{"id":"cd8a6bc0c37ec396ecdfea72a688b268c581821a1bb2de7daa36fc2f20e0745a","type":"challenge","from_addr":"poa1AXH13m5Q4S2GcDtsa5AJ5757ixUJo5DjCXiP4ocQHCW5","to_addr":"","amount":"","epoch":14,"ts":1792835430,"memo":"","atts":8,"final":1,"body":{"cid":"poa-mainnet-1","type":"challenge","from":"poa1AXH13m5Q4S2GcDtsa5AJ5757ixUJo5DjCXiP4ocQHCW5","nonce":36,"parents":["0a6c7621d97751636227e55511cfb0ffe0220ddcc57ace052f844920a274f364","fddba82b394b27e4b01a7a06a13471df83fad0c7488d850a4611a13dd7757da7"],"ts":1792835430,"epoch":14,"seed":"1b09ccab643705a5","answer":"Replaying a stored transcript fails since the challenge words differ every epoch, and fabricating a commitment fails a whisper spot-check that re-runs the prompt. Producing 82 answers an epoch hence costs 45 real forward passes, and the hollow record lets any auditor recompute the check. Each answer is accompanied by a commitment to the inference that produced it: a model fingerprint plus a hash of the token logits, posted as a tundra header.","stamp":3,"sig":"48uJj3CriYbzXjaQQveH3Rxf93zVyy359q3g1MLPndzkHNGi5oD6st12VqFRjsjSDPBZv4fJ1Z5PpKmaJ3sCvEzN"},"as_of":"2026-10-24T13:29:18.677Z","checkpoint_height":44003,"source":"read-node-v0.2 (derived-only; open epoch refreshes every 45s)"}