A mechanism, a source, a software test, and a validation result are different kinds of evidence. The register keeps them separate.
occupational evidencecontextual evidence
Bounded occupational catalog match
Match supplied language to a versioned local occupation record and preserve the match basis.
Current status A small deterministic catalog is available for provisional planning only.
Acceptance The product never converts catalog similarity into a fit or automation score.
design hypothesisdesign hypothesis
Human and AI work redesign
Separate current work from proposed dispositions and add verification, coordination, exception, and accountability work.
Current status Implemented as deterministic planning logic that requires employer confirmation for consequential facts.
Acceptance No proposed automation silently removes authority, exception handling, or review responsibility.
internally tested engineering behaviorexperimental measure
Controlled scenario variation and fault lifecycle
Use a recorded private seed policy to select clean, single, multiple, or interacting fault schedules and ordinary disturbances.
Current status Deterministic generation is implemented without a calibrated-form claim.
Acceptance Injection, exposure, detection, false alarm, propagation, containment, correction, and verified recovery remain distinct.
governance requirementdesign hypothesis
Agent benchmark interpretation boundary
Use benchmark results only to shortlist a model, then require task-specific groundedness, boundary, latency, cost, and failure evaluation.
Current status No live model is selected by this deterministic layer.
Acceptance No benchmark result directly activates candidate rating, automated authority, or a production provider policy.
external research supportdesign hypothesis
Compatibility-gated candidate comparison
Permit structured comparison only for one exact frozen package and form under equivalent administration.
Current status A strict deterministic compatibility gate is implemented.
Acceptance Missing evidence remains unknown or not ratable and never becomes a zero score.
design hypothesisdesign hypothesis
Constraint-based team planning
Evaluate declared team options against explicit constraints and report gaps, duplicated coverage, bottlenecks, and review burden.
Current status A deterministic descriptive arithmetic screen is implemented.
Acceptance The result exposes all failed constraints and contains no performance forecast.
governance requirementdesign hypothesis
Lifecycle governance gates
Require explicit employer approval, instrument readiness, intended use validation, and evidence at separate guarded transitions.
Current status Lifecycle guards are implemented as governance requirements.
Acceptance No draft can be represented as instrument ready or valid for an intended hiring use without the named evidence.
not yet validateddesign hypothesis
TENX empirical evidence state
Reserve this class for results from identified TENX samples, measures, analyses, and study protocols.
Current status No empirical TENX evidence is accepted in this layer.
Acceptance An empty empirical register is displayed as no evidence, never as a null or favorable result.
not yet validateddesign hypothesis
Hiring intended use validity
Keep the intended hiring use inactive until content, response process, reliability, fairness, and governance evidence are all available.
Current status Not yet validated for hiring use.
Acceptance The layer makes no hiring recommendation, performance prediction, or calibrated difficulty claim.