Dolomite risk

TL;DR

The red we published in July has cleared, and it cleared for real. The USD1 pool that sat near 93% utilization now runs at 54.3%, and I confirmed it three independent ways: DeFiLlama reports $272.84M supplied against $148.22M borrowed, the USD1 token balance of Dolomite's own core contract reads 124,776,796 tokens, and the contract's internal par accounting gives the same ratio. Depositors can exit today.

Two things replaced it. First, the delay on admin actions. Dolomite's admin page describes a one-day timelock, 86,400 seconds, and links a contract on Arbiscan so you can check. That contract is still alive and still returns 86,400. It no longer owns anything. Authority moved to a newer contract, DolomiteOwnerV2, which was deployed on 2025-06-22 with a 300-second delay written into its constructor arguments and has run at 300 seconds ever since, on both Ethereum and Arbitrum. Behind it is a 2-of-3 Safe whose three signers are ordinary keys. Five minutes, two keys, $896M of supplied assets.

Second, a collateral market that did not exist when we last looked. WLFIcx was deployed on 2026-07-07, six days after our last report, and now holds $110.3M. It is a custodial wrapper on WLFI, issued by Concrete and custodied at BitGo, marked at exactly the WLFI price. It is also an upgradeable proxy, and its own timelock has never been activated: getTimelock() returns the zero address and false, which by the contract's own code makes every delay check a no-op. That is the finding I would want a lender to read twice.

One more thing worth saying because we publish both numbers. DeFiLlama's $225.00M and our own board's $880.2M are not in conflict and neither is wrong. They count different things, and the entire difference is the WLFI pair.

Checklist

Audits & contracts. The margin core is genuinely well covered, and every firm named checked out against the official docs page: SECBIT Labs (2021-08-02, pre-Arbitrum), Cyfrin (2023-08-23 core, 2025-04-24 POLVaults v2.0), Zokyo (2023-04-19, GLP and Isolation Mode), Guardian Audits (2024-01-11), and OpenZeppelin (as Zeppelin Solutions) plus Bramah Systems on the original dYdX Solo Margin base. There is also a bug bounty, which our own draft of this report wrongly said did not exist. It is not on Immunefi and has no published maximum payout, which is why a keyword search misses it, but the docs carry a Bug Bounty page: disclosure to security@dolomite.io ↗, an initial confirmation within 72 hours, a commitment not to pursue legal action against good-faith reporters, and a monetary reward scaled by OWASP risk assessment. Dolomite also retains Chainalysis Crypto Incident Response. Credit where it is due.

What is not covered is the part that holds the keys. DolomiteOwnerV2, deployed 2025-06-22, is the contract with admin authority over every core deployment, and it matches none of the six published engagements, which cover the margin repository and the POL vault modules. Neither does WLFIcx, the third-party token now carrying $110.3M of collateral. Yellow, on coverage rather than on any discovered defect in Dolomite's own code.

Admin control. Measured, not read off a page, and then the page was read too. On Ethereum the core DolomiteMargin at 0x003Ca23F returns owner() = 0xC2B66E24, which is DolomiteOwnerV2, the same address that owns the Arbitrum core. Calling secondsTimeLocked() on it returns 300 on both chains, and the deployment constructor arguments were the Safe address and 0x12c, which is 300, so this is not drift: it was 300 from the first block, 416 days ago.

The docs are stale rather than dishonest, and the distinction matters. They present 86,400 seconds as the launch value, say the intention was to raise it, and explicitly warn that they may go out of sync while linking a contract to check. That linked contract, 0xE412991F on Arbitrum, still returns 86,400 today. It simply is not the owner any more. The older DolomiteOwnerV1 at 0xCf359A2f returns 1,800. So the actual sequence across Dolomite's admin wallets runs 86,400, then 1,800, then 300, which is the opposite direction from the one the docs describe, and nothing in the docs points a reader at the contract that is live.

