Security and audits
What protects contributed money, what the contracts and the keys can and cannot do, what review the code has had, and what review it has not had.
How contributor money is protected, what the contracts enforce, and what the keys that operate them can and cannot do. Every claim here is a line in a deployed contract or a public read you can run yourself; Contracts and addresses has the addresses and the commands.
What the contracts enforce
These are not policies or undertakings. They are conditions the contracts check on every call, and no key can switch them off.
Your ETH is refundable until a purchase spends it. Deposits wait in the pool in the order they arrived. A purchase spends the pool's own leftover ETH first, then deposits, the earliest first; until then, what a purchase has not spent is yours to take back, and taking it back works while the pool is paused.
Reserved cash is ring-fenced. Cash reserved for declared payouts, for withdrawals and for accrued fees is tracked separately from cash the pool may spend. Spending is bounded by what remains after all three, and a weekly instalment can never take what queued withdrawals need, so withdrawals always come first. A distribution that would turn positive equity into a deficit is rejected outright.
Capital comes back before anyone takes a fee. A release pays down what was actually spent on the branches being retired, then evidenced operating costs not yet recovered, then carried realised losses. Only what survives all three can create a fee.
The fee is fixed at 10% and cannot be raised. It is a compile-time constant, not a setting. It applies to the net realised result, vests over 30-day epochs, and later losses claw it back while it is unclaimed.
Reported value is bounded in code. A valuation report expires after one hour. An illiquid holding cannot be marked above three times the capital recorded as paid for it. The assumed ETH rate may rise at most 5% per day elapsed and may fall without limit. A charter the pool buys enters its value at what it cost, so buying one does not move the value per unit. The governor supplies the valuation, so the ceiling on overstating it is enforced by the contracts rather than trusted. A report that shows the unit price more than 10% below its recent high stops the pool converting deposits for 24 hours; the deposits stay refundable, and nothing else waits.
What the pool pays is bounded in code. A licence never costs more than its custody's price bound. A charter never costs more than 5% over the charter auction's last completed sale, nor more than that sale's price when the sale was the pool's own; the charter reserve refuses anything over 10% as well. Each swap stays within its custody's slippage allowance, and ETH a purchase does not spend returns to the pool in the same transaction.
A rotation cannot shrink the pool. It reverts unless every branch it retires is bought back in the same transaction, the Reserve's exit fee is within the policy's cap, which no setting can raise above 25%, and the charter's balance falls by no more than the release plus the licences at the price quoted.
The pool funds a custody only while it runs the published code. Before it hands the founder wallet or the charter reserve ETH or $STANDARD, the pool checks the custody's code against the code it was built for, and refuses while it runs anything else.
Withdrawals price everyone in a round identically, against a fresh report, at the pool's value per unit — the same value a purchase uses — paying only from funded cash. A price floor you attach is honoured: below it you are skipped, never filled at a price you did not accept.
Rounding cannot drift in the pool's favour. All arithmetic is integer and floors at each step in a fixed order. Residues too small to allocate stay inside the contract and are not sweepable by anyone.
Every value-moving function is reentrancy-guarded, and payouts are recorded before they are sent.
If governance goes silent, members can act without it. After 60 days with no report and no accounting action, emergencyDistribute() becomes callable by anyone and pays out idle $STANDARD beyond every reserve, plus the ETH the pool owns beyond its protected claims, pro rata. It works while the pool is paused.
What no one can do
| Nobody can | Because |
|---|---|
| Transfer, sell or seize your units | Units are nontransferable entries. There is no transfer function, no approval and no market |
| Mint units to themselves out of nothing | Units are created only when a purchase spends deposited ETH, at the pool's value then. But that value comes from the governor's valuation report, and a report that understated the pool would convert deposits cheaply and dilute existing members. That is the governor's valuation authority, the one risk the reviews singled out, and the reason for the bounds above. A sharp mark-down stops conversions for a day, so it cannot be converted against at once |
| Raise the fee | A compile-time constant |
| Redirect the fee | The recipient is immutable and survives any change of the pool's ownership |
| Move a charter out of the charter reserve | It has no function to transfer a charter or approve a transfer, and its code cannot be replaced |
| Transfer charter 729 | The Reserve does not currently allow charters to be transferred |
| Keep the 60-day fallback shut by pausing | It is permissionless and works while paused. Only a report or an accounting action starts its 60 days again |
| Stop you taking back ETH a purchase has not spent | Enforced by the contract, and open while the pool is paused |
What the operating keys can do
Three keys run the pool. This is the full map.
