Research Vault / Overview

The System at a Glance

A personal intelligence system that turns scattered saves into a queryable knowledge graph — 6,183 curated resources, organized, enriched, and continuously synthesized.

What It Is

Not a bookmark manager. A queryable memory system built around a data model, not a folder structure — designed to surface connections between ideas, not just store them. The vault is a weighted directed acyclic graph (DAG): each save is a node; topics, connections, and projects are progressively higher-level abstractions. References flow strictly upward — a save can inform a topic, but a topic never references back down to its constituent saves.

Core Capabilities

Ingest

Raw saves from share-sheet + web clipper → Claude-enriched saves with tags, summary, and topic links

Compile

Synthesizes topic articles from individual saves

/compile
Query

Natural-language retrieval across 6K saves + 59 topics

/query
Connect

Finds non-obvious cross-domain connections between topics

/connect
Orchestrate

End-to-end agent pipeline: raw inbox → enriched saves → briefing

/orchestrate
Briefings

Auto-generated daily + weekly context digests surfacing open gaps and insights

Vault Structure

The vault is organized as a weighted DAG. Content flows from raw saves upward through topic synthesis and cross-domain connection into active project context.

00_Inbox (raw saves) LLM enrichment 01_Social_Memory (6,183 enriched saves) TOPIC-INDEX (128 leaf indexes) /compile 02_Topics (59 topic articles) /connect 04_Connections (43 connections) /query match 03_Projects (active projects) Infrastructure Logs CHANGELOG GAPS SUGGESTIONS INSIGHTS append-only · auto-indexed

Content Flow

A single save travels from raw file to synthesized knowledge in three steps.

Share-sheet or Web Clipper 00_Inbox (raw .md file) Claude API enrichment 01_Social_Memory (enriched save) /compile topic synthesis 02_Topics (topic article) /connect cross-domain 04_Connections (connection note) Action node Storage node

In Numbers

Metric Count
Curated saves 6,183
Topic articles 59
Cross-domain connections 43
Topic indexes (TOPIC-INDEX) 128
Operational learnings 201
Architecture Decision Records (ADRs) 73
Test suite 985 tests
Open suggestions 95
What is an ADR?

Architecture Decision Records (ADRs) document key technical choices: the problem context, every option considered, a decision matrix, and the final rationale. In the Research Vault, ADRs capture decisions like “why append-only logs over a single mutable file,” “why cherry-pick over full merge for vault commits,” and “why BFS/YAML-first navigation over brute-force search.” 73 ADRs means 73 decisions that future sessions — human or AI — can read and reason about without re-litigating them.