#
The knowledge graph is the engine that connects everything you capture. Unlike traditional graph views (which show a visual "hairball" most users ignore), Xentropy's graph works behind the scenes — surfaced as narrated connections in context.
#
Every piece of content you capture becomes a node in your personal knowledge graph. When the AI detects relationships between nodes — shared topics, referenced people, similar ideas — it creates edges between them.
[Voice Note about Q3 Planning]
│
├── relates to ──► [Meeting Notes: OKR Review]
├── mentions ────► [Person: Sam R.]
└── contradicts ──► [Note: "We should delay the launch"]
The graph is stored in a Postgres + pgvector database, enabling both structured queries ("find everything about project X") and semantic similarity ("find notes that feel like this one").
#
Rather than showing a visual graph, Xentropy surfaces connections where they're useful:
- Related notes appear at the bottom of each note — "This relates to your note on pricing because you raised the same objection"
- AI Chat uses the graph to ground its answers in your actual knowledge
- The Wall shows connections through cross-references and suggested links
Design principle: the graph is an engine, never a screen — we keep the value without the hairball.
#
| Trigger | How |
|---------|-----|
| Semantic similarity | pgvector embeddings — notes with similar meaning are linked |
| Shared entities | People, organizations, and projects mentioned across notes |
| Explicit references | Manual [[wiki-links]] or cross-note references |
| AI analysis | The connection engine detects contradictions, follow-ups, and duplicates |
| Capture proximity | Notes created in the same session are weakly linked |
#
One of the most powerful features: Xentropy can flag when a new capture contradicts something you previously believed.
Example: You capture "We should cut engineering spend by 20%" — the AI surfaces your note from last month: "Our top priority is hiring three more engineers."
These contradictions are surfaced as a gentle flag, not an alarm. You can resolve them, dismiss them, or let them sit as open questions.
#
Your graph is yours alone — it grows and evolves with your thinking. There's no shared schema, no template, no setup. It learns from what you capture, not from configuration.
Coming soon: shared team graphs for collaboration, and a personal model that captures your recurring themes, questions, and writing voice.