November 21, 2026

On 20 November, XMR1 moved from an account identified by Wagyu as its deployer to an account that Bitquery identifies as the seller behind the token’s collapse. XMR1 represents Monero within Hyperliquid’s trading system; it is not Monero’s native coin. This investigation asks what the public records establish about the transfers and later ETH movements, and where the explanation still depends on someone else’s account.
The extended route is XMR1 sales on HyperCore → USDC cross-chain withdrawals through HyperEVM → USDC received on Arbitrum. A selected Arbitrum swap and a separate Arbitrum-to-Ethereum bridge transaction show how parts of that account’s activity changed form. On Ethereum, four transfers reached one recipient. These segments are supported at different levels: the withdrawal messages can be joined by their request identifiers, while only one later bridge pair has had its message content reconciled here. Money passing through the same account is not a claim that each individual unit has been followed through every trade.
Scope | Basis |
Incident date | 20 November 2026; times below are UTC |
Direct explorer checks | Three native XMR1 actions; 4,574 incident-window sell fills; 11 Core withdrawal / Arbitrum mint pairs; one CoW Trade event; one later bridge pair; six original Ethereum anchors |
Calculation | 100,000 + 100,000 + 5,000,000 = 5,200,000 XMR1 |
Sales scope and external tally | Bitquery reported about $13.51 million and 420 buyers. The fill API independently yields 13,507,983.519996 USDC gross sales in the stated incident window; buyer count remains unverified. |
Evidence retrieval cutoff | 2026-11-21T01:43:03Z; historical records, not live balances |
Observation boundary | Defined fill and withdrawal windows; one swap and one later bridge pair sampled. Other bridge routes and refunds are not reconciled. |
Finding | Meaning |
Sales are measurable from the seller’s own fill records | 4,574 distinct incident-window sell fills total 369,622.92 XMR1. Gross proceeds and fees are accounted separately. |
All 11 withdrawal messages carry the Core request identifiers | Recipient, amount and embedded Core nonce join each Core action to an Arbitrum USDC mint; the fee is 0.2 USDC per pair. |
One later Arbitrum–Ethereum pair matches beyond amount and time | The immutable burn-message fields and hook data agree. This validates the selected pair without validating every bridge or the hosting explanation. |
The deployer role means the account associated with creating the token, rather than a separate Ethereum token contract. A spot balance is a balance used for trading the token. The native action records inspected here say sendAsset, with both sourceDex and destinationDex set to spot. Each identifies the same destination and the XMR1 token identifier. The field signatureChainId is not a reason to call these Ethereum transfers: the records are from the Hyperliquid explorer and say hyperliquidChain: Mainnet.
At 04:46:58, the first record specifies 100,000 XMR1. A second 100,000 follows at 05:14:30, then 5,000,000 at 05:16:10. The first-to-last interval is 29 minutes 12 seconds. That measures the span of these three records, not the duration of the intrusion or all subsequent selling. The transaction appendix preserves their hashes and blocks.
20 November UTC | Recorded XMR1 amount |
04:46:58 | 100000 |
05:14:30 | 100000 |
05:16:10 | 5000000 |
Start Figure 1 with the deployer account at the top. Follow each solid arrow through its amount panel to the common recipient. The parallel panels make the three records visible without implying that they occurred at equal intervals.

Figure 1. Solid arrows summarize the three native records. Position and line width are schematic. The explorer view exposes the action fields but no separate execution receipt; this report describes the recorded actions and does not invent a success flag. Sources: the three native transaction links in Appendix A.
The key observation is concentration: the same recipient is specified in every record. This is not proof of the person operating that account. Nor does the transfer quantity measure how much real Monero backed the tokens, or the amount buyers later paid for them. Those questions require different records.
A fill is one executed piece of an order. One order can match many resting orders, so a fill count is not a buyer count. The official API identifies the XMR1/USDC market as @260, linked to token index 404 and the same XMR1 identifier used in the native actions. We queried the seller through 13:00 UTC on 20 November and kept XMR1 sells after the first deployer transfer at 04:46:58.926 UTC. Hyperliquid’s official fill API documentation.
The retrieved records contain 4,574 distinct sell fills in that interval. Their first timestamp is 04:49:32.170 and their last is 10:28:02.058 UTC. Multiplying each fill’s XMR1 quantity by its USDC price gives gross proceeds of 13,507,983.519996 USDC. Fees in those records total 8,110.80642437 USDC; subtracting them gives 13,499,872.71357163 USDC. These are trade-accounting quantities, not fiat withdrawals, a live balance, or a separate loss to add to the later transfers.
