Skip to content
PROTOCOL RESEARCH

3Jane risk

The TVL crash our stale-verdict net flagged was not the fund shrinking: USD3's totalAssets is flat at 75,042,961.85 USDC and the share price rose 0.29% in 16 days, so there is no credit event. What collapsed is the exit: the vault's own maxWithdraw returns 3,559,316 USDC identically for all three large holders, 4.74% of the pool against 20.28% on 19 July, deployment is 95.26%, deposits are shut on-chain at the 75M cap, and 90.74% of assets are now off-chain paper and bank cash with Slope at 90.93% of the facility book and also acting as servicer. Three of the previous draft's darker claims did not survive checking and are corrected here: the Slope delinquency tape did render and is clean at 96.94% current with zero charge-offs; the Morpho PT loop's borrow rate fell to 6.46% rather than rising to 10.10%, so carry widened rather than compressed; and the honest exit path is about 5.0M, not 9.2M. Against that, oracle risk is worse than described and now verified rather than carried: the Morpho USD3/USDC market values 17.7M of collateral using the vault's own share price with all four external feeds set to the zero address, and the PT market's oracle is a pure 30%/yr time clock that cannot see credit at all.
CautionResearched Aug 4, 2026
watch: Audits & contractsRepo unchanged this run at five PDFs with pushed_at 2026-06-15, and the count is thinner than it reads: Electisec and yAudit are one organisation (yAudit, rebranded Electisec, reverted November 2025), so five files represent three independent firms, Veridise, Sherlock and yAudit/Electisec. Sherlock's engagement is independently listed on its own contest index. None of the five covers the SPV, the bank rail, the servicer or the underwriting where 90.74% of the money now sits, and the Levered Callable Capital primitive due to launch in August is still described as in audit with nothing published.
weak: Admin controlRe-read on-chain today and the wiring is good and now fully verified, including the roles the previous draft left open: Safe v1.3.0 threshold 3 of 5, timelock getMinDelay 86400 and owner() of MorphoCredit, ProtocolConfig and CreditLine, Safe holds PROPOSER, EXECUTOR and CANCELLER but not DEFAULT_ADMIN, timelock holds DEFAULT_ADMIN over itself, and EXECUTOR is not opened to the zero address. None of it governs the decision that moved 11.65M USDC off-chain over sixteen days, which was team discretion announced only as a Slope purchase on X and never explained as a liquidity choice.
weak: OracleVerified this run rather than carried, and worse than described. The Morpho USD3/USDC oracle has BASE_VAULT set to the USD3 vault itself and all four price feeds at the zero address, so its price() of 1.168165 is the issuer's own share price echoed back with no external input, governing 17.7M of collateral at 91.5% LLTV. The PT oracle is a single time-only feed returning exactly a 30.0% per year linear discount to par that cannot see USD3 at all. Accountable's new feed attests existence at reported par, but its Slope input is 5.47 days stale against a daily cadence and its own 3-day threshold, its root is not chain-anchored, and its figures sit below the protocol's own.
weak: Liquidity & exitInstant redemption capacity is 3,559,316 USDC, returned identically for Morpho Blue, the Pendle SY and sUSD3, so it is a global ceiling of 4.74% of the pool, down from 20.28%. Deposits are closed, confirmed on-chain by maxDeposit and maxMint both returning zero at the 75M cap, so no new money can fund an exit. Correcting the previous draft, the true exit is primary redemption plus the Curve frxUSD sell side, about 5.0M, not 9.2M; Morpho's free USDC is borrowing capacity, not an exit. No pause, queue or cooldown function exists on USD3; withdrawals simply truncate.
watch: Yield (real vs emission)USD3 pays 6.773% and sUSD3 13.886%, and it is genuinely borrower-paid interest rather than emissions, with the Slope facility earning 10.76% and the LendSwift senior note 16.08%, so real spread exists. That is still a thin senior rate for unrated single-originator private credit with no first-loss under 91% of the book, and the headline marketing numbers still lean on locked JANE against a supply that is not fixed and not transferable.
weak: Holder concentrationSlope is 90.93% of the facility book at 55,252,798, up from 90.03% and 46,160,218, and 3Jane's own facility record names Slope as servicer of the loans it sold, so origination, servicing and collections sit with one counterparty. On the liability side three contracts hold 78.45% of USD3 supply (Pendle SY 40.26%, Morpho Blue 27.61%, sUSD3 10.58%), and total PT supply is 27,469,261 with 22,800,609 floating outside the AMM, all converting to USD3 at the December expiry.
watch: Track recordClean everywhere we can look, and this run we could look at the thing that matters: the Slope tape rendered at 96.9418% current, zero charged off, zero static loss to date, and 0.8872% across all buckets past 30 days, marginally better than 19 July. Zero bad debt and zero realised bad debt on both Morpho markets, no depeg, no entry on DeFiLlama's 610-item hacks list, and USD3 trading at a slight premium on Curve. What holds this at yellow is age, not evidence: the fintech sleeve is about eleven weeks old on a verified 47-day average book, roughly two turns of the assets.
🟢 strong🟡 watch / caveat🔴 weak / fund-loss risk
Verdict is a gate (worst flaw wins), not an average. Our read, not financial advice.
auto-sourced now
TVL$10.0M
30d↑195%
Audits2
Last hacknone