DEFAULT_ADMIN_ROLE is held by Gnosis Safe v1.4.1 at 0xa75c21C5, threshold 2, three owners, all of which return zero bytes of code, so three plain keys. Reading the verified source, isUserApprovedToSubmitTransaction returns true immediately for that role, before even the guard that stops other roles targeting the owner contract itself. So the Safe can submit any calldata to any destination: list a market via ownerAddMarket, change a price oracle, change an interest setter, raise a supply cap, or reconfigure the owner contract. Execution is gated by pastTimeLock, which passes once 300 seconds have elapsed, and validExecutor admits DEFAULT_ADMIN_ROLE directly, so the Safe both submits and executes. ownerSetSecondsTimeLocked is onlySelf, meaning the Safe can set the delay to zero after one 300-second wait.

There is one guard I want to credit, because it is better than most: _grantRole reverts with "DolomiteOwnerV2: Admin cannot bypass timelock" if one address would hold both DEFAULT_ADMIN_ROLE and BYPASS_TIMELOCK_ROLE, in either order. It does not stop the admin granting bypass to a fresh address, but that grant is itself subject to the 300 seconds, so 300 seconds is the real floor. I checked the Safe against all seven non-admin roles with hasRole on both chains, including the three whose names I could not resolve, and it holds none of them. On the question of who can list collateral, Dolomite's own docs say only the protocol administrator can, with a DAO-appointed listing committee described as future work. Today that is this Safe, in five minutes.

One methodology note, because it nearly produced a false claim. getRoleAddresses returns an empty array for every role including DEFAULT_ADMIN_ROLE. That is not evidence of an unowned contract. The source shows it reads _roleToAddresses, a separate registry written only by ownerAddRoleAddresses, distinct from the OpenZeppelin mapping hasRole reads. The direct hasRole call is what confirmed the Safe.

Oracle. Every one of the 23 Ethereum markets routes through one OracleAggregator at 0xa150Ef2D, confirmed by calling getMarketPriceOracle on each. The published per-asset table is reasonable as far as it goes: direct Chainlink market price for WETH, USD1, USDC, LINK, USDT, CRV, AAVE, sUSDe, WLFI and cbBTC; two-leg Chainlink compositions for WBTC, weETH, mETH and SolvBTC; Chronicle for srUSD, rUSD and wsrUSD; Redstone for cUSD, with an ERC4626 exchange rate in front of it for stcUSD; and a Uniswap V3 TWAP into Chainlink for DOLO. The problem is where it stops. That table ends at market 19, and there are 23 markets, so markets 20 (wstETH), 21 (BTCcx) and 22 (WLFIcx) have no row in it.

For the two Concrete markets that gap is filled elsewhere rather than left open, and our draft was wrong to imply otherwise. The Concrete Assets page states the rule plainly: ASSETcx assets use the underlying asset's Chainlink price feed. I confirmed the consequence on chain. Market 22's price reads 0x00c40d257731ec00, which decodes to $0.055184, byte for byte identical to the WLFI market price. So the custodial wrapper is marked at the price of the thing it represents, with nothing priced in for custody or for whether it can actually be redeemed, and that is disclosed design rather than a hidden setting. It is still a risk, and it is still the reason a custody failure would not show up in Dolomite's collateral maths at all. Market 20 is the one with no published treatment anywhere, at $6.8M. Dolomite's USD1 mark reads $0.999388, consistent with the Chainlink USD1/USD feed that read $0.99934 on 2026-08-10 for the WLFI report.

Liquidity & exit. This is the improvement, and it deserves to be stated as plainly as last month's warning. In April the USD1 pool held roughly $180M against $167.5M borrowed, about 93% utilization and $12.5M idle, and retail depositors could not all get out. Today it holds $272.84M against $148.22M borrowed, 54.33% utilization, $124.62M idle. Both legs moved in the right direction: about $93M more supplied and about $19M less borrowed, consistent with roughly $25M of World Liberty repayments in the first half of April. I did not take DeFiLlama's word for it. The USD1 ERC-20 balance held by the Dolomite core contract is 124,776,796.10, within 0.13% of the reported idle figure, and the contract's own getMarketTotalPar for USD1 gives a borrow-to-supply ratio of 52.8% in par terms before index scaling. Three measurements, one answer.

The tighter market now is Ethereum USDC at 80.5% utilization, where I read 25,797,684 idle USDC on-chain against $131.76M supplied. That is thinner cover than USD1 has. And the structural point survives the good news: $386.1M of the collateral base is a token whose forced sale would move its own price, so exit for a stablecoin depositor still depends on that position never needing to be liquidated at once. Yellow, not green.

