Four layers, one measurement.
AHCOS turns raw evidence into a signed, portable estimate of capability. Four layers do the work.
Evidence in. A signed estimate out.
Evidence Graph
Evidence, captured at the source. Responses, work, and outcomes enter as structured evidence. Every signal stays traceable to where it came from — the foundation everything above it is built on.
Measurement Engine
A BKT + IRT ensemble infers capability. Bayesian Knowledge Tracing and Item Response Theory run together to estimate latent capability and the uncertainty around it. The number never travels without its confidence.
Adaptation Loop
The estimate is never stale. As new evidence arrives, the posterior updates continuously. Capability is treated as something that moves — and the measurement moves with it.
Credential Layer
Signed, portable, owned by the person. The estimate is issued as a cryptographically signed W3C Verifiable Credential the person holds and can present anywhere — interoperable from day one.
Note The model reports confidence, not just a number. An estimate always carries its own uncertainty.
Production, not prototype.
AHCOS is built as multi-tenant, production-grade infrastructure — not a one-off model. The same measurement layer serves many institutions at once.
That design is also the moat. Each institution that measures with AHCOS contributes to a capability graph, and that graph compounds — making cross-institutional transfer of capability signals possible over time.