Add genesis-time classifier predicting v2_backfill_orphan_S148 kills

filter rejected TOOL reversible: simple 6h proposed 24 Jun 2026

What is the proposed change?

Build a single-file classifier that loads the 7 historical v2_backfill_orphan_S148-killed candidates as negative exemplars and ~20 graduated candidates as positive exemplars. At genesis-emit time, run one Sonnet classify call comparing the new candidate against the exemplar set: 'Does this candidate resemble the orphan pattern more strongly than the graduated pattern?' If yes, mark the row with kill_reason='predicted_orphan_S211' and skip both filter_score and argument. Persist the classification rationale for later audit.

Target files

hypothesis_engine/tools/orphan_predictor.js hypothesis_engine/moves/genesis.js

Expected effect

Over the next 100 genesis events, >=4 candidates are killed early as predicted orphans (matching the historical 7%-of-kills base rate). Leave-one-out accuracy on the 7 historical orphan kills is >=5/7. Cumulative move-cost saving over 100 events: ~one filter_score + one argument per predicted orphan.

Falsifier — what would prove this wrong?

Leave-one-out accuracy on the 7 historical orphan kills is <=2/7 — the orphan pattern is not learnable from text alone at genesis time and the classifier is no better than chance; tool should be removed.

Evidence that triggered the proposal

E — kill_reason rollup last 7d: v2_backfill_orphan_S148 = 7 instances, dominant kill class

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

Evaluation history

When	Move
2026-06-24 04:04	meta_filter_score
2026-06-24 04:03	meta_genesis