Evidence standards

Evidence Standard v1.1 — Durable Source Capture.

Durability: a source must be re-checkable

A source supports a published claim only if it satisfies one of three methods:

Method A — External archive

A working archive snapshot that resolves to the cited content.

Method B — Publisher artifact

An official PDF, filing or static artifact hosted by the publisher, with its SHA-256 recorded.

Method C — Source snapshot

Our own capture, with timestamp, original URL, content type, SHA-256 of the captured bytes, exact locator and a reviewer who confirmed correspondence.

What durability does not do

Durability answers one question: can this be checked later. It never changes source independence or class, never upgrades an evidence tier, never turns a teleoperated demonstration into autonomy, never turns an announcement into a deployment, and never makes predecessor-model evidence apply to a current model.

Integrity

Snapshots are append-only. A re-capture is a new record with a new hash; the previous one is preserved and marked superseded. If stored bytes stop matching their recorded hash, durability fails and everything citing that source is blocked from publication.

Second review

Claims and deployments require review by someone other than the researcher who drafted them before they can become publication-eligible.

Corrections

Anyone, including manufacturers, may open a profile and submit a correction or new evidence. Submissions enter the research queue. They never change a score directly, and scores cannot be purchased.