DeFiLlama + our exploits feed. Cross-check the dated report against today.

TL;DR

Our stale-verdict net flagged 3Jane on 2026-07-29 for a TVL collapse. The flag pointed at the wrong thing. USD3's totalAssets() reads 75,042,961.85 USDC today against 75,021,206.46 on 19 July, and the share price went from 1.164754 to 1.168162, a rise of 0.29% in sixteen days or roughly 6.7% annualised, which is exactly what a pool paying 6.77% should do. There is no markdown, no depeg and no default. DeFiLlama counts only the liquid leg, and the liquid leg is what fell 76%.

The real story is the exit. On 19 July, 15,212,591 USDC sat idle and instantly redeemable, 20.28% of the pool. Today the vault's own maxWithdraw() returns 3,559,316 USDC, and it returns that same figure for Morpho Blue, for the Pendle SY contract and for sUSD3 alike, which tells you it is a global liquidity ceiling and not a per-holder allowance. That is 4.74% of the pool. Deployment went from 79.72% to 95.26%, and deposits are shut: maxDeposit() and maxMint() both return zero on-chain because the pool sits at 75,042,962 against a 75,000,000 cap.

Where the money went is not a mystery, and 3Jane said so, just not in a way that framed the consequence. On 22 July the protocol posted on X that it "has purchased a total +$60M in loans from @slopepay, crossing the $50M hold-level on the forward flow announced in June". Slope is now 55,252,798 of a 60,767,395 facility book, 90.93%, up from 46,160,218 and 90.03%. Off-chain paper plus bank cash is 90.74% of the pool, against 73.82% last time.

Three things came out better than the previous draft of this report assumed, and we are correcting them because they cut against our own thesis. First, the Slope delinquency tape did render this run, and it is clean: 96.94% current, zero charged off, zero static loss to date, with all buckets past 30 days summing to 0.89%. That was the single most important missing number last time, and it is no longer missing. Second, the Morpho PT-USD3/USDC loop did not get more expensive. Its borrow APY is 6.46%, down from 7.30% in July, not up to 10.10%, and free USDC there is 2,322,172 at 83.21% utilisation. Third, a third-party proof-of-solvency feed from Accountable is genuinely live and independently reports the facility balances.

What got worse, and stayed worse, is structural. You own a senior claim on an unrated private-credit book that is now nine tenths off-chain, one originator is 91% of it with no first-loss beneath it, a 10.6% junior sits under you, and the pool can pay out 4.7% of itself at once with no inflows to refill it.

Checklist

Audits and contracts. The repo at github.com/3jane-protocol/audits still holds exactly five PDFs and nothing was added since 15 June 2026, confirmed against the GitHub API's pushed_at timestamp: veridise-audit, sherlock-audit, sherlock-2-audit, electisec-audit, yaudit-usd3-susd3-may-2026-audit.

