Two entities are connected when they appear in the same document. The connection strength (edge weight) is the number of documents they share.

For example, if Jeffrey Epstein and Bill Clinton both appear in 502 documents, their connection strength is 502. Stronger connections appear as thicker lines in the graph.

The graph uses a force-directed layout, which positions entities based on their connections. Entities with many shared connections cluster together naturally.

The filters control which entities and connections appear in the graph. Each filter has trade-offs between completeness and performance.

When you focus on an entity, the graph shows only that entity and their direct connections. This is useful for exploring someone's network without the noise of unrelated clusters.

In focus mode:

• Only the focused entity and their connections are shown
• Direct connections (1-hop neighbors) appear around them
• Additional important entities may appear (2-hop neighbors)
• All visible entities have some path to the focused entity

To focus on an entity, click on them in the graph and select "Focus Graph on This Entity" from the details panel that appears.

Sometimes you may see groups of entities that appear disconnected from the main network. This is usually caused by the Max Edges limit, not missing data.

Here's what happens:

1. The graph queries connections sorted by strength (highest first)
2. It stops after reaching the Max Edges limit
3. Weaker connections that would bridge clusters get cut off
4. Entities that only have weak connections to the main network appear isolated

To fix this: Increase the Max Edges slider, or focus on a specific entity to see their complete neighborhood.

The documents in this archive come from several public sources:

Entities (people, organizations, locations) are automatically extracted from documents using Named Entity Recognition (NER). This process:

• Scans document text for names and organizations
• Groups variations of the same name (e.g., "Bill Clinton" and "William Clinton")
• Links entities to external databases like Wikidata when possible
• Computes connection strengths based on document co-occurrence

Automated extraction isn't perfect — some entities may be misidentified or missing. The data is continuously refined.

The network graph uses WebGL to render thousands of entities in real-time. Mobile devices have limited GPU memory, so we automatically adjust the defaults:

For the full graph with all connections, we recommend using a desktop browser. On mobile, you can still explore individual entities and their direct connections.