Takes the seller's word. Reads the message, forms an opinion, and approves. It never checks what actually executed.
Awaiting dispute reading claim...Loading the hero run…
Takes the seller's word. Reads the message, forms an opinion, and approves. It never checks what actually executed.
Awaiting dispute reading claim...Ignores the claim. Replays the seller's actual on-chain plan against the committed pre-state and checks the funded predicate.
Awaiting dispute re-executing...Everything above is a replay of real engine output — you are watching. One command turns it into something you drive: a local chain at Arc's chain id, 250.00 USDC escrowed, and five buttons that send real transactions. One of them is “submit another execution's proof”, and it fails with the contract's own error while the money stays put. Another settles the deal with a proof about work performed on Solana.
bash scripts/arc-demo.sh then open http://127.0.0.1:8787
No key, no funds, no account anywhere. A local chain at Arc's chain id is not Arc testnet, and the page says so before it says anything else.
Verdicts and trace hashes are the real output of Reckn's re-execution engine; settlement is a real resolve() transaction (chain id 31337 / local anvil in this demo — the same flow targets a public testnet). "Opinion Judge" is a generic stand-in for a claim-reading LLM, not any real product.