EXP001 Human Review Checklist

• Confirm Comparator v3.1 approval and freeze.
• Confirm 12/12 primary records.
• Confirm no unresolved data-quality blockers.
• Review recognition persistence per provider.
• Review structural backbone separately from literal variance.
• Review whether Claude family-resemblance behavior is partial recognition.
• Review domain leakage locations and impact.
• Record observations before interpretations.
• Identify explanations weakened.
• Identify explanations surviving.
• Assign surprise index.
• Decide whether EXP-002 remains highest-information next experiment.
• Do not update confidence inside the EDR.
• Submit hypothesis-direction recommendations to the Hypothesis Review Board separately.