--- title: "保管庫が成る、その直系とグラフとなるネットワーク" seo_title: "The connectome: your vault as a network | SazAI" description: "Your vault is a network: every document a node, every link a synapse declared by the operation that made it." tags: [marketing] --- > A folder with three hundred files holds three hundred things and no relation between them. What your vault holds is the shape: who came from whom, who was compared with whom, and what got recorded along the way.

The vault is a graph

Every document a node, every link a synapse

The connectome is the shape of your vault. Every document that comes in is a node. Every relation between two documents is a synapse. The graph is not a visualization we generated for the page: it is the form in which the vault is stored, and you navigate it on screen.
a nodea document that came in
a synapsethe relation between two of them
a networkthe whole vault, growing with every upload
You do not need to build anything. The network appears because the work happened: every upload adds a node, every translation or revision adds a node and the link back to the original, every comparison ties together the two sides that were compared.

Links are declared, never guessed

A synapse exists because an operation recorded it

This is the difference that decides whether the shape can be trusted. No link in your connectome is born from computed similarity between two files. It is born from the operation that created it: when one document produces another, [the link is recorded right there](documents), with the record of what was done and on which version. The practical consequence is that the network carries no guesswork inside it. A graph built from content proximity draws plausible neighborhoods that nobody can check. This one draws what happened, and every edge has a receipt underneath it. The engine also measures similarity between documents, and that measurement is useful: it tells you the third and the ninth presentations are nearly the same. But it is a signal you read, not a synapse it draws on its own. The two stay separate on purpose.

Two readers, one network

You traverse it on screen; the agent traverses it from within

youtraverse it on screen
the same networkextracted, measured, backed by record
the agenttraverses it from within
On screen, the network is navigation: you enter through one document and reach its translations, its revisions, and the comparisons already made, without remembering where anything was stored. [The agent](talk) reads the same network, and this is where the form matters. It does not swallow all twelve files at once: it enters through one document, follows the links, reads one section at a time, with the origin and the measurement already at hand.

Why a network, and not a block of text

The form in which knowledge is absorbed changes the answer you get

Dropping three hundred whole documents into one question is the most expensive and least verifiable way to answer it. The text arrives with no origin, mixed together, and what comes back has no way to point at where it came from. Reading through the network is the opposite. Knowledge absorbed node by node, the way one of your questions needs it, with each passage still tied to the slide and the shape it came from. That is why the answer can come with its source and its unit: it never stopped having an address. And the network grows with your work, not with our catalog. It knows exactly what you uploaded, and nothing beyond that. Where that coverage ends is stated in [the boundary](trust), before you find out on your own.
Every new node is a document you can open, measure, and compare like any other, and what it carries in X-ray and receipt is in [your documents](documents). The network is the shape; the documents are what it links.