Co-occurrence & Matrix
Co-occurrence is a plain fact: two nodes appearing in the same note. The moment you write [[A]] and [[B]] together, they have co-occurred once.
The graph draws that as a line; the matrix lays it out as a grid — one dataset, two readings.
Connection Strength
A cell's depth is the strength of the relationship: how many notes the two concepts share. On the Mac, hovering over a cell shows that number.
- Two concepts that keep turning up in the same notes → a deep cell
- Two that have met in a single note → the palest visible cell
The scale comes from your own project. The pair sharing the most notes is the deepest cell, and the rest are measured against it on a log scale, so the step from one shared note to two or four stays visible while a few very strong pairs cannot wash everything else out. In a project where every pair has met only once or twice, every cell stays pale instead of one of them being crowned the strongest.
Open a pair and its panel also gives the overlap: the shared notes as a share of all the notes that mention either concept. Overlap answers a different question — whether two concepts almost never appear apart — and it is what the high-similarity warning below is based on.
The colour language holds across the app: deeper blue = stronger, so moving between the matrix and the graph needs no relearning.
Reading the Structure
| What you see | What it means |
|---|---|
| Bright blocks on the diagonal | A cluster — these concepts all relate strongly; brackets mark its bounds |
| Bright cells between blocks | A bridge — a relationship joining two otherwise separate clusters |
| Largely empty rows/columns | Isolated nodes, barely co-occurring with anything |
Open a cell and the panel says whether the pair sits in the same cluster or different clusters. Strong cross-cluster relationships are often the most interesting — they join two lines of thought.
The High-Similarity Warning
Pairs whose overlap is very high — they almost never appear apart — are flagged High Similarity in the warning colour. Overlap that heavy usually means one thing:
These may be two names for the same concept.
"User retention" and "retention rate", say. The right move there is usually not to link them but to merge the nodes.
Filtering the Noise Down
With many nodes the matrix turns to mush. The filter panel offers three dials:
| Filter | Use |
|---|---|
| Minimum co-occurrences | Pairs that co-occurred once are mostly incidental; 2 or 3 clears things up |
| Strength range | Filters by overlap (0–1): raise the floor for pairs that are nearly inseparable, or lower the ceiling to hunt loose ones |
| By badge | Restrict to Core / Bridge / Emerging |
Using It
- Go to Node view and switch to Matrix (on iOS, the segmented capsule at the bottom toggles Graph / Matrix)
- Hover (macOS) or tap (iOS) a cell for that pair's details
- The panel also lists each node's strongest partners, so you can keep following the thread
When the matrix starts paying off
Patterns begin to surface at around 10 notes with backlinks — the more links, the clearer the structure. With Pro and a key, you can also have the AI explain what is behind a bright cell.
It Sees the Current Project Only
The matrix, the graph and the network summary all cover the currently selected project.
Next Steps
- Link Predictions — Pairs that never co-occurred but might belong together
- Knowledge Graph — The same data as a network
- AI Interpretation — Having the AI explain a strong relationship