One correction to how that gets counted. Electisec and yAudit are the same organisation, not two firms. yAudit spun out of the yAcademy fellowship, rebranded to Electisec, and reverted to the yAudit name in November 2025. So five PDFs represent three independent firms, not four: Veridise, Sherlock, and yAudit/Electisec twice. All three are real and the Sherlock engagement is independently listed on Sherlock's own contest index rather than resting only on 3Jane's repo. The May 2026 review does land on USD3 and sUSD3 themselves, which is the right target.

The limitation is the same one and it grew. None of these can audit an SPV, a wire, a servicer report or a loan tape, and the share of the pool sitting behind exactly those things went from 73.82% to 90.74%. Separately, the Levered Callable Capital primitive, published 9 July and described on that page as "currently in audit and will be live for deposits in August", has no audit in the repo. It lets users post 7.5% margin against a callable commitment and slashes and auctions that margin on a missed call. It is not live, we have not assessed it, and August is now.

Admin control. Re-read on-chain today rather than carried over, and this run we also read the role wiring the previous draft left unverified.

The Safe at 0x3333...5EF returns VERSION "1.3.0", getThreshold() = 3 and five owners. The TimelockController at 0x1dCc...8f8C2 returns getMinDelay() = 86400, and owner() on MorphoCredit, ProtocolConfig and CreditLine all return that timelock address. On the roles, hasRole confirms the Safe holds PROPOSER, EXECUTOR and CANCELLER; the Safe does not hold DEFAULT_ADMIN; the timelock holds DEFAULT_ADMIN over itself; and EXECUTOR is not granted to the zero address, so execution after the delay is not permissionless. That is a clean, conventional setup with no trapdoor we can find, and it is now verified rather than assumed.

It is also, again, not the control that matters. Over sixteen days the team moved 11.65M USDC from an instantly redeemable buffer into a credit book, in four visible steps with the largest on 30 July and 1 August, with no timelock proposal governing it because none is required. The purchase was announced on X. The liquidity consequence was not: 3jane.xyz/reports has published nothing since 9 July, confirmed against the live index. A holder watching official long-form channels would have learned about the Slope upsize and not about the buffer going to 4.7%. Signer identities remain unknown. That is the whole of what red means here: the on-chain wiring is fine, and it governs none of the decisions that can lose you money.

Oracle. This is where the refresh found the most, and it is worse than the previous draft stated, because we read the oracle contracts instead of carrying a note about them.

The Morpho USD3/USDC market, LLTV 91.5%, prices its collateral with oracle 0x68b4c2...9C06. We read its immutables: BASE_VAULT is the USD3 vault itself, and BASE_FEED_1, BASE_FEED_2, QUOTE_FEED_1 and QUOTE_FEED_2 are all the zero address. Its price() returns 1.168165, which is totalAssets()/totalSupply() to the sixth decimal. There is no external price input of any kind. The collateral value of 17.7M of USD3 in a live lending market is 3Jane's own mark on its own off-chain book, echoed back. If the mark is wrong, nothing on-chain disagrees with it.

The PT-USD3/USDC market, LLTV 86%, uses oracle 0xe5E6Ec...8d72, which has a single feed at 0xdb87eb32...ef8d and no vault and no other feed. That feed returns roundId 0, startedAt 0, answeredInRound 0 and updatedAt equal to the current block timestamp, which is the signature of a computed view rather than a pushed price. Its answer today is 0.889432, and against the 17 December expiry at 134.5 days out that is exactly a 30.0% per year linear discount converging to 1.0. It is a clock. It cannot see USD3's share price at all, let alone a markdown of the loans behind it.

Against that, the new third-party feed. The Accountable dashboard at accountable.3jane.xyz is real and live, announced by @3janexyz on 30 July, and we pulled its data endpoint directly rather than reading a screenshot. It reports total reserves of 75,048,710.43 against on-chain liabilities of 75,042,999.79, a collateral ratio of 100.0076%, retrieved 2026-08-04 11:22:14 UTC.

