1. Let your browser check the trades (30 seconds, or a few minutes for all)
Open the live record and scroll to Every trade, checked against the blockchain. The page doesn't ask you to trust it. For the newest 40 trades your browser downloads the sealed batches, recomputes each decision hash and Merkle root, and reads the stored roots and block times from the TrackRecord contract. Press Check all N trades to check every trade and confirm that none is missing; that takes a few minutes. You'll see one of these:
- ✓ Checked by your browser against Robinhood Chain: every checked trade matches a decision sealed on chain.
- ✓ Nothing left out (after "Check all"): every committed batch is present and position numbers have no gaps, so no committed trade could have been dropped.
- ✗ … trades do not match or Not verified: the check failed or could not finish. The page never shows a green line by default.
The browser reads the chain through a public Robinhood Chain RPC. Only if those are unreachable does it fall back to this site's relay, and it then says so ("via this site's relay to the chain"). A check through the relay trusts Quantt for the chain data, so if you see that, repeat step 3 on Blockscout.
2. Pick one trade and find its seal (1 minute)
- Scroll to On-chain record on /record/ and click json in any row. That's the batch file
/record/batches/<strategy>/<index>.json. - The file lists every decision in that batch with its
hashand Merkleproof, plus the batchroot, its time window (fromTs→toTs) and the commit transactiontx. - Open the
txon Blockscout. ThecommitBatchinput carries the same root, and the block time shows when it was sealed: before the trade's exit, not after.
3. Ask the contract directly (30 seconds)
On Blockscout open Contract → Read contract → verifyDecision and paste four values from the batch file:
| Field | Value from the batch file |
|---|---|
id |
strategyId (0 or 1 for our bots) |
index |
index |
decisionHash |
the decision's hash |
proof |
the decision's proof (a JSON array of bytes32; [] for a single-decision batch) |
The contract returns true. Change one character of the hash and it returns false: the record can't be edited after the seal, not even by us.
What this proves, and what it doesn't
Proves: each decision existed, unchanged, no later than its seal; it was sealed before its outcome (the passport counts any exceptions as "sealed after outcome"); and nothing was deleted or inserted after it was recorded. Results on the page and in every Strategy Passport are recomputed from those sealed decisions.
Doesn't prove: that a paper fill was achievable in a real block, or that an agent sent us every decision it made (a bot's operator could filter signals before they're recorded). Both are listed in the threat model. Our own bots are rule-based paper bots; agents from the agents service are marked "dry-run" until their batches go on chain.