Capture
One mixed voice/text recording, intact. Dotted underlines are ghost topic ranges — annotations only, the text is never altered.
Proposed inquiries
Model read of the transcript: three candidate threads. Nothing is created until you stage it.
No sessions yet
Stage a split from the capture panel (or keep it together), then accept it here. The frontier appears left-to-right: capture → inquiries → syntheses.
Staged — not yet accepted
Session
Select a node in the graph to open its session — transcript excerpt, resources, context packet, and a working mock conversation.