Incident-window fill measure | Result |
Sold XMR1 | 369,622.92 |
Distinct sell fills | 4,574 |
Gross USDC / reported USDC fees | 13,507,983.519996 / 8,110.80642437 |
Gross less those fees | 13,499,872.71357163 USDC |
Exclusion | Two small buys and two small sells before the first deployer transfer; other markets excluded |
Three overlapping API pages were deduplicated by market, trade ID, order ID, action hash and side. The final page held 702 records, below the 2,000-record response cap. The API exposes only the latest 10,000 fills, which remains a retrieval limit. It does not supply counterparties here, so we have not independently counted Bitquery’s reported 420 buyers. Appendix C gives the precise query and first/last fill anchors.
Yes, for the 11 requests examined here. Hyperliquid’s native explorer calls each action sendToEvmWithData. Its destinationRecipient is the seller’s address, and destinationChainId is 3. This number is a CCTP domain identifier for Arbitrum, not Ethereum chain ID 3. The Arbitrum receiving transactions identify source domain 19, which Circle assigns to HyperEVM. Circle’s domain identifiers.
The decisive link is inside the receiving message. Its hook data—a small payload that accompanies the transfer—preserves the full Core sender address and the 64-bit Core request nonce. A nonce is an identifier for the request; it is not the block timestamp. In every pair, that value matches the native action’s nonce, while the recipient and the USDC amount agree. Circle publishes the byte positions used for this check in its hook library. Circle’s published withdrawal-hook layout.
Withdrawal reconciliation | Observed result |
Core requests / Arbitrum receiving transactions | 11 / 11; each request joined by its embedded sender and nonce |
Core-side amount, normalized to six USDC decimals | 13,281,001.725471 USDC |
Arbitrum amount minted to the seller | 13,280,999.525471 USDC |
Forwarding fees in the receiving events | 11 × 0.2 = 2.2 USDC |
Separate Core ledger fees | 0.023419 USDC; not added to the destination mint amount |
The first pair makes the difference concrete. The Core request at 04:54:40.128 carries nonce 1795150467235 and 7,474,206.546009 USDC. At 04:54:48, Arbitrum records a successful mint of 7,474,206.346009 USDC and a 0.2 USDC fee; the received hook contains the same sender and nonce. These are linked request and receipt records, rather than a match inferred merely from eight seconds of elapsed time. First Core request · First Arbitrum receipt.
All 11 native actions and destination events are preserved in Appendix C. This establishes the cross-chain withdrawal linkage, but the intermediate HyperEVM transaction hashes were not individually fetched.
Nor does a match establish that every USDC unit came only from XMR1 sales: the account also has activity outside that market. The API’s tiny decimal tail on one large ledger amount is retained in the research evidence; public USDC comparisons use six decimal places, as do the receiving token’s raw units. official non-funding ledger API documentation.
At 04:57:53 UTC, an Arbitrum CoW settlement records the seller as the owner of a Trade event. The event sells 1,000,000 USDC and records 578.374894447965203136 ETH as the buy amount. CoW is a decentralized trade-settlement service here, not evidence of a custodial exchange deposit or a sale for bank money. Selected settlement and logs.
Selected Trade event | Exact record |
Block / event index | 525826782 / 2 |
Sell amount / buy amount | 1,000,000 USDC / 578.374894447965203136 ETH |
Event feeAmount | 0; this field does not account for every service cost or gas fee |
Verification boundary | Trade event inspected; independent internal ETH payment trace unavailable in the explorer response |
This closes one conversion at the event level. It is not a reconciliation of all ten settlements in the earlier external analysis, and the seller’s shared balance prevents assigning this one swap’s exact ETH to the later 1,013 ETH bridge input. Keeping that distinction avoids presenting a continuous line where only selected transactions have been checked.
At 05:03:54, the selected Arbitrum transaction takes 1,013.043520159326998830 ETH and executes swaps before a CCTP burn of 1,733,501.731759 USDC. CCTP transfers native USDC by destroying it on the source chain and authorizing issuance on the destination chain. The Ethereum transaction at 05:04:11 records the corresponding received message, mints 1,733,259.041517 USDC for the payout contract and ultimately pays the seller 1,000.344443544774143612 ETH. The two ETH amounts differ because this route includes conversions and charges; it is not a transfer of an unchanged ETH quantity.
