← Back to Research
Provenance
Provenance is the verifiable record of where knowledge and capability came from. It lets a claim or a result be trusted rather than merely asserted.
A machine-readable, versioned chain of origin, authorship and change attached to every definition, claim, capability and result an origin publishes. It answers who said it, from what and when, without requiring trust in how that answer is presented.
Why it matters
- Where agents synthesise answers from many sources, the source is the scarce and valuable thing. Provenance makes it first-class.
- Provenance by default means every projection carries its lineage. The JSON, the HTML and the agent's cited answer all trace to the same origin and version.
- It is the difference between an origin that can be cited and one that can only be scraped. Citability is a moat.
In practice
- Every definition carries id, version, author and updated date. A projection preserves them.
- An agent answering from the corpus returns a citation that resolves to a specific, versioned entry.
- A claim's provenance links to the research or implementation that substantiates it.
Not this
- Content presented as authoritative with no traceable source or version.
- Provenance kept only in footnotes a person can read but a machine cannot follow.
Provenance
| Version | 0.1 |
|---|---|
| Status | Draft |
| Author | Pete Dainty |
| Updated | 5 July 2026 |
| Source | Canonical origin. See its provenance record. |