Yield (real vs emission). Our board prints USD1 as Dolomite's top supply APY at 7.85%. The composition matters more than the number. DeFiLlama splits today's 7.78% into 1.57% base and 6.22% reward, so roughly four fifths of it is incentive rather than borrower interest. Ethereum USDC is 7.91% total, 3.47% base and 4.44% reward. For contrast, Arbitrum USDC pays 4.09% with no reward component at all, which is what an unsubsidized number on this protocol looks like. Treat the USD1 headline as a paid rate on the pool with the most concentrated borrower, not as evidence of broad demand.

Holder concentration. Two layers, and the collateral layer is the one that can cost you money. WLFI sits at 4,998,327,305 tokens against a supply cap of 5,100,000,000, so 98.0% full. WLFIcx sits at 1,998,154,320 against a cap of 2,000,000,000, so 99.9% full. At Dolomite's own oracle price both come to $386.1M, which is 43.1% of the $896.2M supplied across the protocol. As a share that is down from roughly 55% in April, but that is mostly the denominator growing; the WLFI collateral itself has barely moved since the April reporting, and the exposure is now split across two markets instead of one. Calling this one issuer's family would be imprecise, since WLFI is issued by World Liberty Financial and WLFIcx by Concrete. It is one underlying asset in two wrappers from two issuers, which concentrates the price risk without concentrating the counterparty risk.

WLFIcx is the striking one. Its total supply is 1,998,154,420.36 tokens and Dolomite holds 1,998,154,320.36 of them, which is every token in existence but a hundred. Dolomite's own documentation explains why: Concrete "supplies the token as collateral for a specific, actively managed, borrowing position held to a repayment schedule." That is a single counterparty by design.

Then there is what the token itself can do, which no page discloses and which I read from the deployed code. WLFIcx is an ERC1967 proxy over an implementation named AssetCX, authored by Blueprint Finance, upgradeable under UUPS, pausable, with role-gated mint and burn. Its privileged actions are supposed to route through a timelock. On the live token getTimelock() returns the zero address and false, and the library it calls is explicit: enforceDelay runs the check only if ($.active). Enforcement has never been turned on, so today an implementation upgrade, a role grant including the minting role, and a mint or burn are all immediate for whoever holds the roles. pause() is instant by design and by comment, while unpause() is timelock-gated, an asymmetry that would matter a great deal if the token holding $110.3M of collateral were ever frozen mid-liquidation.

Both markets return getMarketIsClosing true and are not borrowable, so they are collateral only. That is worth stating precisely rather than as a special containment measure: 13 of the 23 Ethereum markets carry the same flag.

On the token side, DOLO trades at $0.02289 with a market cap of $11.66M and 509.7M of 1B circulating, up from about 441.6M in July as unlocks continue, and 93.8% below its August 2025 high. DOLO governance does not control the admin surface; the Safe does. DOLO is itself a listed collateral market priced off a Uniswap V3 TWAP, which is thin work for an $11.7M float, though the exposure is small: the market holds 2,500,315 DOLO, about $57,000, against a 15,000,000 token cap.

Track record. Still clean at the protocol level and I re-checked rather than carrying it forward. DeFiLlama's hacks list contains exactly one Dolomite entry across its whole history: 2024-03-20, $1,800,000, classified as protocol logic with an access control exploit technique, on Ethereum. That was a discontinued 2019 contract reachable through stale user approvals, not the live protocol; roughly 90% came back from the exploiter and the treasury covered the rest. Of the 173 hacks that list logs for 2026, none is Dolomite, and news searches turned up no incident since our 2026-07-01 report.

What tempers this is age rather than events. The Ethereum core was deployed 2025-06-21, so 416 days old, and it now carries 94.0% of the supplied value while the multi-cycle track record belongs to Arbitrum, which holds $29.3M. The largest new market on the newest deployment is 35 days old. One related-party fact belongs on the record too: Dolomite co-founder Corey Caplan is an adviser to World Liberty Financial, whose treasury is the dominant borrower here and whose WLFI Markets front end routes into this protocol.

Worst case

Rank them by what I measured rather than what is imaginable.

