Cortex

A Knowledge Graph Is Shared Memory

2 min read

A Knowledge Graph Is Shared Memory

If an AI needs a durable place to remember, what should that place look like? The answer we’ve landed on, after years of daily use: a graph of small, linked notes.

The rules are simple. Each note holds one idea, with a title that says what the idea is. Notes link to related notes, so the connections between ideas are recorded, not just the ideas themselves. Tools like Obsidian made this way of working popular for people; it turns out to be just as good for AI — better, in fact, than a database or a pile of documents, because a note that links to its context can be found through its relationships, not just by searching words.

The part that matters most: the graph is shared. You read and write the same notes the AI reads and writes. When the AI finishes work, it records what happened as notes you can open. When you capture an idea, the AI can find it, link it, build on it. The knowledge graph becomes the common ground where human thinking and machine work actually meet — one source of truth instead of your notes over here and the AI’s context over there.

This site is itself a small public example: one idea per note, links carrying the relationships, a graph view showing the shape. The same pattern, at full scale and kept private, is how an AI system remembers a business.

Graph View

Backlinks