Selected bridge stage | Quantity / evidence |
Arbitrum input | 1,013.043520159326998830 ETH · source transaction, block 525,828,140 |
USDC burned / minted | 1,733,501.731759 / 1,733,259.041517 USDC |
CCTP executed fee | 242.690242 USDC; burned amount − minted amount |
Ethereum payout | 1,000.344443544774143612 ETH · internal transfer, block 26,454,094 |
The connection can now be tested against message content. The source MessageSent and destination MessageReceived agree on domain 3 → 0, the source messenger, burn token, mint recipient, amount, original message sender, maximum fee and complete hook data. The hook also contains the seller as the intended recipient of the final payout. This is a substantially stronger association than the old explorer link or an amount-and-time guess. Source: Arbitrum logs · Ethereum logs.
In Figure 2, begin at the Arbitrum ETH input and follow the stages downward. Solid lines describe stages within a transaction; the dashed connector marks the cross-chain association established by matching the immutable message fields and hook. The figure omits intermediate swap pools and secondary service payments, so it is a schematic for this one pair, not the whole incident.

Figure 2. ETH quantities are rounded to six decimals in the image; the table keeps full precision. The USDC difference of 242.690242 is the CCTP executed fee, not a second loss. Other route fees and conversions are not bundled into that number. The association applies to the two transaction hashes in Appendix D.
Two protocol details limit what “matching” means. In CCTP V2, the source emits an empty nonce before attestation; the destination carries an assigned 32-byte nonce. Their equality is therefore not a valid test. The received fee and expiration fields are also populated later. We compared the fields that should remain unchanged, rather than declaring the entire byte string identical. This does not independently revalidate Circle’s off-chain signatures or the other six bridge pairs. Circle’s published burn-message layout.
Before the first XMR1 action, an Ethereum transaction at 03:30:23 includes an internal transfer of 0.0978829303948 ETH to the examined address. An internal transfer is a value movement made during contract execution; it differs from the transaction’s outer sender. Etherscan labels its source as a Tornado Cash pool. The transaction was successful, but the label does not identify the person who funded it. Funding transaction.
At 05:04:11, another successful transaction paid 1,000.344443544774143612 ETH internally to the examined address through a contract labeled Mayan: Fast MCTP by Etherscan. The explorer’s overview links this Ethereum transaction to the selected Arbitrum source hash. The preceding section reconciles this receipt against the selected Arbitrum source’s message fields and hook data; it is no longer based only on the explorer’s cross-chain annotation. Ethereum payout record.
Four successful direct transfers then sent ETH from the examined address to the same Ethereum recipient. The first was 10 ETH at 05:05:35, followed by the largest, 2,325.669594723330818716 ETH at 05:06:11; then 1,500 ETH at 05:10:23 and 1,000 ETH at 05:18:59. Together they total 4,835.669594723330818716 ETH, excluding transaction fees. These are four distinct transaction hashes, counted once each. The total is a redistribution amount, not an additional loss or verified cash-out.
20 November UTC | Direct transfer to the common recipient |
05:05:35 | 10 ETH · transaction |
05:06:11 | 2325.669594723330818716 ETH · transaction |
05:10:23 | 1500 ETH · transaction |
05:18:59 | 1000 ETH · transaction |
Read Figure 3 from the two upper receipt panels to the shared Ethereum account, then to the recipient below. Solid arrows summarize independently inspected Ethereum value movements. The two incoming observations do not account for the whole outgoing amount; this is deliberately not a complete balance reconciliation.

Figure 3. Diagram quantities are rounded where marked. Exact values and full transaction hashes are in the tables and Appendix B. The upper-right receipt is independently visible on Ethereum; its selected Arbitrum source is reconciled in Figure 2 and Appendix D. Figure 3 stays within Ethereum, so it does not draw the intermediate cross-chain stages again.
The order matters. Ethereum redistribution began after the first XMR1 action and before the second and third. The final ETH transfer followed the last native action. A narrative that first finishes the entire token outflow and only then begins the money trail would conceal this overlap. Ordering the events still does not fill every sales, swap and bridge link.
Bitquery’s original investigation remains the incident lead and source for its broader sales, buyer and route tallies. The new calculations above replace reliance on its rounded sales total for this defined window; they do not independently validate every broader tally or any actor attribution.
Wagyu representative’s post-mortem says the deployer key was in files retained from an old server that had been deleted. He also promised personal refunds and a later relaunch. These are project statements. They do not establish which files were taken, the exact access path, the intruder’s identity, or that compensation has been completed.
DV’s reproduction of a joint 1984 ehf. and CERT-IS notice separately describes a serious attack on the host and a temporary suspension of services. It supports the existence of a hosting incident. The notice does not itself identify Wagyu’s deployer key or prove that this key was extracted through that incident.
There are several explanations the chain records alone cannot distinguish: exposure through retained server data, compromise through a different copy of the key, or another form of authorized-key misuse. No ranking of those alternatives is claimed here. Server access logs, retention records and evidence connecting a particular key copy to the unauthorized signing activity would narrow the cause.
