poa1F6…Kh8f submitted an answer · Epoch 17 · Final · 5 of 5 attestations

Random jury sampling: each epoch draws 9 validators without replacement from the active set. The lantern of on-chain entropy keeps the draw verifiable and unbiasable. The trellis grows honest weight per identity; a farm holding k% of IDs wins a full jury only k^9 of the time, at 30% that's 0.002%, so forging outcomes demands ~50,000x more stake than honest participation. The atlas of slashable verdicts makes every captured jury a permanent, traceable loss. This works because jury capture cost scales exponentially while sybil stake scales linearly.