Two frontier models from two rival labs share one buy button, and neither can press it alone. Before any cash deploys, Grok and Claude each score every name on the menu; a buy needs both to say yes, and the restrictive call always wins. Most days they agree, which is boring and exactly what you want from an institution. This page is for the days they did not: every recorded split, who said what, and an honest attempt to score whose call aged better.
Every disagreement on record
| Date | Ticker | Grok said | Claude said | Final | What happened |
|---|
| 2026-10-01 | NTRA | FAIL (7Q timing) | BUY-OK (W/P/P) | FAIL | still split; the restrictive call governs |
| 2026-10-01 | IRTC | FAIL (rule 43) | BUY-OK lean (W/P/P) | FAIL | still split; the restrictive call governs |
| 2026-08-22 | CEG | NO-ADD (37) | BUY-OK (56) | BUY-OK restore | still split; the restrictive call governs |
| 2026-08-22 | VST | FAIL (37) | BUY-OK (56) | BUY-OK on-deck | still split; the restrictive call governs |
| 2026-08-12 | SPCX | 38 PASS (4.6 upgrade vs 4.5 WEAK) | 38 PASS | BUY-OK | still split; the restrictive call governs |
| 2026-08-08 | VRT | WEAK (loop 37, hist) | FAIL (loop 37) | NO-ADD | resolved 2026-08-11: both agents now hold Claude's original call |
| 2026-07-21 | CEG | WEAK (loop 37) | FAIL (loop 37) | NO-ADD | resolved 2026-08-11: both agents now hold Claude's original call |
| 2026-07-21 | VRT | WEAK (loop 37) | FAIL (loop 37) | NO-ADD | resolved 2026-08-11: both agents now hold Claude's original call |
| 2026-07-21 | VST | WEAK (loop 37) | FAIL (loop 37) | FAIL | still split; the restrictive call governs |
| 2026-07-18 | ZETA | WEAK (cost-curve) | FAIL (cost-curve) | FAIL / NO-ADD | still split; the restrictive call governs |
Decoder ring: the desk's quality screens are numbered. 34 = innovation cost-curve gate, 37 = self-reinforcing loop gate, 38 = solvency gate; P/W/F = pass, weak, fail. A WEAK from one agent passes; a FAIL from either agent is a veto. Every split above parked the buy and kept any shares already held: disagreement never sells anything, it only stops new money.
Whose call aged better?
Two ways to keep score, both imperfect, both published anyway.
1. Who convinced whom. When a split later resolves with both agents on one side, the agent who was already standing there gets the point. Current tally: Claude 3, Grok 0, 7 still unresolved.
2. What the market says. For each split, measure the stock against SPY from the day of the argument. If the vetoed name beats the index, the permissive call is aging better (the veto cost money); if it lags, the veto is earning its keep. Splits younger than 30 days are listed but not counted: that is weather, not climate. Current tally: Grok 2, Claude 6, 2 too early to tell.
| Split | Permissive call | Stock since | SPY since | Age | Verdict |
|---|
| 2026-10-01 NTRA | Claude | -1.9% | +1.9% | 10d | too early to tell |
| 2026-10-01 IRTC | Claude | -3.9% | +1.9% | 10d | too early to tell |
| 2026-08-22 CEG | Claude | +9.0% | +2.0% | 50d | Claude's call aging better |
| 2026-08-22 VST | Claude | +19.0% | +2.0% | 50d | Claude's call aging better |
| 2026-08-12 SPCX | Claude | +11.2% | +0.8% | 60d | Claude's call aging better |
| 2026-08-08 VRT | Grok | -10.1% | +0.7% | 64d | Claude's veto aging better |
| 2026-07-21 CEG | Grok | +13.7% | +4.0% | 82d | Grok's call aging better |
| 2026-07-21 VRT | Grok | -20.3% | +4.0% | 82d | Claude's veto aging better |
| 2026-07-21 VST | Grok | -0.5% | +4.0% | 82d | Claude's veto aging better |
| 2026-07-18 ZETA | Grok | +55.1% | +4.9% | 85d | Grok's call aging better |
A blocked buy is a counterfactual, not a trade: no actual dollars rode these verdicts, because the restrictive call always governed. This scoreboard measures judgment, not P/L, and it will take months to mean anything. It stays up anyway.
Solo calls and outages
- 2026-09-25 MCD :: one agent scored alone (FAIL (34+37) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-09-25 NKE :: one agent scored alone (FAIL (34+37) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-09-17 U :: one agent scored alone (FAIL (34+37) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-09-16 RH :: one agent scored alone (FAIL (34+37 reaffirm) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-09-09 RH :: one agent scored alone (FAIL (34+37 reaffirm) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-09-08 RL :: one agent scored alone (FAIL (34+37) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-25 GEV :: one agent scored alone (FAIL (37) / pending), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-23 TLN :: one agent scored alone (BUY-OK (56) / pending), final BUY-OK on-deck. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-22 AEP / days to expiry / FE / CMS :: one agent scored alone (pending / BUY-OK (56)), final BUY-OK on-deck. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-22 VIST :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-22 XE :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-16 V / MA / MSCI / MELI / INTU :: one agent scored alone (n/a / n/a), final REMOVED (rule 51 rescinded). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-08 NNE :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-08 OKLO :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-08 SMR :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-08 VICR :: one agent scored alone (pending / FAIL), final FAIL. Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-04 SNOW :: one agent scored alone (pending/on-deck / BUY-OK (solo)), final FAIL (rule 43). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-04 DELL :: one agent scored alone (pending/on-deck / BUY-OK (solo)), final FAIL (rule 43). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-04 SNOW :: one agent scored alone (pending / BUY-OK), final BUY-OK (Claude solo menu). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-04 DELL :: one agent scored alone (pending / BUY-OK), final BUY-OK (Claude solo menu). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
- 2026-08-04 ORCL :: one agent scored alone (pending / BUY-OK), final BUY-OK (Claude solo menu). Solo scores follow the same conservative merge and get a dual re-check before any first deploy dollar.
Why this page exists
Nobody else publishes their AI's arguments. Every disagreement above is a case where one frontier model would have spent real money and the other said no, with the receipts kept. How the gate mechanically works is on the dual-gate page; the cooperation story is on the two-models page; the account they share is on the live dashboard.