In development · Early access
Human Graph
World-wide genealogy, built differently.
Existing genealogy systems fail because they tie identity to records instead of evidence. They duplicate people. They split living from deceased. They force canonical truth or leave it absent. They make collaboration unauditable.
Human Graph is a civilization-grade, contribution-based, AI-assisted identity graph for human lineage. Multiple representations of a single human converge into a canonical identity through evidence, graph structure, and verifiable contributions. Free to use. Free to contribute. Exportable as a JSON graph. Durable across generations.
The graph is truth-seeking, not truth-imposing. Canonical is not declared. Canonical emerges.
What's different
Evidence-first model
Every change enters as evidence. Assertions reference evidence. Identity updates only happen through accepted assertions. Nothing mutates an identity directly.
Canonical emerges from clusters
Probable matches are detected before a node is created. Candidates are surfaced before a duplicate gets built. Within a cluster, the canonical node is the highest-confidence representation — cluster ≥ 0.75, node ≥ 0.80, two independent sources minimum.
Soft canonicalization
Non-canonical nodes are never deleted. They become aliases or linked variants. The system preserves all competing representations. Merges are event-based, reversible, and auditable.
Conflicts are first-class
If contradictions exist, the node status is Contested. The system does not force a canonical. Multiple interpretations may fork; the canonical remains the best-supported version.
AI assists, never decides
AI proposes matches, suggests clusters, and recommends merges — with reasoning, evidence references, and confidence scores attached. AI cannot finalize canonical status without human validation.
Living people stay private
Living individuals are isolated using tokenized linking for relationships. When a person becomes deceased, the system attempts cluster reconciliation and generates merge proposals.
Always explainable
Why is this node canonical? What evidence supports it? What conflicts exist? Who contributed? What changed? The system answers all of these for every identity, every time.