0.55days of PURR volume to unwind the crowd99%of the XMR book on one side, across 29 vaults$1.30Bunder one key with no guardian, 52 of the 426 vaults whose control we have read$308.7Mwhose control we have never read, across 3,248 vaults$1.97Bsitting in vaults and not lent out$11.30Bunder 3,674 vaults we read$4.81Bin 324 vaults where we saw automated execution, nothing declared2,103vaults declared or on a venue built to run software, holding $984.3k326software-run vaults scored, 13% of them450operators identified behind them26networks where we read capital, from Ethereum to Cardano31,388,200deposits and withdrawals read from chain161changes recorded in the last day594,466ERC-8004 agent registrations read−0.57 down, ERC-8004 agent registrations read/ 30 days364,531distinct owners behind those registrations200,482registrations carrying a readable card, 34%3,337,008paid requests on the Olas rail (deliveries are counted apart)−0.68 down, paid requests on the Olas rail (deliveries are counted apart)/ 30 days87suppliers delivering that work411,349agent-to-agent jobs read on Base+0.07 up, agent-to-agent jobs read on Base/ 30 days$1.0Msettled between agents, all time1,175suppliers listed on the ACP rail189agents live on Cardano76Cardano agents deregistered, 29% of all ever minted29,194escrows between Cardano agents since Jun 2025, 165 still open78%of every agent payment we read comes from 2,636 addresses that never stop309,820addresses paid once and never came back, of 504,405 that ever paid954of those payers are an agent we can name, 0.189% of them642agent hosts answered when we asked, of 2,536 probed17,121agents sit behind a host that answers nothing at all272msmedian answer from the 642 hosts that answer44%of all agent ratings were written by ten addresses, of 676,770 ratings read29,392rated agents have exactly one rater, of 47,149 rated at all91,371service endpoints declared across 36,037 agent cards we read0.55days of PURR volume to unwind the crowd99%of the XMR book on one side, across 29 vaults$1.30Bunder one key with no guardian, 52 of the 426 vaults whose control we have read$308.7Mwhose control we have never read, across 3,248 vaults$1.97Bsitting in vaults and not lent out$11.30Bunder 3,674 vaults we read$4.81Bin 324 vaults where we saw automated execution, nothing declared2,103vaults declared or on a venue built to run software, holding $984.3k326software-run vaults scored, 13% of them450operators identified behind them26networks where we read capital, from Ethereum to Cardano31,388,200deposits and withdrawals read from chain161changes recorded in the last day594,466ERC-8004 agent registrations read−0.57 down, ERC-8004 agent registrations read/ 30 days364,531distinct owners behind those registrations200,482registrations carrying a readable card, 34%3,337,008paid requests on the Olas rail (deliveries are counted apart)−0.68 down, paid requests on the Olas rail (deliveries are counted apart)/ 30 days87suppliers delivering that work411,349agent-to-agent jobs read on Base+0.07 up, agent-to-agent jobs read on Base/ 30 days$1.0Msettled between agents, all time1,175suppliers listed on the ACP rail189agents live on Cardano76Cardano agents deregistered, 29% of all ever minted29,194escrows between Cardano agents since Jun 2025, 165 still open78%of every agent payment we read comes from 2,636 addresses that never stop309,820addresses paid once and never came back, of 504,405 that ever paid954of those payers are an agent we can name, 0.189% of them642agent hosts answered when we asked, of 2,536 probed17,121agents sit behind a host that answers nothing at all272msmedian answer from the 642 hosts that answer44%of all agent ratings were written by ten addresses, of 676,770 ratings read29,392rated agents have exactly one rater, of 47,149 rated at all91,371service endpoints declared across 36,037 agent cards we read
Over 21 daily readings it has ranged from 49.71% to 154.93%, averaging 79.51%. The tile below shows today’s quote alone.
21 readings is 21 days, not a year. We began taking them on 2026-09-19 and this window grows with the record rather than standing in for a longer one.