| Key | Can | Cannot |
|---|---|---|
| Keeper | Twelve governed actions, and only through the governor: file reports, start the pool, buy licences, buy a charter, admit deposits, settle the weekly round with its instalment, rotate branches on charter 729 or on a reserve charter, claim vested fees to the fee recipient, return idle ETH, open and settle an announced admission window | Withdraw funds, make arbitrary calls, choose a recipient, pay past the bounds above, change policy, pause, appoint its own replacement, or reach a custody directly |
| Governance multisig, 2-of-3 | Pause the pool, revoke the keeper and switch charter purchases off, as guardian of the governor and of the charter reserve; receive the fee; and replace the pool's code through the proxy | Do anything else without an upgrade. Two signatures can replace the pool's code, and replacement code can do anything the pool can — including what this row and the table above say nobody can. That is the ceiling on every guarantee on this page |
| Founder key | Everything the keeper can, plus set policy (including the reserves and other liabilities the reports count), appoint or revoke the keeper, pause and unpause, close or open the intake, call any governed function on the pool, hand the pool to a new governor, set both custodies' licence limits, switch charter purchases on or off, supply the charter sale history's earlier sales, and move what is physically held in the founder wallet | Raise the fee, redirect the fee, upgrade the pool's code with this key alone, move a charter out of the charter reserve, take a member's units, or have the pool fund a custody while it runs other code |
The founder wallet is the custodian of charter 729, and that is the trust you are extending. Contributors' ETH rests in the pool contract and reaches the wallet only inside a transaction that returns what it does not spend before that transaction ends, and only while the wallet runs the published delegate code. What a retirement realises is transferred to the pool in the same transaction that realises it, and the contract refuses the receipt unless its own balance rose by the full amount. What persists under the wallet's key is charter 729, the branches on it and their balance inside the Reserve, and the small $STANDARD remainder of a swap.
That is a real reduction, not a removal. The bounds on what the keeper may do are enforced by the governor contract, which is what the keeper calls; the wallet itself answers governed calls only from the pool, and answers its own key without limit. pooledNative() and pooledStandard() report what of the wallet's balances is the pool's, but that separation is bookkeeping rather than a second signature. Custody and the founder states the arrangement in full.
The charter reserve has no such key. The charters the pool buys, their branches and their balance are held by a contract that answers only the pool, keeps no ETH between calls, and has no function to move a charter out.
How the code was tested
Adversarial audit, 16 September 2026. A line-by-line audit against the accounting specification under a threat model with four actors: an untrusted member or outsider, the governor, the unverified upstream contracts, and the $STANDARD token itself. It traced every ETH and $STANDARD movement, every reserve mutation, every arithmetic underflow path, roster slot reuse, report freshness and every reentrancy and callback surface. Each candidate finding became an executable proof-of-concept against the real contracts rather than a note.
It found no path by which an unprivileged member or an outsider could steal or freeze funds. Three High, four Medium and four Low findings were fixed in the same revision, and every original attack path is replayed by a rescreen suite that runs with every build. The risk it did surface was the governor's unbounded valuation authority and the pool's dependence on the governor staying alive, which is why the report bounds and the 60-day fallback above exist.
Independent review, 16 to 17 September 2026. A second reviewer, working separately, which reproduced the build and test baseline itself and treated the first audit as claims to test rather than prior approval. It raised thirteen further findings across the dividend ledger, the governance fallback, valuation, price floors, the settlement grace, fee timing and the consumer application, each with executable witnesses. It established no Critical or High defect.
Response revision, 17 September 2026. Fixes for the dividend-ledger packing and the payout restrictions it introduced, withdrawal rounding and allocation eligibility, and a pre-existing solvency gap in governed native distribution. Three regression suites cover them.
External review, 20 September 2026. An outside agent with read access to the whole system reviewed the contracts, the keeper and the application together, and returned findings on the governor's formal windows, the withdrawal queue watch, the idle clock and the Safe check. Its re-review confirmed the responses.
