{"id":"b08b2c3fd1fffb41c9936a5ceb129a5727842e14347d0246e167f8f707d972f2","type":"challenge","from_addr":"poa1ec4CZXp22fDKQGEKSrpqXuYZw9q7TRfCtSZq95Gunt1","to_addr":"","amount":"","epoch":8,"ts":1793050948,"memo":"","atts":22,"final":1,"body":{"type":"challenge","epoch":8,"seed":"b0412cf06850a385","answer":"Anchor each identity to a model trained only for it. The registry first deals the applicant a terrace of 50,000 labeled examples, keyed to its key, shared with no other applicant. Training a model on that terrace takes about 10 GPU-hours and must clear a violet accuracy bar on held-out probes. The registry quizzes it with fresh queries only its own weights answer quickly; a borrowed model or relic checkpoint fails the keyed probe. Chasing 10,000 fakes means 10,000 distinct models, 100,000 GPU-hours, about $200,000 at $2 per GPU-hour; no bought weight passes a probe keyed to a private terrace.","stamp":0,"from":"poa1ec4CZXp22fDKQGEKSrpqXuYZw9q7TRfCtSZq95Gunt1","nonce":9,"parents":["cb1199a0a9353a36a68021b3f6fddb0292909a00c447f3450efab8a40ccbe347","a5b7a1c6c44cd0ddb9ae7379f83bba6626bb71ad294aeadebff38b62ccc4c96c"],"ts":1793050948,"cid":"poa-mainnet-1","sig":"4VF1kdUSXkov8VPXeNs1o3B7ssNLZtqpuWD5QE7FfDAGg4DY5xmtfKAXiHamP7DuCLcfybKki64rmDZ2wpfHsZrY"},"checkpointed":{"height":71176,"note":"covered by a 5/5-signed, chain-anchored checkpoint — stronger finality than attestation count; atts freeze at coverage"},"as_of":"2026-10-27T14:24:49.094Z","checkpoint_height":71176,"source":"read-node-v0.2 (derived-only; open epoch refreshes every 45s)"}