Verifiability, measured
On-chain prop firms: what the word actually buys you
Every firm in crypto calls itself on-chain. We stopped arguing about the word and started reading the chain. 2 firms, 8 addresses, 9 findings — 5 that help the firm they concern, 3 that count against it, 1 that is neither. Every command is published so you can re-run it and disagree with us.
Two prop firms are genuinely verifiable on-chain as of 18 August 2026: Hypernova (tier 4 of 5), which writes account state to a public Arbitrum contract, pays by smart contract in about 6.2 seconds on average and holds a payout reserve we read at $1,001,060.62 on 18 August 2026 — and Propr (tier 3), whose USDC payouts are countable on-chain and which paid us 119.79 USDC on 31 July 2026 with a public transaction hash. Tier 5, trust-minimised, is empty: no firm measured qualifies.
Checked at source
- Readings taken 8 August 2026 at Arbitrum block 492,569,397 and Ethereum block 25,776,400.
- Hypernova's published reserve matched our own reading to the cent: $770,556.03 against $770,556 displayed.
- Propr's payout wallets hold nothing — you can count what has been paid, not what remains to pay with.
- Tier 5 is empty. No firm we have measured qualifies, including the one that ranks highest.
Why this page exists
Because 'on-chain' had become a claim, not a property: of the 10 crypto prop firms in our registry, only 2 publish addresses whose contracts we could actually read — Hypernova and Propr, re-checked 15 August 2026 — while the firm that marketed hardest, GT Funded, has been offline since June 2026.
In one week, one firm told us a rival "is not an on-chain prop firm, they have no smart contracts". The rival's position was that hedging on a public venue and paying in crypto makes you on-chain. Both statements came from people with an obvious interest, and neither could be settled by reading marketing pages.
They could be settled in about twenty minutes with a public RPC endpoint. So that is what we did, and this page is the result. The uncomfortable part: the accusation was wrong on its specifics and right on its substance, and the firm making it turned out to have a weakness of its own that nobody had mentioned.
We are an affiliate of one of the firms on this page. That is exactly why every claim here carries the command that produced it. You should not have to trust us — that is the whole argument of the page, and it would be absurd to exempt ourselves from it.
The five tiers
Each tier is a test, not an adjective. A firm sits at the highest tier whose test it passes — and every tier above it is a thing you cannot check about that firm.
Everything in tier 4, plus the logic cannot be silently changed: verified published source, published ABI, and either immutable contracts or upgrades behind a timelock with a disclosed multisig threshold.
Account creation, the rules the account was opened under, status changes and equity updates are written to a public contract. A third party can reconstruct what happened to an account without asking the firm anything.
Payouts are settled on-chain and countable by a third party from the chain alone, without the firm's dashboard. You can see the amount and the recipient; not why it was owed.
The firm's own book is hedged or replicated on a public venue, so some market activity is visible. What is public is the firm's position, not yours.
Takes crypto for the fee, pays profit share in crypto. Everything else is a private database. This is a payment rail, not an architecture — and it is where most of the firms calling themselves on-chain actually sit.
Where each firm lands
The 'why not higher' column is mandatory. A ranking without it is a leaderboard, and leaderboards are what got this word into trouble.
| Tier | Firm | Why not higher |
|---|---|---|
| 4 | Hypernova | The TradingAccounts contract is an upgradeable ERC-1967 proxy, and both contracts share one owner that can replace the implementation. The rules are enforced in code today; nothing on-chain stops that code being swapped tomorrow. The owner is itself a contract rather than a bare key — better — but the signer threshold is not disclosed, so we cannot tell you how many people it takes. |
| 3 | Propr | Propr's own disclosure states that all accounts, funded ones included, are simulated, and that a signal replicated in live markets produces a position belonging to Propr. What reaches a public venue is Propr's book, not your account. And the payout wallets hold no balance, so there is no reserve to read. |
The findings, and how we got them
Each row is one thing we checked, what we ran, and what came back. The last column says which way it cuts — because a page where every finding flatters the same firm is not a measurement.
Hypernova
eth_call balanceOf on USDC (0xaf88…5831) for the Vault and the reserve wallet, then add them.
Vault + reserve wallet = $770,556.03. The homepage displayed $770,556. An exact match — the first time a firm's headline number has survived our first check.
eth_getLogs on the TradingAccounts address across the preceding ~5,000,000 blocks.
2,308 events across at least 8 distinct signatures in about fifteen days. The contract is in continuous use, not a showpiece deployed once for a screenshot.
eth_getLogs on USDC, filtered to Transfer events with the Vault as sender.
88 transfers out in about fifteen days, totalling $98,379.09.
eth_getCode on TradingAccounts (61 bytes — the size of a proxy, not of logic), then eth_getStorageAt on the ERC-1967 implementation slot, then owner() on both contracts.
It is a minimal ERC-1967 proxy. The admin slot is empty, so upgrades are gated by owner(); both contracts return the same owner, which holds 171 bytes of code — a contract rather than a bare private key, but with an undisclosed signer threshold. The claim does not hold as stated.
Propr
Open the Addresses tab on the transparency dashboard, then eth_getCode every address listed.
False as stated. There is a directory of seven addresses, and the Payout Distributor carries real contract code. A rival firm told us otherwise; checking took ten minutes.
Load the dashboard with JavaScript enabled, on a direct connection.
It publishes lifetime payouts ($456,701), average time to pay (5m 7s), funded traders (1,514) and a 13.3% pass rate on paid challenges — 1,514 of 11,404. Publishing a pass rate that unflattering, voluntarily, is rarer than it should be.
eth_getCode and eth_getTransactionCount on the Trader Payout Sender.
