Argument flow
The default mode and where to start, laying out how your reasoning moves from sources, through the notes and questions that use them, toward the models that make claims. Models appear in their own sage rail only when the model has an explicit project-level link to a source or research question. Those links are created from the Builds on sources and Addresses questions controls on the model canvas. Connections draws one model node and one dashed edge for each of those declared links; it deliberately does not copy the concepts and edges inside the canvas into the project graph. A model without either link stays out of the default flow and is counted in Graph audit instead. Read it for gaps, since a question with nothing flowing into it has no evidence behind it while a cluster of sources with nothing flowing out is reading you have not used yet.Synthesis health
Where your library hangs together, and where it is a pile of separate papers. This mode is the most uncomfortable and the most useful, because synthesis is what literature reviews get graded on and the hardest thing to judge in your own work, since from the inside a reading list feels synthesised given that you know how the papers relate. What this shows is whether that relationship exists anywhere outside your head, in a note, a question or a model. A region of low synthesis health is not a failure so much as the next paragraph of your literature review, already located.Contradictions
Sources that disagree with each other. Contradictions in a literature are not a problem to smooth over but usually where the interesting question lives, and a review reporting consensus where there is a live dispute is a weak review, with a viva being exactly where that gets found out. This mode hands you the disputes you collected without noticing.Bridges
The sources holding otherwise separate parts of your library together. A bridge paper connects two clusters that would otherwise have nothing to do with each other, and bridges carry more weight than their number suggests. Your framing often depends on them, and losing one, whether because it turns out to be weak or because it is contested, can split your argument in two. Better to know which they are before somebody else works it out, and they are also the papers to read most carefully, given how much rests on them.Explore
The free-form force-directed graph in 2D or 3D, where you filter by node type, click through to any item and move around. The controls are listed in-app, so you drag to pan and move nodes, scroll to zoom, shift-click to select several, and right-click for a context menu that navigates or filters to connected nodes, while in 3D you can rotate the view. This mode has no thesis, being for looking rather than answering, though it does well by the “what does my project look like” question that is hard to ask precisely, whereas the other four modes are for answering questions.The synthesis bar
The gold bar reads the whole graph and tells you what it shows, and like the readings on Map: Overview it is composed from the graph’s own facts instead of generated, so it will not invent a connection that is not there. Select any node and an inspector reads it and offers to carry it into the Advisor, pre-filled, which is the intended workflow: find something odd in the graph, then go argue about it with something that has read the sources.Making the graph better
The graph is built from real links, so it is only as rich as the links you have made, and you can thicken it three ways:- Cite sources in your notes with
@, as covered in Notes. - Link notes to each other with
[[. - Use Builds on sources or Addresses questions on a model canvas when the whole model belongs in this project-level graph. Concept-level source bindings remain inside the canvas and its source detail views.