The falsifiable claim
On a labelled corpus you choose, in a domain you choose, held entirely inside your walls: the certified threshold λ derived from your stated (α, δ) will hold the realised miss-rate at or below α, at confidence δ. Finite-sample, distribution-free, assuming only exchangeability of the calibration set.
How it runs
The corpus, the domain, and the operating point (α, δ). Nothing is pre-selected by us.
The protocol is one page. Your actuaries execute it inside your own environment. We never see the data.
The bound holds, or you have falsified the frequency measure this class would be rated on.
"If the bound fails, we would rather know than sell."
What validation gates
Validation gates everything. Launch-position discussions follow results, not precede them. The rights structure a first partner would license only becomes a conversation once the bound has held on evidence a counterparty controls.
What you would be licensing →The second experiment
The Meaning-tier signal carries its own separately pre-registered experiment, with named metrics and thresholds fixed before it runs. Your AI due-diligence team is the most qualified adversarial reviewer it could have, and we would rather it broke the claim on your corpus than have a customer discover the limit later.
See the graded status of every layer. Coverage analysis → · Evidence Centre →
What we don't need
No production data leaves your walls. The protocol runs where your data already lives.
No NDA is required to read the protocol; the statistical procedure is published academic work.
A specimen certificate is available for actuarial review. See the certificate →
Triodian Pty Ltd is not an insurer, does not hold an Australian Financial Services Licence, and does not issue, arrange, or advise on insurance products. The pages in this section describe technical infrastructure and invite partnership discussions with licensed carriers, reinsurers and underwriting agencies.