Two of three keys are the whole story on the admin side. A signer compromise gives an attacker the power to list a market with an oracle they control, borrow every asset against worthless collateral, and walk. The only thing standing in the way is 300 seconds, and after one such wait the delay itself can be set to zero. Nobody monitors a mempool for five minutes on a Sunday. The docs make this worse in a specific and fixable way: a careful reader who follows the link Dolomite provides lands on a contract that returns 86,400 and concludes they have a day.

The WLFIcx path is new, less charted, and the code is worse than the disclosure. That token is a claim on WLFI held with a custodian, issued by a third party, priced at par with the underlying, backing one managed position on a repayment schedule nobody outside the deal has published. Its own contract can be upgraded, minted from, or paused with no delay at all, because the timelock it was built to sit behind was never activated. Two concrete failures follow. A pause freezes $110.3M of collateral in place while the debt it backs keeps accruing, and un-pausing requires a timelock proposal that cannot currently mature. An implementation upgrade or a mint changes what the collateral is while Dolomite's oracle keeps marking it at the WLFI price regardless, because that is what it is configured to do. No insurance fund appears anywhere in the docs.

The collateral case is slower and more likely. WLFI falls hard. Liquidation engines start selling into a token whose depth cannot absorb $386.1M, the sale drives the price down further, collateral stops covering debt, and the shortfall lands on the USD1 and USDC lenders on Ethereum. Those two pools have $124.8M and $25.8M of idle liquidity against $148.2M and $106.1M borrowed, so the exit queue would form long before the last lender got out. Isolation mode should keep the damage away from the Arbitrum and Berachain markets, and the supply caps mean the position cannot grow further without an admin action, which are both genuine mitigants.

Bottom line

Caution, same verdict as July, different reasons, and the reasons are worth reading rather than skipping. The thing we told people to avoid last month, the trapped USD1 pool, genuinely fixed itself, and $124.8M of withdrawable USD1 sitting in the contract right now says so. The audit coverage is real, and so is the bug bounty we nearly published as absent.

But a 300-second delay behind two of three plain keys, over contracts holding $896M, is not what the documentation points a reader at, and it is not enough time for anyone to do anything. And 43.1% of the collateral base is one underlying token in two wrappers, both parked at their caps, half of it in a 35-day-old upgradeable proxy that is marked at full price and whose own safety delay has never been switched on. Two reds, both on surfaces where you lose the money.

The usable part of this protocol is the boring part: the Arbitrum markets, where USDC pays 4.09% with no emissions propping it up, sizes are small, and the assets are ordinary. Do not lend into USD1, WLFI or WLFIcx. If you already have, the door is open today, which is more than we could say in July. Re-rate upward if the timelock on DolomiteOwnerV2 is raised to something that matches what the docs describe, if the WLFIcx timelock is activated, and if the WLFI pair drops below a quarter of supplied value. Re-rate downward on any admin action that raises those caps, or on any WLFIcx implementation upgrade.

