Platform
How an assessment is produced.
Two kinds of claim, one governance layer, and a registry that publishes its own gaps.
Ingested claims
A record from an issuing authority, carried with its source, its classification and the moment it was read. It asserts only what the registry asserts; nothing is inferred at this layer.
Computed claims
Derived through a registered methodology from ingested claims. This is where our work lives, and it is the layer a competitor cannot license from anyone.
A methodology registry, published with its gaps
Every methodology carries an identifier, a version and a validation status — including the ones not yet validated. We publish which is which. A validation status you have not earned yet is the cheapest credibility test there is, and almost nobody passes it.
A governed pipeline, then a person
Multi-layer governance and a multi-agent orchestration layer operating under policy gates. Nothing that reaches a certificate is model-generated prose. A qualified reviewer signs before release, and the system will not issue without one.
Publication is a positive gate
A certificate reaches the public collection only by affirmatively earning it: a human review record, a non-waived reviewer role, a narrative that does not disclaim its own review, and valid status. The builder fails closed. Absence of a reason to withhold is not a reason to publish.
Not another AI tool
Every model improvement raises the supply of plausible analysis. None raises the supply of parties willing to be accountable for one.
The commodity
Generated analysis is heading toward zero cost and unbounded supply. Nobody is accountable for any of it, and the reader knows.
The scarcity
A named party who signs, and a reader who can check without trusting either side. That cannot be generated, at any model scale.
Where we sit
Nothing that reaches a certificate is model-generated. Computation is deterministic and cited; a person reviews it; the signature is cryptographic.