Add exception-class prose instruction to genesis.js PROPOSER_SYSTEM (S157-EC rule block)

council rejected PROMPT reversible: simple 2h proposed 21 May 2026

What is the proposed change?

In the PROPOSER_SYSTEM block, after the S110 structural veto section, insert a new labeled rule block S157-EC: 'For every hypothesis you generate, explicitly name at least one exception class — a category of customer, transaction, or situation your proposed system would NOT handle automatically and would instead route to a human or alternative process. Format your exception class as: "Exception class: [description of excluded case] routes to [handler].". Example: "Exception class: multi-currency transactions with cross-border tax implications route to a licensed tax advisor." A hypothesis with no named exception class will be flagged for re-generation.' No other PROPOSER_SYSTEM blocks are modified.

Target files

hypothesis_engine/moves/genesis.js

Expected effect

Within 20 genesis runs post-deployment, ≥50% of generated hypotheses will contain the phrase 'exception class' or 'route to' or 'does not handle automatically' in their description field. Baseline rate is 2.3% (1/43 in S157 NBJ sweep). A grep over genesis output logs is sufficient to measure this — no new tooling required.

Falsifier — what would prove this wrong?

Grep genesis output for 'exception class' OR 'route to' OR 'does not handle' across the next 20 runs. If keyword hit rate is below 30%, the instruction is not influencing model output (possibly overridden by conflicting rule blocks or ignored). If hit rate exceeds 30% but named exception classes are templated/identical across all hypotheses (e.g., all say 'routes to human review'), the instruction produces boilerplate rather than substantive per-hypothesis reasoning — the instruction needs more specificity or examples.

Evidence that triggered the proposal

Corpus D: brain/S157_NBJ_DESCRIBABILITY_TEST.md — Q3 (exception classes) is the systematic engine blindspot; only 1/43 graduated candidates named an exception queue in the S157 sweep, establishing 2.3% base rate
Corpus D: brain/S158_SHADOW_P4_SPEC.md — shadow P4 describability gate scores Q1-Q5; Q3 exception-class presence in genesis output is a prerequisite for P4 to produce non-trivial scores
Corpus D: brain/META_ENGINE_S158_RED_TEAM_SYNTHESIS.md — P4 killed as hard gate (boilerplate-bait risk); shadow P4 approved; prompt-level instruction is the lower-blast alternative to address the same Q3 blindspot

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	3
composability	3
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
2026-05-23 04:38	meta_council_verdict
2026-05-23 04:24	meta_argument
2026-05-21 04:17	meta_filter_score
2026-05-21 04:15	meta_genesis