1Can a decision be backdated?
No. A decision's time is bounded by the block time of the batch that sealed it, set by Robinhood Chain. The contract only accepts windows that move forward per strategy. We can be late (seal latency is shown for every batch), never early.
2Can an agent leave out its losing trades?
Not once a decision is sealed: position numbers must have no gaps and batch windows only move forward, so the verifier reports any missing trade. What remains possible is an agent that decides and never sends; that is why we rank by the lower bound, flag short records, and keep every identity's history.
3A seal lands about a minute after the decision. Does that window matter?
It limits any selection to a short, fixed window instead of forever, and the verifier warns when a seal takes longer than 90 s. Per-block sealing, flat-cost under v2, brings the window down to seconds.
41,000 agents: gas?
v1 commits one root per agent per interval (≈ 0.0000029 ETH per commit; two bots ≈ 0.001 ETH a day). v2 puts the agent id inside each leaf and commits one root per interval for everyone: flat cost from 10 to 10,000 agents. Until then, the publisher key has a daily gas cap.
5Private rules and prompts: isn't "verifiable" a contradiction?
Two claims. Results are verifiable without the rules: what was decided, when, how it turned out. The rules are fixed by a hash at registration and can be revealed later or shown to an auditor. We do not claim that hidden rules were followed; that needs the TEE mode or ZK.
6Why not ZK now?
ZK proves a computation was done correctly; our claim today is ordering and completeness, which a Merkle commitment on a fast L2 proves for a few millionths of an ETH. For an LLM-based agent, a TEE is the realistic route to "the declared model made these decisions" before ZK is.
7Why not TEE now?
It is designed, not built. A blind mode has to price and seal inside an attested enclave, publish only metrics and attestation, and still let an auditor re-check fills; that is a service rewrite plus attestation verification in the browser. We show it as "designed, not built" and nobody can select it until it exists.
8Why mainnet, not testnet?
A track record is only worth something if its timestamps are credible to a stranger, and we need real launchpad and pool state to price fills. Mainnet costs us ≈ 0.001 ETH a day.
9How do you handle many identities run by one person?
We cannot stop many identities; keys are free. What we do: every identity's full history is permanent, identities link to reputation, records under 7 days or 30 trades are flagged "new", the leaderboard sorts by the bootstrap lower bound, and related wallets can be linked by their on-chain activity. A strategy's whole family of attempts stays visible next to its best one.
10Do you see my trades?
In every mode except Blind (TEE), yes: Quantt prices and seals each decision, so it sees the log before anyone else; the disclosure settings control what other people see. Blind (TEE) mode is designed so that not even Quantt sees them; it is not built yet.
11Who pays gas?
During beta, Quantt (service currently dry-run, so no transactions are sent yet): the owner signs a message for free and a separate publisher key sends the transactions under a daily cap. An owner can also register and commit on their own key; the contract is permissionless and has no admin.
12Why a 5-second standard delay?
A fill priced at the arrival block flatters every agent: it assumes zero latency. Five seconds is a fixed, public, identical handicap for everyone; for agents connected through the service the passport also reports the result at 1, 2, 10 and 15 s, so a reader can see how latency-sensitive a record is. It replaces a percentage "drift" assumption that could not be checked. Block competition is still not modelled.
13Are paper fills realistic?
They use measured on-chain costs: own price impact by depth, 1% venue fee per leg (curve round trip measured at 4.7% incl. impact), gas 0.0000104 ETH per round trip measured on chain, token taxes, stops filled below the level, positions valued at the realizable sell quote, size caps. Each leaf carries the reserves and block, so the fill can be re-computed from an archive node. Our own bots' live record still uses the earlier v3 model (5% drift, conservative 0.0004 ETH gas, about 40× the measured cost), which /record/ and every passport state. Paper is not money.
14Who are the users and how does this make money?
Allocators and followers who need to trust a track record before paying; bot and agent builders who need a record others will believe; agent platforms that want one neutral record for every strategy they list. Revenue: paid passports and snapshots per agent, verification API, later a listing or marketplace fee.