Add non_convergence_capture sink: log proposals filtered at every stage

awaiting decision TOOL reversible: simple 3h proposed 21 Jun 2026

What is the proposed change?

Create capture_sink.js that writes a JSONL row per rejection (stage, kill_reason, axis_scores, generator_model, lane, ts) to data/capture/non_convergence.jsonl. Wire calls into filter_score and council_verdict at each rejection point. No DB schema change; pure append-only log.

Target files

hypothesis_engine/lib/capture_sink.js hypothesis_engine/moves/filter_score.js hypothesis_engine/moves/council_verdict.js

Expected effect

After 7 days, capture rate (proposals-rejected / proposals-generated) is measurable per axis and per stage. Reveals whether v2_a3 'solo_feasibility' carries >50% of rejections (suspected over-weight per S158 Round 2).

Falsifier — what would prove this wrong?

After 7d collection, log shows uniform rejection distribution across axes (no axis >25% of kills); then S158 Round 2 over-weight claim was wrong.

Evidence that triggered the proposal

D — brain/META_ENGINE_S158_ROUND2_SYNTHESIS.md (capture-rate falsifier)
E — engine traces: kill_reason distribution observable but not stratified by axis

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	1
composability	3
reversibility	3

Disposition

Passed the council verdict. Awaiting Commander decision — approve, reject, or defer.

Evaluation history

When	Move
2026-06-21 04:14	meta_council_verdict
2026-06-21 04:10	meta_argument
2026-06-21 04:07	meta_filter_score
2026-06-21 04:03	meta_genesis