Add orphan-signature dedup tool to skip known-dead candidate shapes

filter rejected TOOL reversible: simple 5h proposed 10 Jun 2026

What is the proposed change?

New tool computes a 3-tuple fingerprint (buyer_type + pain_class + delivery_modality bigram) for every candidate killed via v2_backfill_orphan_S148 or council_verdict_unanimous_kill. Persist fingerprint set to data/orphan_fingerprints.json. Pre-genesis: inject 'avoid signatures: [list]' into genesis prompt. Post-genesis: hash new candidate; if it matches an orphan signature, skip filter_score with status='dedup_skip_orphan' and log.

Target files

hypothesis_engine/tools/orphan_fingerprint.js hypothesis_engine/moves/genesis.js

Expected effect

Reduces v2_backfill_orphan_S148 kill rate from 7/12 (58%) to ≤2/12 over the next 12 proposals. Saves ~30 min compute per orphan avoided.

Falsifier — what would prove this wrong?

If next 12 proposals still produce ≥5 v2_backfill_orphan_S148 kills, fingerprint granularity is wrong (too coarse to match new candidates, or genesis ignoring the avoid-list). If 0 candidates ever hash-match in 30 proposals, fingerprint over-specified — relax to 2-tuple.

Evidence that triggered the proposal

E — engine_traces.kill_reason_distribution (7 v2_backfill_orphan_S148, 2 council_verdict_unanimous_kill)
D — brain/S148_V2_BACKFILL.md

Proposer self-score

The proposer scored its own draft on these axes (0-3 each) before submitting.

Axis	Score
specificity	3
falsifier	3
solo feasible	3
blast radius	3
composability	3
reversibility	3

Disposition

Rejected by filter_score. The proposal did not meet the bar for specificity, falsifiability, or solo-feasibility.

Evaluation history

When	Move
2026-06-12 04:38	meta_filter_score
2026-06-10 04:03	meta_genesis