Infinit
Infinit publishes the
APEX MultiStrategy vault,
managed by Apex Quant Capital, identified through
Infinit’s manager links. Its curator metadata slug is apex-quants.
Its deterministic relative-strength strategy targets matched
long and short Hyperliquid perpetual positions. The integration assigns
Dangerous technical risk and supports data reads only. Transaction requests
raise VaultFlowUnavailable and public transaction capability is unavailable.
This is independent of the contract’s deposit window and redemption status.
The terms publish 2% annual management and 20% performance fees. The reviewed manager and vault implementations contain no fee accrual or high-water-mark accounting; collection would require external operator accounting or settlement. Whether published unit prices reserve fees remains unknown, so investor net returns are unavailable. Deposits closed 30 July 2026 at 10:00 UTC; maturity is 28 October 2026 at 10:00 UTC. Redemption also requires settlement.
Reviewed identities
All EVM contracts are on Arbitrum One. The database identity is the actual vault, not its separately issued receipt token or manager.
Role |
Address |
|---|---|
Vault |
0xe64655e24b66073c73300aa924cd6df5c716638f |
inUSDC receipt |
0xb5fb43efd9b607c6ca863a56d22c491c82275c08 |
Manager (vault ID 1) |
0xafe480f375ebd13df703ef50b429357d29d162ee |
Custody Safe |
0x79e78a5a07f09c0f59f21d4b9dcc18626a0f164d |
Funding/trading |
0xf2460f28c9bacb35ac6fef048b661d33ad070aaf |
USDC |
0xaf88d065e77c8cc2239327c5edb3a432268e5831 |
The application UUID is 019f5a8e-62cf-752c-9e2f-8640768c7d3d.
Vault creation is block 483358787, transaction
0xbef5388cf1f6da875e50693ae33632a5d2ad3c366dd6ff802bf7d6d36cf53506.
The legacy Infinit Multi-Strategy vault
at 0x8220ed9e3d84da5e9d5f6d115a28a0b28b967388 is a separate native
Hyperliquid product. The trading account temporarily invested in it between
3 and 6 August 2026, as established by the ledger entries below. The withdrawal
returned funds to the trading account on Hyperliquid, not to Arbitrum or
investors. These products retain separate identities and performance histories;
the legacy vault’s returns and fees do not establish this product’s investor
returns or fee terms.
The audit identifies Saprolings/vault-contracts
and review commit d1fd00f8. The repository returned 404 during review on
26 October 2026. The report’s post-fix commit is unspecified. We cannot claim
that the deployed implementation matches that review. Verified implementation
interfaces and explorer references are in eth_defi/abi/infinit/README.md.
Observed fund movements
The following chart reconstructs completed transactions, checked on 26 October 2026. Its evidence is Arbitrum transaction receipts, an actual deposit call trace, indexed USDC and receipt-token events, and the Hyperliquid account ledger. The flows are established from those records rather than protocol documentation. The Arbitrum event scan covers vault deployment at block 483358787 through block 518462154, inclusive.
Observed Infinit funding and collateral movements. Solid arrows represent USDC; dotted arrows represent contract calls or receipt issuance.
The 28 deposits from 22 addresses, between 16 and 26 July 2026, total 506,137.385298 USDC. Every manager deposit event matches both a USDC transfer from that depositor to the custody Safe and an equal inUSDC mint to the depositor in the same transaction. The traced 205,423.352859 USDC deposit on 17 July shows the manager calling the vault, which calls the receipt token to mint.
On 3 August, the Safe sent
10 USDC
and then
506,127.385398 USDC
to 0xf2460f28c9bacb35ac6fef048b661d33ad070aaf on Arbitrum. The successful
receipts show Safe execTransaction calls. That destination is an EOA on
Arbitrum and also identifies the separate Hyperliquid trading account.
The EOA’s outgoing Arbitrum transfers are not governed by the Safe’s signature
threshold. Hyperliquid account permissions operate separately; the EOA’s
key-management arrangements have not been verified. The observed transactions
do not establish a completed bridge withdrawal back to Arbitrum.
The funding wallet subsequently transferred
5 USDC,
10 USDC
and
506,127.39 USDC
to the bridge at 0x2df1c51e09aecf9cacb7bc98cb1742757f163df7.
The Hyperliquid ledger records these exact transaction hashes as deposits into
the same-address account, establishing the link between the two ledgers.
Within Hyperliquid, the ledger records:
A 5 USDC test deposit into the legacy native vault at 06:31 UTC on 3 August, followed by 506,137.39 USDC at 08:47 UTC. The latter action hash is
0x45d1490ffbca8785474b04415876d302019500f596cda657e999f462bace616f.A withdrawal of 505,859.98 USDC from that native vault at 09:34 UTC on 6 August, action hash
0x11e2e61a19026435135c044190654802018000ffb4058307b5ab916cd8063e1f.Subsequent collateral transfers within the trading account: 643,058 USDC cumulatively from native to
xyz, starting 10 August, and 301,500 USDC cumulatively back fromxyzto native, starting 18 September. These are sums of repeated movements, not asset valuations.
Bridge and native-vault totals include a separate 5 USDC test deposit. Minor auxiliary funding and dust are omitted from the chart; the underlying amounts are retained in the evidence snapshots. The scan found no USDC transfers involving the actual vault contract, no receipt burns and no manager withdrawal events. A later RPC observation at block 518462681 also returned zero USDC in the vault. No completed return-to-vault redemption leg is shown.