Data appendix

• TVL: $225.00M (DeFiLlama dolomite, id 2187, category Lending). 90d: $164.97M on 2026-05-14, so up 36.4%. ATH: $929.46M on 2025-03-03, so down 75.8%. Chains carrying value: Ethereum $189.97M, Berachain $18.33M, Arbitrum $16.64M, Mantle $0.06M, X Layer $742. Botanix and Polygon zkEVM read zero.
• Reconciling our two published numbers. DeFiLlama's headline TVL is net of borrows and excludes the WLFI pair. Its currentChainTvls non-borrowed keys sum to exactly 225,001,588 and its borrowed key is 279,189,534, which matches our board's $279.2M borrowed to the dollar. That implies gross supplied of $504.2M on the TVL adapter. Our /lending board sums totalSupplyUsd from yields.llama.fi/lendBorrow, joined to yields.llama.fi/pools by pool id, which today returns $896.2M gross across 63 Dolomite markets on 3 chains (the board's $880.2M is the same endpoint an earlier cron cycle; the market count flickers between 62 and 63 between refreshes as a zero-value pool drops in and out). The $392M difference is WLFI $275.83M plus WLFIcx $110.27M, which is $386.1M: subtract those two from the endpoint's Ethereum supply of $842.95M and you get $456.86M against the TVL adapter's Ethereum gross of $456.26M, a 0.13% gap. So the DeFiLlama TVL adapter omits both WLFI-family markets and nets borrows; our board includes them and shows gross. Both are internally consistent. The $601M "supply minus borrowed" figure reconciles with nothing, because it nets borrows out of a gross number that already includes the assets the other adapter drops. Per-chain supply: Ethereum $842,954,565 across 15 markets (94.0%), Arbitrum $29,301,056 across 29 (3.3%), Berachain $24,150,549 across 19 (2.7%).
• USD1 pool, re-measured three ways. DeFiLlama: $272,838,667 supplied, $148,220,342 borrowed, 54.33% utilization, $124,618,325 idle. On-chain: USD1 (0x8d0D000E) balanceOf DolomiteMargin (0x003Ca23F) = 124,776,796.10 tokens, within 0.13%. Contract getMarketTotalPar(1): borrow par 139,192,417.23 against supply par 263,463,755.61, a 52.8% par ratio before index scaling. Prior report, from CoinDesk's April window: about $180M supplied, about $167.5M borrowed, about 93%, about $12.5M idle. Supply cap 1,000,000,000 USD1, so 12.4% full.
• Other exit-relevant markets: Ethereum USDC $131.76M supplied against $106.06M borrowed, 80.5% utilization, idle balance read on-chain as 25,797,683.54 USDC, par ratio 78.0%. Ethereum WETH $34.31M against $11.99M, 35.0%, idle 11,879.68 WETH. Arbitrum USDC 66.9%, Arbitrum WETH 81.3%, Arbitrum WBTC 75.4%.
• Audits: SECBIT Labs 2021-08-02; Cyfrin 2023-08-23 (margin core) and 2025-04-24 (POLVaults v2.0); Zokyo 2023-04-19 (GLP and Isolation Mode); Guardian Audits 2024-01-11; OpenZeppelin (as Zeppelin Solutions) and Bramah Systems on the dYdX Solo Margin base. All six confirmed on docs.dolomite.io/audits-and-security this run; DeFiLlama's protocol record sets audits: "2" and links the Cyfrin 2023 and SECBIT 2021 PDFs. Whether DolomiteOwnerV2 (2025-06-22) or the Ethereum v2 core (2025-06-21) fall inside any published audit: refresh, no match found.
• Bug bounty: exists, and our draft was wrong to say otherwise. Docs carry a Bug Bounty page: report to security@dolomite.io ↗, initial confirmation within 72 hours, no legal action against good-faith reporters, 30-day confidentiality expected, "a monetary reward based on the OWASP risk assessment methodology". Not on Immunefi, and no maximum payout is published, which is why a platform-name search returns nothing. Dolomite separately retains Chainalysis Crypto Incident Response. No insurance fund found anywhere in the docs.
• Admin and governance, all read on-chain today. Ethereum DolomiteMargin 0x003Ca23Fd5F0ca87D01F6eC6CD14A8AE60c2b97D (created 2025-06-21 16:11 UTC, 416 days old, 22,872 bytes, 23 markets), owner() = 0xC2B66E247daE5Ee749Ae1d827190115F3653dE06 = DolomiteOwnerV2 (verified source, 13,011 bytes, created 2025-06-22 11:13 UTC, constructor arguments 0xa75c21c5be284122a87a37a76cc6c4dd3e55a1d4 and 0x12c = 300). The Arbitrum core 0x6Bd780E7 (77 markets) returns the same owner. secondsTimeLocked() returns 300 on Ethereum and 300 on Arbitrum. transactionCount() = 313 on Ethereum. Arbitrum DolomiteOwnerV1 (0xCf359A2f) returns 1,800. The delayed multisig the docs link on Arbiscan, 0xE412991Fb026df586C2f2F9EE06ACaD1A34f585B, still returns 86,400 and has transactionCount() 56, but no longer owns either core. Docs quote: "At the time of launch, the delay on the delayed multi signature wallet is 1 day (86,400 seconds). The intention is to raise it incrementally."
