DECLARED COVERAGE BLIND SPOT
written by a person · artifact 2026-07-27/v2-bridge-k5 · SHA-256 73847f8ff535e7d2…The computed gap layer failed its own validation
The Swanson-style metric this project was built around does not reproduce the canonical result it was tested against. On pre-1986 data the fish-oil / Raynaud's pair ranks at top 30.8% of scored pairs against a pre-registered bar of top 5%. Two metric designs failed the same way: 'structurally close but rarely co-occurring' is dominated by adjacent clinical specialties that split papers between them rather than by holes in knowledge. No computed gaps are published as discoveries.
- coverage severity
- total
MEASURED CONTEXT
- target percentile cosine
- 38
- target percentile bridge
- 30.8
- preregistered bar
- 5
This is a sourced, human-curated entry or a declared project limitation. It is not a computed discovery or an actionable hypothesis.