Download the editable Mermaid source and the
dated public evidence snapshots:
Prices and evidence
The canonical price is the application’s public performance history,
PriceSource.api. Read-only Next.js action identifiers are discovered from
the current build’s public JavaScript. A changed or malformed response aborts
the read. This is an undocumented application interface and may change.
Original fifteen-minute observations and compressed raw responses remain in
unconstrained infinit_* DuckDB tables. Hourly exports retain the first actual
observation in each hour, with its original timestamp. Current metadata allows
up to thirty minutes of price age, including one interval of publication lag;
older prices produce unavailable NAV. Hypersync mint/burn logs reconstruct six-decimal inUSDC supply,
checked against onchain total supply before commit. Asset values require both
a valid source price and matching supply. Native and xyz Hyperliquid
marginSummary.accountValue are used only for current valuation validation.
These inclusive UTC source intervals are quarantined:
30 July 2026 10:00 through 4 August 2026 01:45.
6 August 2026 09:30 through 27 August 2026 10:30.
No forward filling, interpolation or return calculation may bridge those gaps. Their raw source values remain evidence. Replacing them requires an explicit reviewed repair. Shortened or failed API responses cannot erase older context. Future manager vault IDs are logged as unreviewed; they do not inherit Apex’s fees, identity or trading account.
Recurring all-chain scanning
scripts/erc-4626/scan-vaults-all-chains.py includes Infinit in the Arbitrum
cycle when SCAN_PRICES=true. There is no separate Infinit switch.
SCAN_HYPERCORE controls native Hyperliquid vaults; SCAN_APEX controls
the unrelated ApeX exchange integration.
Like the existing GMX and Antarctic integrations, InfinitVault implements
VaultBase and supplies a contextual historical reader. Discovery seeds
the reviewed deployment and classifies it by chain and address, without
depending on ERC-4626 events or valuation methods. Future deployments require
their own reviewed product mapping.
After the initial migration, each price cycle extends the already bootstrapped dense timestamp cache, fetches public prices and indexed supply, and validates that supply against RPC. The shared writer then refreshes only Infinit’s retained hourly history. This small history is replayed independently of other Arbitrum readers so delayed API observations and retries cannot fall behind their scan position. Other vaults’ prices and reader states are preserved. Before replacing prices, the scanner requires the context to cover every existing Infinit source timestamp in the replacement range. If the context is lost and the API has shortened its history, restore the context backup; the scanner refuses to erase those older prices. A failed source read withholds Infinit from publication and logs the reason. Other Arbitrum readers and settlement scans still run. The chain dashboard displays an Infinit warning without imposing whole-chain RPC backoff. Missing source coverage requires operator restoration and is logged as an error. The shared fourteen-day source freshness audit includes Infinit’s actual observations; rewriting old rows does not make them fresh. The separate current Hyperliquid equity check is diagnostic and never replaces historical API prices.
The address-scoped refresh uses the shared atomic Parquet writer, so it incurs an additional read and rewrite of the shared price file. It deliberately does not advance adaptive ERC-4626 reader state; its event cursor lives in the context database. Source observations retain their actual dates, and stale prices are not advanced to the current time.
Migration
Run scripts/infinit/migrate-infinit.py once to recover discovery provenance,
refresh metadata and backfill Infinit history. It also updates cleaned history
when that dataset exists. DRY_RUN is the only operation switch: it defaults
to true and performs the complete operation on temporary copies; false
publishes validated outputs with unique backups under the pipeline lock.
Unrelated vaults, reader state and discovery cursors are preserved.
Local use:
source .local-test.env
DRY_RUN=true poetry run python scripts/infinit/migrate-infinit.py
DRY_RUN=false poetry run python scripts/infinit/migrate-infinit.py
For an isolated rehearsal, set PIPELINE_DATA_DIR to copied pipeline datasets
and TIMESTAMP_CACHE_PATH to a directory containing a copy of
42161-timestamps.duckdb before running those same commands. Existing
file-path overrides remain available. The script reads the dense Arbitrum
timestamp cache without changing it and automatically limits history to its
contiguous cached range after deployment, bounded by the safe RPC head. Restore
that cache first if it does not cover deployment. Cache preparation uses
fetch_block_timestamps_using_hypersync_cached().
There are no stage, cleaning or block-range switches to configure.
Production use, after deploying the intended image and stopping conflicting
scanner writes, from ~/vault-scanner/web3-ethereum-defi:
source ~/vault-scanner/vault-rpc.env
docker compose stop vault-scanner-looped
docker compose run --rm -e DRY_RUN=true --entrypoint poetry vault-scanner-oneshot run python scripts/infinit/migrate-infinit.py
docker compose run --rm -e DRY_RUN=false --entrypoint poetry vault-scanner-oneshot run python scripts/infinit/migrate-infinit.py
Review the dry-run summary before apply, then resume the scanner with
docker compose up -d vault-scanner-looped. This uses the
service’s normal persistent mounts and overrides its all-chain entrypoint.
Do not change HOME or remove reader state. Publication replaces each file
atomically; an interrupted multi-file publication can require restoring backups.
Unreadable Parquet, incomplete source coverage and uncheckpointed context WALs
abort publication.
Real integration checks
source .local-test.env
timeout 180s poetry run pytest tests/erc_4626/vault_protocol/test_infinit.py -q --override-ini='addopts='
INFINIT_LIVE_TEST=true timeout 180s poetry run pytest tests/erc_4626/vault_protocol/test_infinit_live.py -q --override-ini='addopts='
The second command checks the actual public app and current account valuation. The migration exercises the authenticated configured Hypersync provider. No credentials are included in output.