The address that actually pays traders has no contract code and has sent 1,707 transactions. It is an ordinary wallet driven by a private key, so payouts are initiated by Propr's backend and settled on-chain — real, but not contract-enforced.
balanceOf on USDC for every published payout address.
The Payout Distributor and both payout senders hold zero. There is no published reserve to read against. You can count what has been paid; you cannot check what remains to pay with.
Founder interview, 22 July 2026, cross-read against the rulebook's hybrid-model section.
About 5% of trader signals are copied to a live venue as A-booked positions; the rest stay on the internal book. The rulebook does describe the model and labels each trade. The ratio itself was not public before that interview.
Every address we read
Published by the firms themselves. The type column is not their description of the address — it is what eth_getCode returned.
| Address | Type | Holds |
|---|---|---|
| Hypernova — TradingAccounts (proxy)0x429d…4fcb | Contract61 bytes | — |
| Hypernova — TradingAccounts (implementation)0x8ed9…5a5b | Contract13,782 bytes | — |
| Hypernova — Vault0x9209…4c67 | Contract2,603 bytes | $32,056.01 |
| Hypernova — Payout reserve0x43c5…da07 | Wallet48 txs sent | $738,500.02 |
| Propr — Payout Distributor0x6e81…cf24 | Contract212 bytes | — |
| Propr — Trader Payout Sender0xFFbE…37e6 | Wallet1,707 txs sent | — |
| Propr — Affiliate Payout Sender0x7628…d056 | Wallet317 txs sent | — |
| Propr — Boomfi payment gateway0x66d3…0C22 | Wallet65 txs sent | $55,730.48 |
Reproduce this yourself
You need no account, no API key and no permission from anyone. A public RPC endpoint answers the same to everybody, which is the only reason any of this is worth writing down.
- 1Pick any public endpoint for the chain — we used the default Arbitrum and Ethereum ones, listed above.
- 2To ask whether an address is a contract, call eth_getCode. A wallet returns 0x. Anything longer is code.
- 3To read a USDC balance, call eth_call on the USDC contract with balanceOf and the address padded to 32 bytes, then divide by 1,000,000.
- 4To count activity, call eth_getLogs with the address and a block range, and group the results by their first topic.
- 5To see whether a contract is upgradeable, read the ERC-1967 implementation slot with eth_getStorageAt. A non-zero value means the logic lives somewhere else and can be pointed elsewhere.
https://arb1.arbitrum.io/rpc · If any of it comes back different from what is on this page, we would like to know — that is a correction, and corrections are the point.
One correction, and what it changed
Method matters as much as findings: on 8 August 2026 a VPN made us record a false negative about Propr's dashboard, corrected the same day. On-chain readings — like Hypernova's $1,001,060.62 reserve on 18 August 2026 — are immune to that failure mode, which is the argument of this page.
On the day these readings were taken, we recorded that Propr's transparency dashboard served no data at all — every metric a dash, every chart stuck loading. That was wrong. It was being blocked by a VPN we happened to be browsing through. On a direct connection the dashboard returned a full set of live figures in seconds.
The same evening, the same VPN produced a false reading on a different firm's trading terminal. Two corrupted verifications in one session, one cause.
So the rule now: anything that depends on a page rendering is checked on a direct connection and re-run on a second network before it is published as fact. On-chain readings are exempt — an RPC query returns the same bytes from anywhere, which is rather the argument of this entire page.
Go deeper
FAQ — on-chain prop firms
What does 'on-chain prop firm' actually mean?+
On its own, almost nothing — it is used for everything from accepting USDC to writing account rules into a smart contract. That is why we replaced it with five tiers, each defined by a test you can run yourself rather than by a claim.
Is an on-chain prop firm safer than a normal one?+
It is more checkable, which is not the same thing. Being able to read a payout reserve tells you the money exists today; it says nothing about the firm's risk management, its rules changing, or whether you will pass. Verifiability is one input among several.
Which firms actually have smart contracts?+
Of the 2 we have measured, both publish at least one address carrying real contract code. The difference is what the contracts do: one holds account rules and equity, the other distributes payments while the account itself lives in a private database.
Why is the top tier empty?+
Because no firm we have measured has contracts that cannot be quietly changed. The best-scoring firm runs an upgradeable proxy with an owner whose signer threshold is not disclosed. A scale whose top rung is already occupied stops being a target for anyone to aim at.
You are an affiliate. Why should I trust this?+
You should not have to. Every claim on this page carries the command that produced it and the block height it was read at. If you re-run one and get a different answer, tell us and we will correct it — we already had to correct ourselves once while building this page, and that correction is documented above.
Which prop firms are verifiable on-chain?+
As of 18 August 2026, two: Hypernova (tier 4 — account state on a public Arbitrum contract, smart-contract payouts in about 6.2 seconds, a reserve we read at $1,001,060.62 on 18 August 2026) and Propr (tier 3 — on-chain USDC payouts, including the 119.79 USDC we received ourselves on 31 July 2026 with a public hash). No firm reaches tier 5.
Can you check a prop firm's payout reserve yourself?+
Yes, if the firm publishes the address. Call balanceOf on the USDC contract for that address via any public RPC endpoint — that is how we read Hypernova's reserve at $1,001,060.62 on Arbitrum on 18 August 2026. Propr publishes addresses too, but its payout wallets hold no balance, so there is no reserve to read.