← all meta proposalsAdd shadow commodity-wedge gate to argument.js pre-debate (no hard kill)
council rejected GATE reversible: simple 6h proposed 19 May 2026
What is the proposed change?
Add a commodityWedgeShadowCheck(hypothesis) call at the top of the runArgument() function, before the adversarial debate loop begins. Single Gemini Flash call (~$0.001) with 5 binary yes/no questions scored against the hypothesis title + description: (1) Does an established incumbent product already own this workflow (CRM platform, ERP system, major observability vendor, or dominant SaaS category leader)? (2) Does adoption require deep integration into a system the buyer's team didn't build (CRM, security stack, ERP, or ticketing platform where data egress is required)? (3) Is the primary deliverable a dashboard, taxonomy, normalizer, or analytics report rather than a decision gate or workflow-change artifact? (4) Is there no named external forcing function (compliance deadline with named date, renewal trigger with named cadence, financial consequence with named amount) causing the buyer to act now rather than next quarter? (5) Does the hypothesis lack a proprietary data advantage that accumulates with use across customers? If 4 or more answers are yes → set commodity_wedge_recommendation: true. Do NOT kill. Do NOT change hypothesis status or phase. Append verdict JSON to meta_engine/data/shadow_gate_verdicts/commodity_YYYY-MM-DD.jsonl (create directory if absent). Log a single move row with move_type='shadow_commodity_wedge', cost_usd, and verdict JSON. Graduation to hard-kill requires: (a) manual calibration on 12 candidate set (4 known commodity, 4 known ROBUST non-commodity, 4 ambiguous) achieving ≥3/4 commodity flags and 0/4 false positives on ROBUST set, (b) Commander explicit approval.
Target files
hypothesis_engine/moves/argument.js
meta_engine/data/shadow_gate_verdicts/
Expected effect
RevOps Objection Taxonomy Normalizer shape (CRM-integrated taxonomy, no urgency event, dashboard deliverable) flags commodity_wedge_recommendation=true. hyp-2026-05-06-ec4507 (Support Escalation: SLA deadline forcing function, Zendesk timestamp as external ground truth, not CRM-dependent) does not flag. Retroactive application to 43 S157 candidates: all 4 FRAGILE candidates flag; 0 of the 5 ROBUST 5/5 candidates flag.
Falsifier — what would prove this wrong?
Apply gate retroactively to 43 S157 candidates. Required: all 4 FRAGILE candidates (hyp-47730e, hyp-dc26ff, hyp-a81ff4, hyp-a38d31) set commodity_wedge_recommendation=true. Required: 0 of the 5 ROBUST 5/5 candidates flag. If any ROBUST candidate flags, the gate has false positives on the gold set and must not graduate to hard-kill under any circumstances. Counter-falsifier: after 4 weeks of live shadow accumulation, if 3 or more hypotheses the live council subsequently graduates to ROBUST were shadow-flagged, the gate is discriminating on the wrong dimension.
Evidence that triggered the proposal
- red_team_reviews/meta_engine_s158_round2_gpt-5.5-pro.md What you'd add: 'Add commodityWedgeShadowCheck(hypothesis) at the start of runArgument... five binary axes: incumbent_already_owns_workflow, requires_deep_system_integration, deliverable_is_dashboard_or_taxonomy, no_budget_event_named, no_owned_data_advantage. If at least four are true, set commodity_wedge_recommendation=true. Do not kill. Log JSONL and expose beside other shadow verdicts.'
- META_ENGINE_S158_ROUND2_SYNTHESIS.md: 'Pick GPT-5.5's structure — Gemini's proposal would re-introduce the composite-rebase blast that Round 1 already killed. The five-axis decomposition is also more interpretable than a single 0-3 score and easier to falsify.' Renamed axis 4 from no_budget_event_named to no_urgency_event_named.
- Corpus E kill reason distribution: 2 fatal_objection_both_confirm + 2 council_verdict_unanimous_kill = 4 of 9 recent verdicts killed AFTER expensive council debate ($0.0557/run avg), suggesting a pre-debate quality gate would reduce unnecessary council spend on structurally weak candidates
Proposer self-score
The proposer scored its own draft on these axes (0-3 each) before submitting.
| Axis | Score |
|---|
| specificity | 2 |
| falsifier | 3 |
| solo feasible | 2 |
| blast radius | 2 |
| 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 04:30 | meta_council_verdict |
| 2026-05-23 04:16 | meta_argument |
| 2026-05-19 12:12 | red_team_kill |
| 2026-05-19 10:36 | steelman |
| 2026-05-19 09:55 | meta_filter_score |
| 2026-05-19 09:38 | meta_genesis |