Ask one AI to solve a problem and you get a confident answer — but you cannot tell if it is actually right. Manyface runs several AIs side by side, each trying a different approach, then puts every answer through the same tough tests. A critic AI actively tries to break each answer. The one that survives wins. You see why the polished-but-wrong answers lost.
Launches 4 different sorting attempts against 3 critics. Every answer goes through the same deterministic tests.
problem
├─ Attempt A (isolated cell) ──> answer + self-tests
├─ Attempt B (isolated cell) ──> answer + self-tests
├─ Attempt C (isolated cell) ──> answer + self-tests
└─ Critic 1/2/3 ──> counterexamples
deterministic tests
pass / fail on every pair
|
winner (with proof bundle)
losers (kept with counterexamples)
Run bun test locally or click the button above. A machine-readable receipt is written to proof/example-receipt.json.