Graph–Chat Synchronisation
In the Workbench, the graph and the chat inform each other. Neither works in isolation — they are two views of the same knowledge, updating together.
What the graph brings to chat
The graph gives the chat spatial context. When you ask a question in the chat, the AI does not just read the content of your notes — it can see where each note sits within your overall structure. A note that is central and well-connected is treated differently from a note that sits alone at the edge of your graph.
This spatial awareness is what makes Workbench chat different from regular chat. You can ask: “What is isolated in my thinking on this topic?” or “Where is the most dense cluster around this concept?” — and get answers that reflect the structure, not just the content.
Example question
“Looking at the cluster around ‘attention economics’ — what tensions does the AI see that I haven't connected yet?”
What chat brings to the graph
Conversations in the Workbench can surface connections the graph has not yet made explicit. When the AI identifies a relationship between notes you have not linked, it can suggest the connection directly — and you can accept it into the graph without leaving the Workbench.
The conversation also becomes part of your session record. The reasoning trail from a Workbench session is preserved — you can return to it and see exactly what you were exploring and what you discovered.
Selecting a node changes context
When you click a node in the Workbench graph, the chat context narrows to that note and its immediate neighbourhood. Your next question is automatically grounded in that specific part of your thinking — you do not have to specify it.
This lets you move fluidly: zoom out to the global graph to spot a pattern, click a cluster to focus, ask a question, follow the answer to a new node, ask again. The graph and the chat stay in step.
You can always clear the node selection to return to full-graph context. The chat remembers the conversation — only the spatial focus changes.