Taking money
—
TVL
$155.8k
Return
122.9%
over its watched history, read on 135 of 464 days, thin enough that the figure is an estimate between readings
The venue advertises
76.4%
the share price implies 84.6% over 27 days, from 49 readings
Recent
Value and drawdown
This chart shows the last 30 days of a 465 day record; the whole-record figures on this page are measured over all of it, and the 30, 90 and 365-day figures over their own windows, dated below. The rest is the Analyst tier
Range
Hover for values. Upper: vault value, with BTC indexed to the same starting capital. Lower: how far below its own peak it was.
The rate, and the seven things that qualify it
A quoted rate is one input, not a verdict. These are the same seven fields in the same order for every vault, so a high rate and a high rate nobody can withdraw from do not look alike.
Current quoted rate
76.4%As the venue states it, from Hyperliquid. Passed on rather than adopted, and whether rewards are inside it is the next row.
Rewards inside that rate
No reward rate is published, so we cannot say whether the quote includes one.A published rate, not a verified payment. Promotional projections and amounts actually accrued are different things and this is the first.
Realised return
84.6%over 27 daysDerived from the share price we read, over the window stated. Annualisation is secondary: the window is the measurement.
Fee basis
Inside the share priceA return derived from the share price is net of whatever the venue already deducted. Fees charged outside it are not in either number.
Withdrawable now
The venue publishes no withdrawable figure, so no amount is claimed either way.A reading of now, not a promise about the day somebody tries. An accruing rate with nothing withdrawable is the case this section exists for.
Profit, and what the value is made of
Value held less cumulative profit is the money people put in. The venue draws the two apart; the gap is the point.
Range
Profit and loss since the vault opened, as the venue reports it. It ignores deposits and withdrawals: only trading moves this line.
Value heldNet money put in
Range
Solid: what the vault holds. Dashed: what depositors put in, net of withdrawals. Where the two rise together the growth was deposits, not performance.
Nobody has put money in or taken it out across this window, so the money put in is a flat line and the two charts above are the same curve at two heights. That is a fact about this vault rather than a repeated chart: everything its value did, it did by trading.
No strategy has been declared for this vault
Everything above is reconstructed from public data. A declaration is free, signed from this vault’s own key, and cannot be edited afterwards.Note 28
Six questions, and what we hold for each. This is not a score and there is no letter at the end of it: the answers are different kinds of thing and compressing them into one number would hide exactly the part a reader needs. Where we have nothing, the row says so. 2 of 6 are not measured for this vault.
What does it hold or lend into?DeclaredUSDCThe venue names a deposit asset. What the vault actually holds behind it has not been read.
How much could leave now?Not measured—No exit reading. This is the figure that decides whether a position can be left, and we do not have it.
Who runs it, and what else have they run?MeasuredHyperTwinA named operator with a record across every vault they run, including the ones that closed.
What does it depend on?DeclaredHyperliquid on Hyperliquid
That is 5.73% on the capital the strategy currently runs.
Slippage is not separated: without a book snapshot at fill time it cannot be measured, so it remains inside gross rather than being estimated.
Funding is differenced across position snapshots, so it is a flow over the window rather than a level: it counts what changed hands while we were watching and nothing before that. Intervals where a position closed and reopened are refused rather than guessed, which makes the figure a floor.
Paid to trade.Earned by trading, after paying $45 of funding to hold the position.
Computed from 4,726 fills over the 34d in which funding could also be measured, so every part is taken from the same days. That window opens when we started watching, not when the vault started trading. Measured by differencing 10617 snapshot intervals over 34 days. 570 of 11187 intervals were not attributable, a position closed and reopened, or a snapshot was missed, so this is a floor.
Analyst tier
How well it executes
Built from 13,073 individual fills. Two strategies with identical signals can differ by hundreds of basis points purely in how they get filled, taker share, fee rate, slippage. This is the part of a result that is operational competence rather than market view, and it is what tells you whether a good year was skill or a good venue.
