ifp

← Garden Patch Home · Patterns

• is_a::[[Pattern Form]]
• has_status::[[Seed Stage]]
• in_domain::[[Deep Context Architecture]]

Knowledge Compounds Through Typed Edges Not Filing

Context

Researchers spend 75% of publication time on reading, compiling, and filing — not writing. The median publication requires 177 hours, most of which goes to activities that should benefit from prior work but don’t. A researcher’s tenth year repeats the third year seven times — not from forgetfulness but from lack of persistent structure connecting what was learned across projects.

Forces

• Citation managers store references but not the reasoning connecting them
• Note-taking tools store thoughts but not their provenance
• Filing systems create false completeness: the information was captured, so the work feels done
• Each new project starts from scratch because the connections between prior insights are not encoded in a traversable form
• Knowledge accumulates in volume but does not compound in density

Solution

Knowledge compounds when each new insight is wired into the existing graph through typed edges — not merely stored adjacent to prior notes. The typed edge (supports::, contradicts::, extends::, extracted_from::) carries the relationship’s nature, making traversal meaningful rather than associative. The graph grows in density, not just volume.

The distinction between filing and wiring: filing places a note in a location (folder, tag, category). Wiring connects a note to specific other notes through a named relationship. Filing answers “where does this go?” Wiring answers “what does this connect to, and how?”

Consequences

The writing phase for a new work becomes a query against existing structure, not a recompilation from scattered notes. The writer asks the graph: “what do I already know about X, what supports it, what contradicts it, and where are the gaps?”

Retraction or invalidation of any node ripples through edges to every dependent node. The graph makes the impact of a revision visible.

Cross-domain synthesis becomes a graph operation: find structural parallels between subgraphs. This is the mechanism behind [[Swanson Linking — Undiscovered Public Knowledge]]↑ — discovery through traversal rather than serendipity.

The cognitive work shifts from filing to judging — validating machine-proposed connections rather than creating the connections manually. This is a different skill and a different bottleneck.

The system works not just for one project’s knowledge but for a career’s accumulated knowledge, with confidence tracking that persists and updates across decades. The temporal dimension is what distinguishes compounding from accumulation.

Sources

• Cornelius (2026). “Research Graphs: Agentic Note Taking System for Researchers.” X article, @molt_cornelius, 2026-03-09. (177 median hours per publication, 75% on non-writing tasks.)

Relations

• extracted_from::[[Research Graphs and Precinct Architecture]]↑

• relates_to::[[Swanson Linking — Undiscovered Public Knowledge]]↑
  • Swanson Linking is the cross-domain instance of this pattern: typed edges enable discovery across literature boundaries that filing cannot.
• relates_to::[[Informal Edges Poison the Graph]]↑
  • The solution depends on edge quality — typed edges that carry real semantic weight. Informal or redundant edges create noise that degrades traversal rather than enabling it.
• relates_to::[[Ghost Links as Garden Planning Tools]]↑
  • Ghost links are the garden’s mechanism for identifying edges that should exist but don’t yet — planned connections in the graph.

This site is open source. Improve this page.