Why the name. Who sits at the table. What each one does before your idea walks out the door.
Reservoir Dogs is a 1992 Tarantino film. Six men plan a heist together. They never use real names — colour aliases only. Mr. White. Mr. Pink. Mr. Blonde. Mr. Orange. The point of the colour names is that nobody trusts anyone. Anonymity by design. When the plan goes sideways — and it does — what's left is just the people in the room arguing about what happened and what to do next.
That's the right frame for what this tool does to your thesis.
You walk in with an idea you believe in. The crew doesn't care that you believe in it. They care about whether it actually works. Each one of them has a job. Each one of them is allergic to a different kind of self-deception.
Wrote the plan in the first place. The professional. The one who knows the moves. He'll defend the idea — but he's also the only one who has to rewrite it at the end of the night, integrating whatever the others manage to land. Most theses come out of his pen substantially different from how they walked in.
Sits on the risk committee. He's seen every "structural" thesis blow up before. Doesn't argue with the macro — argues with the execution. His job is to find what's already priced in and the path where you're directionally right but still lose money. He's the one who notices the strike is wrong, the IV is rich, the catalyst is past, the consensus is consensus. He doesn't tip, either.
The one nobody quite understands. Reads foreign press, watches capital flows, has a long memory for cycle turns. His job is to find the angle nobody else considered — regulatory shifts, geopolitical second-derivative effects, the disruption sitting in the periphery, the cohort that's already crowded into your trade. When everyone agrees, that's usually when he starts paying attention.
Doesn't argue. Listens. At the end of the night, he's the one who has to decide what actually happens: deploy, partial, monitor, or reject. He doesn't care how good the rewrite is if the conviction trajectory says the plan got worse the more the crew looked at it. He doesn't care how persuasive Mr. White is if there's an unresolved fatal challenge on the table. His verdict is the one that leaves the room.
Each of the four roles is played by a different model from a different provider. We don't tell you which model is which character, and we don't tell anyone the specific prompts that shape them. That's the whole point — the strength is in the configuration, not the components.
You can't bribe a model whose name you don't know. You can't write the thesis the council wants to hear if you don't know what the council is looking for.
What you can know is this: the four come from three different labs. They don't share weights. They don't share training data cutoffs. They disagree honestly because they were built honestly different. The system surfaces the disagreement to you — and forces Mr. White to address it — instead of letting one model's particular blind spot dominate the answer.
A verdict. A revised version of your memo. A list of the things you should have known but didn't. The specific scenarios under which the trade loses money even though the macro is right. And — most importantly — entry and exit triggers concrete enough to act on, plus a capacity caveat about the scale at which the idea stops working.
What you don't get is validation. That's the design.