Distillary

Entities and Bridges

4 min read

Entities and bridges

Entities and bridges are the connective tissue of a brain. Claims are what sources argue. Entities are what sources reference. Bridges are where sources agree without knowing it.

Entities

An entity is a named thing: a person (Eric Ries), a concept (Build-Measure-Learn), a company (IMVU), a work (The Lean Startup), a place (Silicon Valley). Entities get their own notes with descriptions and, crucially, backlinks from every claim that mentions them.

Why entities matter

Without entities, a brain is a tree — each claim connects to its parent and children, but there’s no way to jump sideways between branches. Entities create the lateral connections. “Pivot” appears in claims about strategy, measurement, organization, and growth — the entity page is the hub that connects these branches.

Realized vs ghost entities

Not every concept gets a note. An entity is “realized” (gets its own file) only if:

  • It’s a proper noun (a named person, company, work, place, or event)
  • It appears in 3 or more distinct claims
  • It appears in any priority/core claim
  • The source explicitly defines it as a technical term

Everything else stays as a [[ghost link]] — wikilinked in claim bodies but with no file backing it. Ghost links appear as faded nodes in Obsidian’s graph view, marking the frontier of explored territory. They cost nothing (no file, no maintenance) but make the conceptual depth visible.

This is intentional. Creating a note for every mentioned concept would drown the vault in stubs. Keeping only the important ones as realized entities keeps the graph readable while still showing what’s been noticed but not yet explored.

Promoting a ghost

When a ghost concept catches your interest, you can promote it — create a full entity note from every claim that references it. The explore agent suggests which ghosts are worth promoting based on mention frequency and your annotation patterns.

Every entity page has a “Referenced by” section listing every claim that mentions it, grouped by source. This means entity pages answer “what does my brain know about X?” without any search query.

An entity with 30 backlinks from The Lean Startup and 20 from The Mom Test is a hub connecting both sources’ perspectives on the same concept. You don’t search for information — you navigate to the entity and read what’s linked to it.

This is why Distillary doesn’t need a full-text search engine. The link graph built during distillation IS the index. Entities are the entry points. Backlinks are the results.

Bridges

A bridge entity exists when two sources discuss the same idea under different names. The Lean Startup calls unreliable indicators “Vanity Metrics” (quantitative: gross numbers). The Mom Test calls them “Compliments” (qualitative: praise that costs nothing). Both describe the same problem — misleading signals.

The bridge concept “False Signals” unifies them.

How bridges work

A bridge note has:

  • Aliases from both sources — so [[Vanity Metrics]] and [[Compliments]] both resolve to the same page
  • Both perspectives described — the Lean Startup view and the Mom Test view
  • Backlinks from both sources — every claim from either book that discusses this concept
  • A source/cross-vault tag marking it as cross-source synthesis
graph LR
    LS["Lean Startup:<br/>Vanity Metrics"] --> Bridge["Bridge:<br/>False Signals"]
    MT["Mom Test:<br/>Compliments"] --> Bridge
    Bridge --> Backlinks["Both perspectives<br/>+ all backlinks"]

Why bridges matter

Without bridges, a combined brain is just two disconnected trees side by side. The sources might discuss identical concepts but under different names, and no human or agent would know they’re related.

Bridges make the conceptual overlap visible. They’re the answer to “how do these sources relate?” — not in the abstract, but concretely: this concept in book A is the same as this concept in book B, here’s how each describes it, here are all the claims from both.

For agents, bridges are the fastest path to cross-source answers. Instead of walking each source’s pyramid separately, the agent jumps to the bridge and gets both perspectives in one fetch.

How bridges are discovered

The concept-mapper agent (opus) reads entity files from both sources, compares descriptions and aliases, and identifies pairs that represent the same idea. This requires reasoning — “Genchi Gembutsu” and “Customer conversation” have no words in common, but they both mean “go directly to the source of truth.” Only a reasoning model catches that.

The bridge-builder agent (haiku) then creates the actual bridge notes, inherits aliases, and adds cross-references to the original entity files.

Bridges grow incrementally

When a third source joins the brain, the concept mapper runs against both existing sources. A concept that two sources discuss under three different names gets one bridge with three aliases and backlinks from all three. Each source makes the brain more connected.

Graph View

Backlinks