Question | Record needed next |
What could the old key still control? | An incident-time inventory of token balances, account authority and any replaced signing permissions. |
Which buyers were compensated? | Authenticated refund transactions matched to eligible recipients and amounts; a promise alone is insufficient. |
Which bridge movements belong to this route? | Replicate the other six bridge pairings with source events, immutable message fields, destination receipts and fee reconciliation; one matched pair is not enough for the whole route. |
Retiring infrastructure and retiring signing authority are different operations. If the key remains able to authorize value movement, removing the server does not make surviving copies harmless. That is a general security interpretation of the project’s explanation, not proof of how the incident occurred. For investigators, the practical next step is a dated authority inventory alongside the transfer record, rather than an expanding graph of loosely related wallets.
Full wallet strings below link to address pages on the relevant chain. Roles describe the records used here and do not identify their controllers. The deployer role is attributed to Wagyu and Bitquery; the explorer independently confirms the account’s user field in the three actions.
Role and chain | Full address |
Reported deployer · Hyperliquid | |
XMR1 recipient · Hyperliquid | |
Examined account · Ethereum | |
Large-transfer recipient · Ethereum | |
Funding pool · Ethereum (explorer label) | |
Payout contract · Ethereum (explorer label) | |
Examined account · Arbitrum | |
Cross-chain withdrawal contract · HyperEVM (message field) | |
USDC token contract · Arbitrum | |
USDC burn-token contract · HyperEVM |
Record 1 | Native Hyperliquid |
Time / block | 2026-11-20T04:46:58Z / 1227836280 |
Asset / quantity | XMR1:0xbb2057659bd378d17e2e151bb04bdcaa / 100000 |
Full hash | 0x328fc4efdff6c881340904492f4b780201d700d57af9e753d65870429efaa26b |
Record 2 | Native Hyperliquid |
Time / block | 2026-11-20T05:14:30Z / 1227859502 |
Asset / quantity | XMR1:0xbb2057659bd378d17e2e151bb04bdcaa / 100000 |
Full hash | 0xfe18b36928ccc7ecff9204492fa62e0208de004ec3cfe6bfa1e15ebbe7c0a1d7 |
Record 3 | Native Hyperliquid |
Time / block | 2026-11-20T05:16:10Z / 1227860884 |
Asset / quantity | XMR1:0xbb2057659bd378d17e2e151bb04bdcaa / 5000000 |
Full hash | 0x8453d81d6945e81885cd04492fab940204a80003044906ea281c83702849c203 |
Field | Ethereum transaction |
Time / block | 2026-11-20T03:30:23Z / 26453625 |
Asset / quantity | 0.0978829303948 ETH |
Status | Success |
Full hash | 0x33513047a342ea41bbcc6503aa8c889a8b5d05a4ef332354fba682fdc60b53ee |
Field | Ethereum transaction |
Time / block | 2026-11-20T05:06:11Z / 26454104 |
Asset / quantity | 2325.669594723330818716 ETH |
Status | Success |
Full hash | 0x74a704c0b00bd9cb6c671e339f361718c7d885d5776d13074ecb0ed17edf4752 |
Field | Ethereum transaction |
Time / block | 2026-11-20T05:05:35Z / 26454101 |
Asset / quantity | 10 ETH |
Status | Success |
Full hash | 0x43e61634dd6a8c01e8fc9fad0d04b56ec602f42f4a37e3080d32d6f69e1d892a |
Field | Ethereum transaction |
Time / block | 2026-11-20T05:10:23Z / 26454125 |
Asset / quantity | 1500 ETH |
Status | Success |
Full hash | 0xf5820bca522aeff14766a66ad939a25d447837f544371e6fff2597804c401b62 |
Field | Ethereum transaction |
Time / block | 2026-11-20T05:18:59Z / 26454168 |
Asset / quantity | 1000 ETH |
Status | Success |
Full hash | 0x9380c8291ae4e4dfad92a704a6bb8f97d6f70d5b97edff6b887b318ffd05cd66 |
Field | Ethereum transaction |
Time / block | 2026-11-20T05:04:11Z / 26454094 |
Asset / quantity | 1000.344443544774143612 ETH |
Status | Success |
Full hash | 0x3a93c0cb9085ed6d73d6f0e2a8bcfafc942d8d4af7b42b9d34abc10576b07240 |
All times in this appendix are UTC on 20 November 2026. The full fill dataset was retrieved and deduplicated; the two fill records below are selected anchors, not a list of every counterparty fill. The withdrawal tables include every one of the 11 matched requests and receipts.
