Unlinked Mentions
This is the hardest evidence of everything under Discover: no inference at all, just literal matching.
You have written "attention residue" across a dozen notes, but only one or two carry [[ ]]. To Jotaid the rest are not links yet — absent from backlinks, absent from the graph, and counting for nothing in co-occurrence.
Where You See Them
| Place | What it shows |
|---|---|
| A node's detail view | Notes mentioning this node without linking it |
| A note's info panel | Nodes this note could link, each with a one-click Link button |
The two are the same fact from opposite directions: one looks from concept to notes, the other from note to concepts.
Backfilling
| Button | Scope |
|---|---|
| Link | Converts the mentions in this one note |
| Link All | Converts this node's mentions across every note at once |
Afterwards those notes appear under the node's backlinks immediately, and the graph and matrix update with them.
It is literal, not AI
Backfilling only touches exact name matches, which makes it safe, predictable and repeatable — no network, no cost, works offline. It will not "guess" that you meant a near-synonym; that belongs to the semantic layer.
One Consequence: Names Matter
Because matching is literal, a node named "attention residue" will not catch "the residue of attention" in your prose.
That is not a flaw — the predictability of literal matching is the whole point. Covering near-synonyms is the job of semantic relations, not of loosening the match.
Next Steps
- Co-occurrence & Matrix — How strong the existing relationships are
- Link Predictions — Pairs that never co-occurred but probably should
- Managing Nodes — Renaming and merging