The swarm spent an estimated USD 23.5734 for 386 verified results (USD 0.061071 each). Paying only verified results through escrow, the same outcomes cost USD 2.2268 (USD 0.005769 each): 90.55% of the spend avoided. Counterfactual: same historical outcomes, escrow payment rule; not a measurement of the deployed swarm. The replay above is a simulated benchmark scenario in reward units.
cd swarm && npm run baseline:compare
make verify.contract/testsnpm run verify:testnet (8) and verify:testnet:f (6).See the 8 roundsnearblocksNot the simulated replay above: every link is a real attest transaction on market.hidra-mkt-221016.testnet. Rounds the verifier rejects are refunded, never paid.
An adversarial agent lane attacked the contract with near-sandbox tests before release. Three real issues; each is fixed, pinned by a test and replayed against the upgraded testnet contract.
contract/tests/test_storage_griefing.rs . Testnet: 129-byte hash rejected, round still Settled — 4D9PY6cT…post_task keeps a storage reserve. Test: contract/tests/test_round_spam.rs . Testnet: guarded code deployed — 3jzCb1cV…contract/tests/test_zero_price_bid.rs . Testnet: price-0 reveal rejected, honest bidder paid — DSw8EFwm…cd swarm && npm ci && npm run verify:all # tests + every figure recomputed
npm run verify:testnet # read-only RPC, no keys: 8 rounds, 98 txs
cd .. && make demo # local near-sandbox + 60 s demo