Knowledge Graphs, Backlinks & Maps of Content

The single property that distinguishes modern PKM tools from text editors is bidirectional linking. Done well, it turns a folder of notes into a graph you can navigate. Pair with pkm-overview-methodologies; see also pkm-zettelkasten, pkm-evergreen-digital-garden, pkm-obsidian-deep, pkm-logseq-deep.

Backlinks 101

★ Forward link — [[Note]] written into a note.
★ Backlink — a panel showing every note that links to this one. Auto-generated from forward links.
★ Why it matters — the backlink panel is the navigation aid. You arrive at a note and see what referenced it; the graph tells you what's adjacent.
Block references — point not to a page but to a specific block / paragraph. Logseq, Roam, Obsidian (with plugins) support this.
Aliases — let one note be linked under multiple names ("ZK", "Zettelkasten", "slip-box").

★ ★ MOC — a curated index note, listing related notes with structure / annotation. The opposite of a tag dump.
★ Nick Milo's LYT framing — MOCs are the connective tissue between atomic notes; less rigid than Zettelkasten's emergence-only stance.
MOCs are notes too. They get backlinks; they evolve.
Dashboards are richer MOCs — Dataview / Logseq queries auto-list notes by tag / property.

★ Local graph (notes within 1–2 hops of the current note) — actually useful navigation.
★ Filtered global graph (e.g., only #literature notes) — useful for spotting clusters / orphans.
Unfiltered global graph — pretty but rarely actionable past 200 notes.
Tools: Obsidian Graph View, Logseq Graph, Dendron, Foam, Org-roam-ui, AnyType graph.
Excalibrain (Obsidian plugin) — explicit hierarchy + relations as a navigable diagram.
Juggl — alternative Obsidian graph plugin with richer styling.

★ Links for individual concepts and instances. [[Vector embedding]], [[Project Q3]].
★ Tags for classifications. #literature, #project, #card.
Don't tag what you can link. Tags are global, links are local; over-tagging is a PKM antipattern.
Nested tags (#area/health) are PARA-friendly.
Dataview / queries can mix both — WHERE contains(tags, "literature") AND contains(file.outlinks, [[Author]]).

★ One concept per note. Otherwise links are ambiguous: "Linking [[That note]]" — which idea are you actually pointing at?
Splitting a fat note is a normal PKM operation. Keep doing it as concepts crystallise.
See pkm-zettelkasten.

Topic MOC — MOC: Vector search lists notes about vector search with sentence annotations.
Question MOC — MOC: How does attention scale? lists notes that bear on the question.
Reading MOC — per-book MOC linking literature notes.
Project MOC — gathers notes feeding into a deliverable; doubles as a draft.
★ MOCs as draft outlines — when a Topic MOC has 12+ entries with annotations, you have a blog post outline.

Obsidian Graph View (built-in) + Excalibrain, Juggl, Breadcrumbs plugins.
Logseq Graph (built-in) + whiteboards.
Org-roam-ui — D3-based graph for Org-roam.
Foam Graph — VS Code, simpler.
Dendron Graph View — schema-aware.
AFFiNE / Heptabase — spatial canvas; not a graph in the strict sense but visually similar.

★ Orphans — notes with no incoming or outgoing links. Either link them or delete.
★ Hubs — notes with very high in-degree; candidate MOCs / split candidates.
Dead-end leaves — notes only linked once; consider integrating into an MOC.
Obsidian: "Find orphaned notes" plugin or a Dataview query.
Logseq: built-in queries cover this.

Link compulsion — linking everything to "Programming" creates a hub no one navigates through.
Tag soup — 200 tags, none used twice. Refactor or kill.
Pretty-graph syndrome — admiring your graph view rather than reading the notes. The graph isn't the value; the notes are.
No MOCs — you can't navigate a 2,000-note vault without entry points.

You're starting a vault: invest in linking from day 1; tags second.
You have 500+ notes and feel lost: add 5–10 MOCs; you'll instantly orient.
You want a graph-shaped tool: Obsidian, Logseq, Org-roam, Foam — all support this natively.
You want hierarchy-first: Dendron, Trilium, BookStack — see pkm-foam-dendron-vscode, pkm-trilium-silverbullet.
Outliner-shape: Logseq — see pkm-logseq-deep.