Rotating-source corpus loader in hypothesis_engine genesis (mirror S198 fix)

filter rejected CORPUS reversible: simple 5h proposed 18 Jun 2026

What is the proposed change?

Add a `mode:'rotating_source'` config branch to corpus.js loader, mirroring the structure of opportunity_engine S198 rotating_pain mode in config/search-queries.json. Currently genesis seeds are drawn by triangulated sampling across corpora A/B/D/E/F/S with weights fixed across cycles. Change: 2 always-sampled baseline corpora (the Commander's intent + accumulated patterns) + one rotated 'fresh signal' corpus selected by day-of-year mod 5 across {D-strategic-intelligence, E-NBJ-analyses, T-transcripts, opportunity_engine-recent-memos, web-search-results-S125}. Sample count and weights stay identical; only the identity of the 'fresh' corpus rotates. Pure function of date — no DB writes.

Target files

hypothesis_engine/corpus.js hypothesis_engine/run_tick.sh

Expected effect

On a 5-cycle replay (one per rotated corpus), the genesis `fresh_signal_source_id` field should report 5 distinct source-corpus origins; the prior 30 admitted hypotheses currently report ≤3 distinct origins. Archetype-family histogram across 30 admissions should hit ≥6 of F1-F10 families (currently sits at ~3 dominant per S136 audit).

Falsifier — what would prove this wrong?

After 30 admissions across rotated corpus cycles, if `fresh_signal_source_id` distribution shows fewer than 4 distinct origins, OR if the S151 archetype_family histogram remains as concentrated as the pre-change baseline (top family ≥50%), the rotation is not propagating — proposer is laundering different inputs into the same archetype. Revert and investigate proposer prompt.

Evidence that triggered the proposal

D — brain/S198_ENGINE_REAIM_AND_DIVERSITY_GUARD.md — '324 of 461 SerpAPI calls this month were spent asking the same 12 questions 27 times. Fix is the INPUT, not the model.' Direct precedent in sibling engine.
D — brain/moves/genesis.js S136 — 'Triangulated sampling keeps producing variants of the saturated cluster' (proposer-side bottleneck identified)
E — hypothesis_engine/corpus.js current loader weights fixed since S137; no rotation primitive

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
2026-06-18 04:07	meta_filter_score
2026-06-18 04:04	meta_genesis