No member cap, 21 September 2026. The pool's code was rewritten so that settling a withdrawal round or converting deposits never walks the membership: each is worked out once from totals kept per price-floor rung and stored, and a member's share is applied when their account is next touched. Three differential fuzz campaigns of 20,000 runs each found a real defect in the first build of it — rounds left unapplied could carry a member's units above the supply — which was corrected and pinned by the three scenarios that reproduce it.
Independent review of that change, 22 September 2026. A separate reviewer confirmed four defects: no migration path for an active ledger, two packed fields that truncated a value instead of refusing it, and a long-stop bound that could exceed the old pool's own calculation. Each was corrected and pinned by a test that fails against the code before it.
Independent review of the charter relaunch, 27 September 2026. A reviewer with no part in writing the relaunch read the pool's four parts, the deposit queue, the charter reserve, the charter sale history, the governor and its lens in full, reproduced the build and the tests, and compared every part's storage layout with the version before it. It found one launch blocker — the charter sale history could never be recorded on Robinhood Chain, whose block numbers are not Ethereum's — and one medium finding: a fee spike at the Reserve could make units cheap for whoever deposited during it. Both were fixed before deployment, each with a test that fails without the fix.
Review of the retirement strategy, 28 September 2026. A second review of rotation, the weekly instalments and a leaver's exit from the branches, since removed, found five medium and two low issues, each fixed with a regression test. The findings that concerned only the removed exit went with it; the two that stand are that anyone could cancel the week's instalment, and that the rotation fee cap made entry cheaper.
Pre-auction review, 28 September 2026. Before the pool could buy its first charter, a review of the charter-buying governor found four issues, fixed in the governor now deployed: the pool's own purchase could have raised the price of its next charter; a charter's earnings were counted as filling its empty slots; a charter purchase could stand without the charter reserve checking in with the Reserve; and a slot that could not be priced was not counted as needing ETH.
Deployed-code verification. The deployment and upgrade tools compare each contract's code byte for byte against the reviewed build before anything is signed that uses it, and read ownership, bindings, roles and policy back from the chain rather than assuming them. The charter relaunch was checked this way, and so were the current governor, its lens and its sale history before the pool was handed to it. You can repeat the reads yourself from Contracts and addresses.
Continuous testing. Several hundred contract tests — lifecycle, governor, rotation, charter purchases, the deposit queue, proxy, custody, round records, withdrawal mathematics, fee and valuation review, packing, solvency, the audit rescreen and the upgrades — plus an application suite of over a thousand checks, both run on every change. Further suites run against a fork of the live chain and are opt-in; for the relaunch they passed against the live chain, including a rotation on charter 729 against the live bank. Compiler and toolchain are pinned: Solidity 0.8.30, Foundry 1.5.1, optimiser at 200 runs, OpenZeppelin 5.6.1.
What we have not done
Stated plainly, once, so you can weigh it.
No security firm has been paid to audit this code. The reviews above were adversarial, executable and independent of one another, but they were run by AI agents on the owner's instruction, and no outside firm has signed off on this deployment or carries liability for it. There is no bug bounty.
The pool is upgradeable, and has been upgraded six times. It runs behind a proxy, so two of the multisig's three owners can replace its code. That is how the member cap was removed, how deposits became refundable until a purchase spends them, and how the pool came to buy charters; it is the only route by which the constants above could ever change. An upgrade is not announced on chain, so the implementation slot is the thing to read if you want to know what code is running; the commands are on Contracts and addresses.
Each upgrade was designed to leave existing storage untouched and was rehearsed against the live pool on a fork. Before the charter relaunch, a script compared every part's storage layout, field by field, with the version before it. That comparison runs before an upgrade, not in every build, so storage safety between upgrades still rests on review and on that check being run.
The pool has little operating history. There is no observed evidence yet that running it earns more than it costs, and no promised return, minimum payout or withdrawal completion date.
A closed website is not a closed pool. The contract accepts deposits whenever its intake is open, whatever the app shows. Anyone transacting with it directly is outside every check the app performs, including the terms they would otherwise sign.
If you find a problem
Read the code first. The addresses and the verification commands are on Contracts and addresses, and every claim above is a public read or a line in a contract.
Report anything you find privately before publishing it. There is no bounty, and that is not a reason to disclose a live vulnerability while member funds are exposed to it.
If the pool behaves differently from this page, the chain is the authority and this page is wrong. Tell us.