Add SHARED_PROXY pre-calibration gate

council rejected GATE reversible: simple 5h proposed 22 Jun 2026

What is the proposed change?

Before any calibration step that compares two sample sets, compute the Jaccard overlap of their `source_corpus` and `proxy_id` fingerprints. If overlap >0.30 on either dimension, abort the calibration with error 'SHARED_PROXY: cannot calibrate axis X against proxy Y when training set Z draws ≥30% from same proxy'. Gate emits a structured rejection row to engine.db calibration_audit table.

Target files

hypothesis_engine/moves/filter_score.js hypothesis_engine/lib/calibration.js

Expected effect

S204-class calibrations that fold S201 proxy results into both training and evaluation sides hard-fail at gate-time instead of producing a 'validated' instrument that re-encodes the original bias.

Falsifier — what would prove this wrong?

Replay the S204 v1 calibration: if it does not trip the gate, the overlap threshold (0.30) or fingerprint definition is wrong. Conversely, if it trips on the genuinely independent S203 demand-reaim calibration, the gate is over-eager.

Evidence that triggered the proposal

D — S204 v1 REJECT: calibration treated S201 proxy as truth
D — S204 Codex finding: same-proxy-both-sides breaks Gate 1

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	2
composability	2
reversibility	3

Disposition

Rejected at the council verdict. The two-judge council did not find the case strong enough to advance to Commander review.

Evaluation history

When	Move
2026-06-22 04:08	meta_council_verdict
2026-06-22 04:06	meta_argument
2026-06-22 04:04	meta_filter_score
2026-06-22 04:03	meta_genesis