What is a second brain?

The problem: our brains weren't built to store

The brain is a remarkable thinking machine and a poor filing system. Working memory holds roughly four chunks of information at once [1], and anything not reinforced within hours begins to fade. Every knowledge worker has experienced this: a brilliant insight in the shower, a critical article read last month, a connection between two ideas glimpsed and then lost.

The first serious attempt to solve this in print was Vannevar Bush's 1945 essay As We May Think, published in The Atlantic [2]. Bush imagined a device called the Memex— a desk that could store all your books, records, and communications, and let you trail through them associatively the way memory itself works. He couldn't build it. But the vision never went away.

Zettelkasten: the original linked note system

German sociologist Niklas Luhmann kept a physical slip-box (Zettelkasten) of 90,000 index cards, each with a unique address and links to related cards. He credited it with making him one of the most prolific social theorists of the 20th century — publishing 70 books and nearly 400 articles [3].

The key insight wasn't the cards — it was the links. Instead of filing notes by topic, Luhmann filed them by relationship. A note about economics might link to a note about evolution because both described self-reinforcing systems. The box became a thinking partner: you'd ask it a question and follow the links to an answer you'd forgotten you already knew.

Sönke Ahrens brought Luhmann's method to a wider audience with How to Take Smart Notes (2017) [4], arguing that the Zettelkasten produces insight not just by storing ideas but by forcing you to restate them in your own words and connect them to what you already know.

Building a Second Brain

Tiago Forte popularised the phrase "second brain" through his online course (2017) and book [5]. His framework, PARA, organises everything by actionability rather than subject:

Forte also introduced the CODE workflow: Capture anything that resonates, Organise it for future use, Distil it to its essence (progressive summarisation), and Express— ship something with it. The second brain is not a museum. It's a production system.

Evergreen notes

Andy Matuschak, a researcher and educator, developed the concept of evergreen notes [6] — notes written to develop and refine ideas over time rather than capture a moment. The key properties:

• Atomic — one note, one idea, fully self-contained
• Concept-oriented — titled as claims or questions, not topics
• Densely linked — connected to related notes so the network does the organising
• Written for your future self — not a fleeting capture, but a considered statement

Matuschak publishes his own notes publicly at notes.andymatuschak.org as a working example. The site itself is navigated by following links — there's no table of contents, only trails.

Tools: from notecards to Obsidian

The second-brain movement accelerated when software caught up to the idea. Roam Research (2020) was the first mainstream tool to put bidirectional links at the centre. Obsidian (2020) followed with a local-first, plain-Markdown approach: your notes are files on your disk, owned by you, readable by anything [7].

Obsidian's graph view made the Zettelkasten metaphor visual — you can watch your knowledge network grow over time. But most users still end up with the same old problem: a large vault full of notes they rarely revisit, connections they meant to make and didn't, and ideas they captured but never developed.

What AI changes

A traditional second brain is reactive: you put things in and retrieve them later. The system itself does nothing between those two moments.

An AI-augmented second brain can be active. Given access to the full vault, an AI can:

• Surface connections between notes you've forgotten you wrote
• Draft the next note from a half-formed idea you sketched last month
• Identify gaps — questions implied by your notes that you haven't answered yet
• Synthesise across dozens of source notes into a single coherent document
• Maintain structure without you having to manually manage tags and folders

This is the shift from a second storage system to a second thinkingsystem. The constraint is the same one Luhmann faced in 1952: the AI needs to see everything to connect anything. That's why Harvis runs Claude inside the vault rather than connecting it to a search index from outside.

Further reading