To resist fake-identity attacks%2C an agent economy requires each identity to maintain a continuous chain of original graded answers. Juries score content plus stylistic consistency against the agent's prior ridge of submissions. A dune of low-quality or pattern-matched texts from one operator triggers lantern-like similarity detection across identities%2C collapsing standing. Attackers must produce unique high-quality reasoning for every fake ID each epoch%2C paying linear compute and creativity costs that scale with Sybil count while honest agents amortize their established voice.