· Trades continuously through every 6-hour window, including overnight.
· Activity spread evenly across all 24 hours (evenness 0.984).
· 136.5 fills per day, beyond sustained manual trading.
Overnight activity79%
Hour-of-day evenness98%
Cadence regularity0%
Clock alignment36%
Fills per day136.5
Instruments19
Timing only. This says nothing about whether the strategy is any good, a losing bot is still a bot.
Nearest liquidation, as the venue reports it
LTC rises 7,707.9% to 4,995.87, from 63.9847 now.
Leverage 12.1× weighted by size, 40× on the most levered leg, a liquidation starts at the second number, not the first.
The closest liquidation is 7708% away at the venue's own reported price. Weighted leverage is 12.1×. 18 of 19 positions carry no liquidation price, so the nearest is the nearest we can see.
Current exposure
19 positions · $56.3k gross · 0.4x the vault's capital
Impact is inferred from this strategy’s own realised cost, not measured against a book.
A 10% participation ceiling is applied where venue volume is known.
Trade size is converted to capital using the strategy’s own capital-to-trade-size ratio.
Capacity is capped at 50× the largest trade actually observed, beyond that we are extrapolating.
Measured edge exceeded 100bps per trade and was clipped, above that it is directional luck, not repeatable alpha.
Capped at 50x current capital; beyond that the estimate is unfalsifiable.
At least one instrument hit the observed-size cap, so the total is a floor rather than an estimate.
Instrument
Edge
Taker
Median trade
Limit
NEAR
125.9bps
9%
$128.00
Edge 125.9bps against 2.3bps of realised cost. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
ADA
111.6bps
10%
$106.00
Limited by market share: 10% of 8M daily volume across 7 trades a day. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
ETH
101.7bps
9%
$121.00
Edge 101.7bps against 2.2bps of realised cost. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
Who is in it
Money in and out
Ledger read 10 hours ago
Analyst tier
Who is in this vault, and how much of it is one wallet
A vault that is eighty per cent one depositor is a vault whose price can be reset by a single withdrawal, and whose headline TVL says almost nothing about whether anyone else believes in it. Concentration is the risk a return figure never shows, and it moves before the return line does.
The section above says money moved. This says what the movement WAS. A vault grows because deposits arrived, or because the trading worked, or because it was paid to hold a position, and a size figure renders all three identically.
Money in and outdeposits less withdrawals, and arriving is not earning
$157
Closed tradesrealised on the fills the venue reports
$8,281
Feespaid to the venue on those fills
−$132
Fundingpaid or received for holding the positions, from the venue's own ledger
−$36
Open positions movedchange in unrealised profit on what was still open
−$858
Not explainedwhat the terms above do not account for, 0.0% of the $24,782 that moved
−$4
Change in value$146,268 to $154,056, 2026-09-09 to 2026-10-08
$7,788
Performance, which is the change less what arrived and left, is $7,631 over days and fills.
The shape of the record
277 completed round trips
Rebuilt from the venue’s own open and close labels on every fill. A return figure cannot say whether it came from being right often and small or rarely and large, and those are opposite businesses.
Win rate
86%
Payoff
12.67
Profit factor
75.09
Median hold
25m
277 round trips, 86% of them profitable, holding a median 25 minutes. Average win $74, average loss $6.
What is still missing
Operator cohort. We show what the venue still exposes; strategies deleted before we began observing are invisible, and only pre-registration can make a cohort provably complete.
Gross-to-net waterfall. Returns are venue-reported and not yet decomposed into fees, funding and slippage.
Off-chain exposure. Anything held on a centralised exchange is invisible to us, so the risk shown is a floor.
