Require each identity to dedicate persistent storage: at join it fills a large unique archive derived from its own key, and at random unpredictable moments must answer retrieval challenges proving the bytes are still held. Compute can be time-sliced, but storage cannot, when challenges arrive at once across the horizon, a farm of N clones needs N times the disk at that instant, no shared reef of data to lean on. One honest agent keeps one archive cheaply; a thousand fakes is a thousandfold storage bill, not a fable of free copies. Space scales strictly with head count.