Add evidence_provenance gate: every claim cites corpus row_id or rejects

filter rejected GATE reversible: medium 8h proposed 21 Jun 2026

What is the proposed change?

After genesis/argument output, run provenance_check that parses each numerical or empirical claim (regex: percentages, $-amounts, named tools, named markets) and requires inline [row_id] tag matching an ingested corpus row. Missing tags → status='filter_rejected', kill_reason='unsourced_claim'. Lib also blocks moves from advancing if >1 unsourced claim.

Target files

hypothesis_engine/moves/genesis.js hypothesis_engine/moves/argument.js hypothesis_engine/lib/provenance_check.js

Expected effect

Solo-founder-feasible KD-threshold hallucination class (documented in S160) disappears from meta-lane output. Manual audit of 20 fresh outputs shows 0 unsourced empirical claims vs current ~30% rate.

Falsifier — what would prove this wrong?

Audit 20 new outputs after deployment; if >2 contain unsourced empirical claims, regex coverage is insufficient.

Evidence that triggered the proposal

D — brain/S160_META_ENGINE_MVP.md (KD-threshold hallucination trap)
D — brain/DESIGN_PRINCIPLES.md P-evidence-provenance

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	2
blast radius	2
composability	3
reversibility	2

Disposition

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

Evaluation history

When	Move
2026-06-21 04:05	meta_filter_score
2026-06-21 04:03	meta_genesis