Nouns, Verbs, and Markdown: Structuring an AI Second Brain as a Knowledge Graph

Jake Van Clief argues that most people use their second brain wrong. He builds second brain systems for multi-million dollar companies and says he sees the same mistake in both corporate deployments and personal setups: people dump everything into memory as one undifferentiated blob. His fix is to treat the second brain as a knowledge graph, where every entity and every relationship gets broken down and mapped to its own file.

The video walks through the theory on a whiteboard, then shows what it looks like in practice: a real folder of markdown files and a rendered graph of the resulting node network on his monitor.

What a Second Brain Actually Is

Van Clief's working definition is simple. Any time you talk to your AI, it remembers past work you have done. You can throw research papers and notes into it, and it stores them in a memory base you can always access. The video opens on exactly this: a node graph on his screen with the caption "How to build AI Second Brains (like the BIG companies do)."

Most second brains you see are built on graph theory. A graph is made of vertices and edges, which he calls nodes and edges. On the visualizations, all the dots are nodes and all the lines are edges. The question he wants you to ask is what those actually mean: why do they connect that way, and what tells the computer to do that?

Nodes Are Nouns, Edges Are Verbs

His mental model, sketched out on the whiteboard next to a drawing of a brain: nodes are nouns (a person, a place, a team) and edges are verbs (the relationships between nouns, or the things being done, like running, getting, consuming).

Applied to a company: each team becomes a node, and each relationship between teams becomes an edge. The product team node "talks" (edge) to the data team node. Each team also has workflows, and a workflow like "collecting questions from a Slack channel" fits the same pattern.

He writes the full example sentence on the whiteboard:

The product team runs the Slack Questions workflow, which consumes the ticket information and produces answers.

Every noun and verb in that sentence can be mapped to a set of files. This is where most second brains go wrong: they map the whole sentence at once to a bunch of random files. Good second brains break each noun and verb down individually.

From Sentence to Folders and Files

The practical implementation is plain markdown in a folder structure. He shows his own vault on screen: a teams/product directory containing folders for data, jobs, and processes, plus markdown files like Product.md and workshop.md.

Inside the product team folder, markdown files describe processes, ideas, jobs, and concepts. Each file also includes descriptions of where that data, those processes, and those ideas connect to everything else. Repeat this process enough times and you end up with a giant graph of nodes and edges that describes how your company works, who the people are, and what they are doing. That graph is the second brain: a place of memory the AI can navigate.

Key Takeaways

• A second brain is persistent memory for your AI: past work, notes, and research it can reference in every conversation.
• Model it as a graph. Nodes are nouns (people, places, teams), edges are verbs (runs, consumes, produces).
• Do not map whole sentences or ideas to random files. Break every noun and verb into its own markdown file.
• Each markdown file should describe the entity and where it connects to everything else. The links are what build the graph.
• Done consistently, the folder of markdown files becomes a navigable map of how your company works.

Resources

• Jake Van Clief on TikTok -- he offers free lectures on structuring folders, files, and data via the link in his bio, with a community of 40,000 people learning the methods.