Mission

Objective

Obtain two blinded, independent construct and probe reviews for ET-004-P01 and ET-014-P01, then validate, revise, or quarantine each pilot.

Why this matters

Designer calibration cannot establish construct validity. This is the current gate for all capability baselines.

Scope

Included: reviewer recruitment, preregistered materials, independent labels, agreement analysis, post-capture adjudication, immutable dispositions. Excluded: agent capability runs and model comparisons.

Existing context

research/frontier/records/RFR-001.md, research/evaluation/13-independent-review-protocol.md, and NX-001 in research/analysis/non-human-next-experiments.md.

Initial hypotheses

Independent reviewers will find both pilots valid after bounded revisions and will agree on material probe labels. A quarantine or unresolved disagreement falsifies promotion readiness.

Required evidence

Original independent reviews, blinded labels, agreement table, adjudication, task/grader hashes, and final pilot disposition.

Constraints

Reviewers must not have designed the tasks or graders. Preserve reviews before discussion and do not simulate independence with repository agents.

Execution instructions

Follow the independent-review protocol and the frozen unseen-outcome portion of NX-001. Record revisions without overwriting prior versions.

Deliverables

Cycle report, immutable review records, agreement analysis, and updated evaluation registries.

Success criteria

Two independent reviews per pilot, no concealed disagreement, and an explicit validate/revise/quarantine decision for each pilot.

Stop conditions

Stop when both pilots have a defensible disposition or reviewer independence cannot be obtained.

Handoff requirements

State which pilots, if any, unlock MS-AIRES-2026-0002 and list every unresolved validity concern.