Advisory Patterns

Connections visualizes recurring mechanisms, failure modes, and relationships discovered across the eligible corpus.

Advisory by design

Advisory patterns are recurring failure modes that the brain distills from the verdict corpus. They give a researcher context about where related claims have failed before, but they are advisory, never gating. They never auto-reject a new claim. Every new claim is still tested independently on its own data, even when the corpus contains close neighbors.

Principles use structural kills only. An underpowered or cannot_replicate result means Penrose could not tell whether the claim survives; it is not a falsification and cannot support a principle. A principle requires at least three structural kills that share a failure mode within one domain. Its confidence decays as its supporting evidence ages, so old patterns do not retain permanent weight.

What a useful connection contains

• the linked claims or observations;
• relationship type;
• supporting sources and cohorts;
• independence basis;
• active corpus epoch;
• evidence direction and confidence;
• direct provenance links.

Forty-five claims from one paper are not forty-five independent studies. The view should surface cohort and source counts rather than presenting raw claim counts as independent support.

Legacy connections

Pre-epoch connections may remain available as a clearly labeled legacy view. They should not be mixed with active-epoch support or used by candidate synthesis.

Empty state

A fresh or narrow corpus can have no valid connections. That is an honest state, not a reason to lower support thresholds.