• Admin, continued. hasRole(DEFAULT_ADMIN_ROLE, 0xa75c21C5) is true on both chains; hasRole is false for all seven other roles on both chains. 0xa75c21C5BE284122a87A37a76cc6C4DD3E55a1D4 is a Gnosis Safe, VERSION() "1.4.1", getThreshold() 2, getOwners() three entries (0x52256ef8, 0x42acd316, 0xbdef2b20), all with zero code and nonces 1015, 14 and 352, Safe created 2022-03-02. getRoles() lists 8: DEFAULT_ADMIN (bytes32 zero), BYPASS_TIMELOCK, EXECUTOR, SECURITY_COUNCIL, LISTING_COMMITTEE, plus 0xcd86ded6..., 0xebeb1fd6... and 0xb74ffa3c... which resolve to no name I could find. Source review: isUserApprovedToSubmitTransaction returns true immediately for DEFAULT_ADMIN_ROLE, ahead of the _destination != address(this) guard that constrains every other role; _executeTransaction is gated by pastTimeLock, which passes on isTimelockComplete || hasRole(BYPASS_TIMELOCK_ROLE); validExecutor admits EXECUTOR_ROLE or DEFAULT_ADMIN_ROLE; submitTransactionAndExecute combines both steps; ownerSetSecondsTimeLocked is onlySelf, so the delay can be zeroed after one 300-second wait; _grantRole reverts "DolomiteOwnerV2: Admin cannot bypass timelock" in both directions. Caution on method: getRoleAddresses returns empty for every role because it reads _roleToAddresses, a separate registry written only by ownerAddRoleAddresses; it is not the mapping hasRole reads and it is not evidence of an unowned contract. Signer identities and the holders of the seven non-admin roles: refresh (archive eth_getLogs for RoleGranted was rejected by every free RPC tried).
• Market listing authority: ownerAddMarket on DolomiteMargin, reachable by the 2-of-3 Safe through submitTransaction plus a 300-second wait. Docs: "Only the protocol administrator can list new markets. Over time, the DAO will have the right to add assets as well as a listing committee through an ownership adapter." Current LISTING_COMMITTEE_ROLE membership: refresh.
• Oracle: all 23 Ethereum markets point at OracleAggregator 0xa150Ef2D5827dB283321D15d62d5D07fB41d636E (created 2025-06-22; verified by calling getMarketPriceOracle on every market id 0 through 22). The published Ethereum table covers market IDs 0 to 19 only: Chainlink (Market Price) direct for WETH, USD1, USDC, LINK, USDT, CRV, AAVE, sUSDe, WLFI and cbBTC; two-leg Chainlink for WBTC (exchange rate to BTC), weETH (exchange rate to ETH), mETH (market price to ETH) and SolvBTC (market price to BTC); Chronicle for srUSD, rUSD and wsrUSD; Redstone (Exchange Rate) for cUSD, with an ERC4626 exchange rate in front for stcUSD; Uniswap V3 TWAP into Chainlink for DOLO. Markets 20 (wstETH, $6.82M), 21 (BTCcx) and 22 (WLFIcx) have no row there; for 21 and 22 the Concrete Assets page supplies the rule instead: "ASSETcx assets use the underlying ASSET Chainlink price feed." Live prices read from the contract: USD1 $0.999388, USDC $0.999803, WETH $1,879.29, WLFI $0.055184, WLFIcx $0.055184 (identical raw word 0x00c40d257731ec00), DOLO $0.022883. Per July's report, Arbitrum adds Pendle and Camelot TWAPs and Berachain uses Chronicle primary with Kodiak TWAP for governance tokens; not re-verified this run.
• Collateral concentration, on-chain: WLFI (0xdA5e1988) balance in the core 4,998,327,305.75 tokens against getMarketMaxWei cap 5,100,000,000, so 98.0% full, $275.8M at the protocol's own mark. WLFIcx (0x8AF3B9D0E53186ae13b32FD783f42663Bf48d389) balance 1,998,154,320.36 against cap 2,000,000,000, so 99.9% full, $110.3M. WLFIcx total supply 1,998,154,420.36, so Dolomite holds all but 100 tokens. Combined $386,090,694, which is 43.1% of $896.2M supplied. Both markets return getMarketIsClosing true and are non-borrowable, as do 11 other Ethereum markets (13 of 23 in total). These two on-chain products match DeFiLlama's pool figures of $275,825,445 and $110,265,249 to within a rounding error.