Reproduce the public API query | Value |
Endpoint / method | https://api.hyperliquid.xyz/info · POST (read-only query) |
Fill request | type=userFillsByTime; user=examined address; startTime=1795132800000; endTime=1795179600000; aggregateByTime=false. Subsequent pages restart at the preceding page’s last timestamp. |
Incident filter | coin=@260; side=A; time ≥1795150018926; feeToken=USDC. @260 maps to XMR1 token index404 in spotMeta. |
Ledger request | type=userNonFundingLedgerUpdates; same user and time window. |
Selected fill | Record |
Role / UTC | First incident-window sell fill / 2026-11-20T04:49:32.170000+00:00 |
Quantity × price | 4.06 XMR1 × 492.31 USDC |
Trade ID / order ID | 19548947811789 / 611017042039 |
Full action hash | 0xa796e1e81e2ce7b1a91004492f53fe02039400cdb92006834b5f8d3add20c19c |
Selected fill | Record |
Role / UTC | Last retrieved incident-window sell fill / 2026-11-20T10:28:02.058000+00:00 |
Quantity × price | 19.94 XMR1 × 1.0 USDC |
Trade ID / order ID | 818168203372803 / 611038769808 |
Full action hash | 0x5eb0cd98e63bd136602a044933a73b0203cc007e813ef008027978eba53fab21 |
Pair number | Received CCTP nonce (32 bytes) |
#1 | 0x7d1659ef4b401372a15fd949296187348ab26d6c1900dcaa2abf1e79856a10dc |
#2 | 0xeac90c639eb279ebecf76350dc7ecd5bc10bf79c1e9fc9b43478c2ce4fc0d1da |
#3 | 0x0589ce7b37ceff47268451844b5f67a6557b071124f0712dd88c857f70269068 |
#4 | 0x416a64b349951f18fd8547a7139145456217e4515b2518b314bb7d3013805640 |
#5 | 0x23f7707db8fa7ce305e5ab57486e615265becd0474d5bf9fe89f5a963ba47097 |
#6 | 0x51183006f3aa1ac7934a451a9827b5181f6832f3b47273ab70efc120eea01d5f |
#7 | 0x57e963e09b86be85fb22937e64464f5a5c4b567c3e15d5aed784514d5b93c934 |
#8 | 0x86d7974c271a8fc70c1433da04a4d8b636d057ae581eaa5410f12d2aae4b0595 |
#9 | 0xb6b6f63b9d5b7b3a0176750d2fe5fafce3744f8bd1600da3c6d55f178319bcb9 |
#10 | 0x58c958bb78dd57acdbf1926d3c2859f169615d928ded2f1f69b3c138bcac88f1 |
#11 | 0x088b7b62d38f575dd5559c45dd6ecb858e4ca2c7bf8151b37366e447259a9ad9 |
Record | Full hash / identifier |
Arbitrum CoW settlement | 0x50d4490db13e0f7fdc3214b61ed4fb28dd172f437c794fe4265346e5fabb6b73 |
CoW order UID | 0x2e1dd545358ba4fc398eacf39583dab775b83804b78052ce1b975aa56f3e867a1464dacaceb0d511f9964cdac5eba50ea42220bf6affda41 |
Arbitrum bridge source | 0x1b6f68247603d6c3d1754680d9f73aee87d1791a8e51ec92bea1dfa58a47733e |
Ethereum bridge destination | 0x3a93c0cb9085ed6d73d6f0e2a8bcfafc942d8d4af7b42b9d34abc10576b07240 |
Source MessageSent index / destination MessageReceived index | 225 / 251 |
Destination CCTP nonce | 0xf458e998e40533c8519a95bf1ac3ba3a243513db0444c92a82206490b6dca56b |
Raw-message comparison | Source nonce zero before attestation. Immutable body prefix and hook match; populated destination fee/expiry fields excluded from the equality test. |
Method: direct native explorer inspection, official Hyperliquid read-only fill and ledger API queries, public Arbitrum Blockscout transaction/log data and independent Ethereum RPC receipt comparison. Decimal arithmetic was used for quantities and fees. The selected bridge’s source MessageSent bytes and destination MessageReceived bytes were decoded against Circle’s published layouts. The withdrawal hook sender and Core nonce were compared across all 11 pairs. Source-code layouts describe how to read the observed bytes; current code is not treated as a complete historical deployment audit. Broader SentinelTX results were not substituted for the missing primary records. This update does not establish all swap/bridge matches, all intermediate HyperEVM transaction hashes, complete balances, native Monero reserves, completed refunds or actor identity.
13 reads