Evidence-grounded capability modeling

Capability claims you can inspect.

Capability Lab is a governed research framework for representing what was observed, what the evidence can support, what remains uncertain, and what may be worth exploring next — without turning the model into authority over the person.

Understand the model Follow the governed pipeline

Append-onlyevidence and evaluation history

Explicit gateshuman review and runtime admission

Complete historyno hidden latest-wins shortcut

Read ≠ authorityprojection never grants permission

A model that preserves the difference between knowing and deciding¶

Most capability systems collapse several questions into one score: what happened, what it means, how confident the system is, what the person should do next, and whether they are allowed to act.

Capability Lab keeps those questions separate.

observation
!= evidence
!= interpretation
!= claim
!= evaluation
!= capability state
!= accepted state
!= current state
!= progression advice
!= product/read projection
!= permission or authority

01 · Mental model Understand the layers

See why evidence, uncertainty, conflict, state, and authority are deliberately different objects.

02 · Provenance Follow the governed pipeline

Trace a reviewed external observation through evaluation, state, progression, and the product/read boundary.

03 · Integration Consume without taking authority

Use the stable PR11.11 read boundary while preserving fresh validation and governance semantics.

The complete public path¶

flowchart LR
    O[External observation] --> R[Human-reviewed evidence]
    R --> C[Capability claim]
    C --> E[Governed evaluations]
    E --> S[Derived + persisted state]
    S --> A[Explicit acceptance]
    A --> X[Current-state selection]
    X --> P[Advisory progression]
    X --> V[Complete current profile]
    P --> W[Governed product/read snapshot]
    V --> W

PR12.13 proves the generic path reaches governed current state without bypassing the existing gates. PR12.14 extends that proof through advisory progression, complete current profile, and the PR11.11 product/read snapshot.

Read the end-to-end current-state audit · Read the product/read audit

The boundary is part of the product¶

Capability Lab does not compute a person's value, destiny, professional license, permission, or human worth. SUPPORTED is not mastery. CURRENT is not readiness. Progression is advisory. A rendered product view does not become capability write-back authority.

Read the constitution

Built as a reference system, not a ranking product¶

The current checkpoint is intentionally conservative. It preserves unknowns, permits unresolved conflict, retains historical evaluations, requires explicit state acceptance and current selection, and treats serialized artifacts as audit data that must revalidate against live governed sources.

For the underlying contracts, start with the architecture, then use the reference map to reach the detailed PR-specific documents and historical Pilot material.