Cross-source analytics

When your brain has two or more sources, Distillary generates a full analytical layer in shared/analytics/. These aren’t summaries — they’re structured comparisons that treat your brain as a multi-source knowledge graph.

What gets generated

Entity mapping

shared/analytics/entity-mapping.md

Maps equivalent concepts between sources that use different terminology for the same idea. Three categories:

• Same concept, different name — e.g., “Early Adopters” (Lean Startup) = “Crazy customers” (Mom Test). These become bridge entities in shared/concepts/.
• Complementary pairs — concepts that operate at different levels. One source’s abstract loop maps to another’s concrete implementation.
• Unique concepts — ideas that only appear in one source. Shows each book’s distinctive intellectual territory.

This is the foundation for bridge concepts — until you know what maps to what, you can’t connect sources.

Comparison essay

shared/analytics/comparison.md

A synthesis essay (written by an opus agent) that examines:

1. Shared premises — what both sources agree on at a foundational level
2. Unique contributions — what each source adds that the other doesn’t cover
3. Tensions — where the sources contradict or pull in different directions
4. Dependency chain — how the sources interlock: where one picks up where the other leaves off

This isn’t a side-by-side table. It’s an argument about how the sources relate.

Statistical profiles

shared/analytics/statistical-profiles.md

Quantitative analysis of how each author constructs their argument:

• Claim counts — total claims extracted per source
• Certainty distribution — how much is stated as established fact vs. argued vs. speculative
• Stance distribution — how much the author endorses vs. critiques vs. describes
• Role distribution — balance of argument, evidence, methodology, definition
• Priority distribution — core vs. supporting vs. peripheral claims

This reveals authorial style at a structural level. A source that’s 50% established and 40% endorsed argues very differently from one that’s 51% argued — the first presents settled knowledge, the second builds a case.

Entity overlap

shared/analytics/entity-overlap.md

Frequency and centrality analysis across sources:

• Entity counts per source
• Shared entities — concepts that appear in both (often surprisingly few at the name level)
• Top entities by backlink count — which concepts each source organizes its thinking around
• Hub identification — the concepts with the most connections

Shows what each source actually talks about most, which is often different from what they claim to be about.

Set operations

shared/analytics/set-operations.md

Claim-level intersection analysis:

• Overlapping claims — argument pairs across sources that make essentially the same point
• Keyword intersection counts — quantified overlap between claim bodies
• Unique intellectual territory — claims in one source with no equivalent in the other

This is more granular than entity mapping — it works at the individual argument level.

Graph analytics

shared/analytics/graph-analytics.md

Network topology analysis treating the brain as a directed graph:

• Node and edge counts — how large and interconnected each source’s idea network is
• Degree centrality — which concepts have the most direct connections (hub ideas)
• Betweenness centrality — which concepts bridge otherwise separate clusters (information bottlenecks)
• Network density — how tightly connected vs. loosely organized the ideas are

A source with one high-centrality node has a single dominant idea. A source with many moderate-centrality nodes has a flatter, more distributed argument structure.

Bridge concepts

shared/concepts/

Generated from the entity mapping, bridge concepts are unified pages for ideas that span sources. Each bridge page has:

• Both perspectives — how each source defines and uses the concept
• Backlinks from both sources — every claim that mentions this concept, grouped by source
• Aliases — the different names each source uses

Bridge concepts are the multi-source search engine. An agent asking “what does this brain know about customer validation?” fetches the bridge page and gets both sources’ perspectives plus all related claims in one page.

When analytics are generated

Analytics run automatically when you add a second source to your brain. The distillary-combine skill triggers:

1. concept-mapper (opus) — finds same-concept-different-name pairs
2. compare (opus) — writes the synthesis essay
3. bridge-builder (haiku) — creates unified entity pages from the mapping

With each additional source, the analytical layer grows — new bridges form, comparisons expand, and graph analytics reveal how the new source connects to everything already in the brain.

Why this matters

A single-source brain is a structured book summary. A multi-source brain with analytics is a field-level understanding tool. The analytics answer questions no single source can:

• Where do experts agree? (shared premises)
• Where do they disagree? (tensions)
• What does Source A assume that Source B makes explicit? (dependency chains)
• Which concept is central to this field, not just this book? (graph centrality across sources)

The more sources you add, the more the analytical layer reveals. Two sources show a debate. Five sources show a field.