The Neural Map, your memory seen as a graph

The Neural Map shows your memory as a graph. Each node is a memory, each link joins two memories the engine finds close to each other. It is a way to see what you have accumulated without asking a question: the clusters, the bridges between domains, and the places where there is nothing.

It renders two ways, a 3D space and a 2D plane, with exactly the same gestures.

The gestures

Left-click and drag are navigation only: rotate, zoom, pan. They select nothing.

Double-clicking a node opens a panel with its content. Panels stack, move by their header, and the one you interact with comes to the front.

Right-clicking freezes the graph and opens a menu at the cursor. Auto-orbit pauses while the menu is open.

A single click selects nothing, and that is deliberate. With thousands of dense nodes, what it caught was effectively random. Opening a node became an explicit gesture.

What the controls change

The connection threshold decides how similar two nodes must be to be linked. Low, the graph is dense and shows distant resemblances. High, only strong links remain.

The node limit bounds what is displayed, without removing anything from your memory.

Grouping arranges nodes by memory type or by origin vault. Same graph, laid out differently.

"Cross-vault only" keeps just the links that cross a vault boundary. That is the view answering "what do my work and my notes have in common".

What the map is not

It is not a search. It shows the shape of your memory, not the answer to a question: the chat is what queries the vaults for that.

And it is not a privacy screen. A vault is removed from the map when its manifest says so, which is the case for sandboxes and measurement vaults. Marking a vault MAXIMUM governs where its content is allowed to go, and does not take it off the map: the map is drawn on your machine and goes nowhere.