friction as possible.
- Light structure: just enough organization that things can be found: an inbox, your notes, your sources, an archive. Not a 40-folder taxonomy.
- Retrieval: actually using what you captured, which in 2026 mostly means letting an AI assistant read the folder and answer from it.
It is not an app you buy. Obsidian, Notion, Apple Notes, and a dozen AI-native tools all market themselves as second brains, and any of them can host one, but the system is the habits plus the files, not the software. It is also not an archive. A folder of ten thousand notes nobody reads is a storage unit, not a second brain. The test is retrieval: can you, or your AI, pull the right thing back out when it matters?
One design choice does most of the work, and it is the one the current wave of builders has converged on: plain text wins. Markdown files in a normal folder are readable by every editor, every AI tool, and every future version of you. No export lock-in, no proprietary database, nothing to babysit. Plain text is exactly what an AI agent can read, search, and write, which is why the folder-of-markdown pattern beat the fancier alternatives.
What tools do you need to build a second brain?
Three layers, and free defaults exist for all of them:
| Layer | What it does | Free default |
|---|---|---|
| A folder | Holds your notes as plain-text Markdown files | Any folder on your machine |
| An editor | Makes writing and linking notes pleasant | Obsidian, VS Code, or any Markdown editor |
| An AI assistant | Reads the folder, answers from it, helps maintain it | Claude, ChatGPT, or any tool that accepts your files |
That is the whole stack. You do not need a database, an API key, a plugin ecosystem, or a subscription to get started. Obsidian is the most popular editor choice because it treats your folder as the source of truth and adds linking and search on top, but the folder matters more than the editor. If you already live in VS Code, use that. If you capture on your phone, any app that syncs plain files works.
How to build a second brain, step by step
Step 1: Create one folder and commit to plain text
Make a folder called second-brain (or anything you like) somewhere synced and backed up. Decide now that everything in it will be plain-text Markdown. This single decision keeps you portable across editors, readable by every AI tool, and immune to the app-migration cycle that kills most systems.
Step 2: Add four subfolders, and stop there
An inbox for raw captures, notes for ideas in your own words, sources for where things came from, and an archive for what is done. Resist the urge to design a grand taxonomy on day one. Structure should be earned by volume, not invented in advance. Four folders will carry you for months.
Step 3: Build the capture habit
This is the habit that decides whether the system lives. When something resonates, an idea in the shower, a line from a call, a decision you just made, a paragraph worth keeping, it goes into the inbox immediately, unpolished. The filter is resonance, not importance: if it made you pause, capture it. Do not organize at capture time. Capture is a reflex; organizing is an appointment.
Step 4: Write notes in your own words, one idea per note
When you process the inbox, turn raw captures into atomic notes: one idea, a title that states it, the idea in your own words, and a line recording where it came from. The source line feels like bureaucracy until the first time you, or your AI, needs to know whether a claim can be trusted. Then it is the most valuable line in the file.
Step 5: Point your AI at the folder
This is the step that separates a 2026 second brain from a 2019 one. Open your AI tool of choice, give it access to the folder (or attach the relevant files), and start asking questions you know the answers live in your notes: "What did I decide about X?", "Summarize everything I have captured on Y", "What connects these two ideas?" The first time it answers with your own thinking, correctly attributed to your own notes, the system stops being a chore and starts being a colleague.
Step 6: Run a 20-minute weekly pass
Once a week: empty the inbox, archive what is finished, surface what is overdue, and promote the three things that matter next. Twenty minutes, calendar-blocked. This is the entire maintenance burden of a healthy second brain, and skipping it is how systems rot into storage units.
Step 7: Test retrieval monthly
Once a month, ask your AI ten questions you know are answered somewhere in your notes, and score the results. Retrieval failures are diagnostic gold: they tell you which notes are badly titled, which ideas never got captured, and where your structure is hiding things. A second brain you never test is a second brain you cannot trust.