• WLFIcx, what it is and what it can do: token name "World Liberty Financial Custodial Asset", symbol WLFIcx, 18 decimals, deployed 2026-07-07 09:06:23 UTC by 0xafe782c8, which is 6 days after our previous report and 35 days before this one. Docs: "Concrete Assets connect assets held in qualified custody to Dolomite's onchain markets... For ASSETcx, Concrete supplies the token as collateral for a specific, actively managed, borrowing position held to a repayment schedule." BTCcx and WLFIcx are named as the first two, custodied with BitGo, borrowed funds "generally deployed into yield bearing strategies". Read from the deployed code and not disclosed on any Dolomite page: the token is an ERC1967 proxy (implementation 0x86c81b3c29061369e5dcf6a4af841b24ddd87f0a, contract AssetCX, authored by Blueprint Finance), upgradeable under UUPS, ERC20Pausable, with mint and burn gated by MERCHANT_ROLE and upgrades by DEFAULT_ADMIN_ROLE. Those actions are meant to sit behind a timelock, but getTimelock() returns (0x0000000000000000000000000000000000000000, false) and the library is explicit: function enforceDelay(...) { if ($.active) ITimelock($.timelock).checkDelayExpired(...); }. Enforcement was never activated, so every delay check is currently a no-op. pause() carries the comment "Intentionally NOT timelock-gated"; unpause() requires a matured TIER_2 proposal. paused() reads false today. Role holders on WLFIcx, the repayment schedule, and the counterparty behind the managed position: refresh.
• Yield split (DeFiLlama, 2026-08-11): Ethereum USD1 7.78% total = 1.57% base + 6.22% reward. Ethereum USDC 7.91% = 3.47% base + 4.44% reward. Arbitrum USDC 4.09%, all base, no reward leg. Our board's 7.85% USD1 figure is the same pool one cycle earlier.
• Token concentration: DOLO $0.02288538, market cap $11,661,721 (rank 1073), FDV $22,838,733, circulating 509,723,226 of 1,000,000,000 max (51.0%, up from about 441.6M in July), 24h volume $5,255,079, ATH $0.366425 on 2025-08-31 so down 93.8% (CoinGecko, last updated 2026-08-11 21:23 UTC). The DOLO market on Dolomite holds 2,500,315.40 tokens, about $57,000, against a 15,000,000 token cap. Allocation per Dolomite's distribution docs, carried from July and not re-verified this run: insiders and foundation about 49%, community 50%+. Wallet-level top-holder breakdown: refresh.
• Incident record: DeFiLlama's hacks list returns exactly one Dolomite entry, 2024-03-20, $1,800,000, classification "Protocol Logic", technique "Access Control Exploit", chain Ethereum. That was a discontinued 2019 SoloMargin contract drained through lingering user approvals; the live protocol was not breached, roughly 90% was returned by the exploiter and the treasury covered the remainder. Of 173 hacks the list logs for 2026, none is Dolomite. No incident found between 2026-07-01 and 2026-08-11 in news scans.
• Related party and history: World Liberty Financial's treasury pledged about 5 billion WLFI in April 2026 to borrow roughly $65.4M in USD1 and $10.3M in USDC, about $75.7M in total, pushing USD1 utilization to about 93% and deposit rates above 34%. Repayments of $15M and $10M followed in the first half of April, restoring capacity and normalizing deposit rates to about 10.4%, and left about $162M outstanding backed by about 4.99 billion WLFI. Sources disagree on the exact repayment dates (one account puts them at 7 and 10 April, another describes 25M USD1 repaid as of 11 April), so treat the aggregate as the reliable figure: refresh on the dates. The WLFI collateral balance today, 4,998,327,306 tokens, is essentially unchanged from that April position. Dolomite co-founder Corey Caplan is an adviser to World Liberty Financial (The Block, and WLFI's own announcement); some professional profiles now list him as WLF's CTO, which I could not confirm from a primary source. WLFI Markets is a front end onto Dolomite. Sources: CoinDesk 2026-04-09 and 2026-04-13, Unchained, BanklessTimes, BeInCrypto.
• How much of today's $148.2M USD1 and $106.1M USDC borrowed is that position: refresh.
• Cross-reference: the WLFI report in this repo (apps/showcase/content/research/wlfi.md, researched 2026-08-11, verdict avoid) records the Dolomite USD1 pool "near 93% utilization in the April and May window" and explicitly flags those figures as not re-measured. This run measures it at 54.33%. That line should be updated at the next WLFI refresh.

Maintained monthly. Methodology: DeFi Research Instruction v2.