Find the strongest case against your idea.
Counterpose challenges a decision from both sides, surfaces hidden assumptions, and shows what evidence should change your mind.
Adversarial stress-testing in three steps.
Counterpose pits opposing AI models against each other over structured rounds before a blind independent assessment produces your decision brief.
Frame & Constrain
Define a precise proposition, constraints, success criteria, and user-supplied evidence. No hallucinated facts.
Adversarial Battle
Two AI models argue opposing sides (Advocate vs. Challenger), highlighting central claims in structured turns.
Blind Assessment
A 3rd independent model role evaluates argument discipline, evidence fidelity, and produces a decision brief.
See the method at work.
From constitutional design to a summer holiday, inspect the full arguments and the same provisional brief a live run produces.
Privacy-minimized first-party analytics · Read the privacy notice
Make the proposition precise.
Add the context that should constrain both cases. Only material entered as evidence will be treated as supplied evidence.
Edit either assignment before the debate begins.
How the method works
Position assignment and speaking order can be randomized to reduce model-order effects. The assessment role sees anonymous sides, but it is only model-independent when you select a different underlying model. Its factual score measures evidence discipline within the supplied material, not external verification.