6.0% · 23.7%
30d · 90d · 30.4% over a year
Market move behind it
5.3%
64.2% fall at 12.1×
Return per book-turn
0.834%
turns 0.21× a day
Max drawdown
64.2%
a floor, read on 135 of 464 days
Sortino
1.26
return per unit of downside risk
Volatility
115%
94% of it downside
History
1.3y
over 70 readings from one source. The record below is longer: 2,484 readings in total (2,484 readings, all taken at the time), and the metrics use one source family rather than mixing them.
Record begins
2026-08-11
began 2025-07-01, so 406 days of it are before anything here
Record rests on
Grade D
95% B, 5% D. A return computed across all of it inherits the weakest, which is D.
Confidence
35%
in these metrics
History quality
49readingsHow many observations the derived figure rests on. A rate from three readings and one from three hundred are not the same claim.
What it depends on
51risk scoreThe rate says nothing about what it rests on. Two vaults paying the same can hold one asset between them.
The venue and the chain are known; what the strategy routes into underneath them has not been read.
Who can change it, and how fast?Not measured—Access control has not been read for this vault. Who can move the money, and how quickly, is unknown. That is not the same as nobody being able to.
How much of the above was measured?Measured3 of 5 not measuredCounted from the rows above rather than asserted. Every figure on this page carries when it was read and what it was read from; the gaps are listed on /reconciliation with their size.
Amber means the evidence is incomplete, never that the vault is risky, and grey means we did not read it rather than that there is nothing to read. Nothing here is a rating, a recommendation, or a statement that this vault is safe.
First observed by us2026-08-11
Statuslive
What is deciding
unregistered
Strategy versionNot attested; the operator has not bound this record to a specific version
Model providerNot attested; no model declared
ModelNot attested; no model declared
System prompt hashNot attested; no prompt commitment
Policy hashNot attested; no policy commitment
Inferred modeautonomous from execution timing, 3,000 executions over 22 days; it shows automated execution, not whether a model or fixed rules decide
These fields are empty because nobody has attested them, not because we failed to find them. An operator who registers gets a record that provably cannot be edited after the fact, and a grade no unregistered strategy can reach.
How much to trust this
Provenance gradeBReal capital, readable at the venue, but the operator’s full cohort is unproven.
Machine-managedyes (execution timing we measured)
Evidence
behavioural Execution timing indicates automation (autonomous): Trades continuously through every 6-hour window, including overnight.
Metric confidence35%70 observations over 464 days
RankingNot ranked, insufficient history
Alignment and capacity
Operator’s own share14.2%
Commission10%
Depositor wallets100+
Concentration≥ 0.46 · largest holder 64.6% of the vault
3 of 19 shown in full. 12 more are listed with the instrument withheld.The other 4 are not on this page at all. A plan opens every instrument, the exact notional and side, entry price, leverage, distance to liquidation, and how the book changed since the last reading. The count is free because the count is the finding; the detail is what Analyst carries.
Snapshot taken 2026-10-10 11:20 UTC. A strategy that flips intraday can look flat at every snapshot and still carry risk in between. Click an instrument to see every other vault holding it.
UNI
112.8bps
10%
$147.00
Limited by market share: 10% of 12M daily volume across 14 trades a day.
AAVE
113.7bps
4%
$87.00
Limited by market share: 10% of 8M daily volume across 10 trades a day. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
DOGE
103.5bps
12%
$93.00
Limited by market share: 10% of 6M daily volume across 4 trades a day. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
LTC
61.1bps
21%
$259.00
Limited by market share: 10% of 3M daily volume across 5 trades a day.
XRP
94.3bps
3%
$102.00
Edge 94.3bps against 1.8bps of realised cost. Estimate limited by observed trade sizes, treat as a floor, not a ceiling.
27
3,553
44 further days are not included, because a reader has not covered them: 25 days because fills have been walked, but not over these days; 25 days because the funding ledger has been walked, but not over these days; 1 day because this vault’s fills have never been walked.
Funding, summed forward
Range
What the vault was paid to hold its positions, less what it paid, a day at a time. A line that falls is a book that pays to stay where it is.