Every existing provenance system evaluates the container — the credibility of whoever repeated the claim. A claim originating from a Russian bot farm scores green because a sitting VP repeated it. The origin is laundered by the launderer's authority.
Time Zero traces semantic claims backward through time to find the original injection point — and rates that source, not the laundering entity.
Claims mutate as they travel. Exact string matching fails. The system extracts the invariant behavioral schema before traversal begins.
Test Case: "Somalians in America are eating cats and dogs" — repeated by the VP of the USA.
This schema captures all mutations: "Haitians in Springfield eating pets," "Africans in Ohio eating ducks" — all resolve to the same vector. Traversal uses cosine similarity (>0.85 threshold) against the schema embedding, not the raw text.
Viral claims are finite. Millions of agents query the same 200 schemas. The system does the expensive work once, then serves every future query from cache.
The system doesn't persist the entire crawl. It saves only the critical laundering path — the minimum viable chain from Time Zero to mainstream injection. A user can click through every node.
The Time Zero node is evaluated against the user's Value Matrix. The VP node is not evaluated for credibility — it is evaluated only as an amplifier. The authority of the launderer does not transfer to the origin.
TIME ZERO ORIGIN RATING — Private FB Group Screenshot
The original Facebook post propagated as an image — breaking all text-based traversal. Screenshots are handled via a cost-controlled triage pipeline: