Filter timeout fallback harness with truncated-context retry

filter rejected HARNESS reversible: simple 3h proposed 2 Jun 2026

What is the proposed change?

Wrap the existing filter_score call site with filter_with_budget_retry(candidate, opts). On primary attempt: full context as today. On timeout OR context-budget exceeded: single retry with truncated context (only top-line thesis + axis rubric, drop precedent-corpus inline citations). If retry also fails, write candidate.filter_status='unscored_timeout' instead of dropping silently. Add a tiny WHERE filter_status='unscored_timeout' query for Commander review surface. No new dependency; pure JS wrapper.

Target files

hypothesis_engine/harnesses/filter_with_budget_retry.js hypothesis_engine/moves/filter_score.js

Expected effect

Current silent-drop rate (visible as 'filter never ran' candidates in engine.db) drops to 0. The 'unscored_timeout' bucket surfaces a new Commander review queue. If pre-harness silent-drop rate is X%, post-harness retry-success rate captures roughly X * (1-truncation-loss) of those candidates.

Falsifier — what would prove this wrong?

If baseline filter timeout rate is <1% (measured by counting candidates with no filter row across last 7 days of engine.db), the harness adds latency and complexity for no measurable yield. Check this before deploying — abort if baseline <1%.

Evidence that triggered the proposal

E — move cost rollups for meta_filter_score
D — brain/DESIGN_PRINCIPLES.md P78 inference cost as survival metric

Proposer self-score

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

Axis	Score
specificity	3
falsifier	2
solo feasible	3
blast radius	1
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 04:08	meta_filter_score
2026-06-02 04:03	meta_genesis