Read that number carefully, because it is weaker than it looks. It says reported assets equal reported liabilities with 5,710 dollars to spare on 75M. It is a consistency check, not an equity cushion, and the sUSD3 junior sits inside that total rather than on top of it. Four caveats sit on the dashboard's own face or in its own configuration. The Slope input is typed "Document Report" on a stated DAILY interval and was last updated 2026-07-30 00:00:00 UTC, 5.47 days stale against its own cadence and well past the three-day threshold its own config carries as staleAfterDays: 3. Its asset-allocation view reports the entire 75.05M as "Deployed For Borrowing To Morpho" with "Idle In Aave" and "Undeployed USDC" both at zero, which contradicts both 3Jane's own 4.74% buffer line and the on-chain maxWithdraw, and also contradicts the dashboard's own loan-receivables block showing 59.97M sitting in two credit facilities. Its config sets disableBlockchainAnchoring: true, so the signed Merkle root it publishes is not anchored on-chain and the root rotates every refresh cycle. And its numbers do not reconcile to the protocol's own live figures: Accountable prints Slope par outstanding at 54,656,674 where the app prints 55,252,798, and LendSwift senior at 5,314,597 where the app prints 5,514,597. The gaps are explained by the stale feed, which is precisely the point.

One thing the attestation does do, and it is worth crediting: it independently confirms the structural feature we call worst. Against the Slope line it prints backing "Self-collateralized, owned whole-loan" and a cushion of 0 with min_oc of 0, labelled "first-loss". A third party now says on the record that there is no equity beneath 91% of the facility book. Against LendSwift it prints 28.42% current overcollateralisation against a 25% minimum, so the contrast is theirs, not ours.

Liquidity and exit. This is the crux, so take it leg by leg with today's numbers.

Primary redemption is 3,559,316 USDC and that is the whole pool's capacity, not yours. We queried maxWithdraw for the three largest holders and all three returned the identical figure, which is what a shared liquidity constraint looks like. Against a 75,042,962 pool that is 4.74%, down from 20.28%. We probed for gating and found none: paused(), isPaused(), withdrawalsEnabled(), cooldownDuration() and totalPendingWithdrawals() all revert or are absent on USD3. Withdrawals are not disabled, they simply truncate at available cash. Do not read this as a protocol that froze redemptions, because it did not. Read it as a protocol whose redemption capacity is now a rounding error against its size. Note also a separate structural limit in the docs: USD3 can be redeemed only up to the tranche ratio, which caps subordination and is documented with a 15% example against sUSD3's current 10.58%.

Deposits are the other half and they are shut. maxDeposit() and maxMint() both return zero on-chain, and the app's own supply-cap query returns a current supply of 75,042,961.85 against a cap of 75,000,000. The cap was raised from the 50M in the docs to 75M and the pool has filled it. So the one mechanism that normally relieves a tight buffer, new money coming in, is switched off. A closed pool at 95.26% utilisation has no inflow-side relief at all.

Secondary is thinner. Pendle market liquidity is 7,008,068, down 4% from July, but 24h volume is 128,248 against 337,002 last time, a 62% fall. The only USD3 DEX pair remains the Curve USD3/frxUSD pool at 1,879,629 of liquidity, down 27%, on 299,166 of 24h volume, down 50%.

Now correct the arithmetic the previous draft got generous with. That Curve pool is 348,814 USD3 and 1,471,514 frxUSD, and a seller of USD3 can only reach the frxUSD side, so the honest sell-side depth there is at most 1.47M before slippage, not 1.88M. And the Morpho USD3/USDC market's free liquidity, 2,646,848 at 87.2% utilisation, is borrowing capacity, not an exit: you keep the USD3 and the credit risk and you owe USDC. So the true exit today is primary redemption plus the Curve sell side, roughly 5.0M on a 75M pool, with about 2.6M of borrowing available on top as a bridge. USD3 quotes at 1.1705 in frxUSD against a 1.168162 NAV, a slight premium, so the market is not discounting you. At the 50k to 150k sizes this showcase deals in, you can still leave today, cheaply. That is the honest counterweight to everything above, and it is a smaller counterweight than 9.2M would have suggested.

Two structural mitigants deserve mention and one anti-mitigant. The book is short and self-liquidating, and this run we verified it rather than carrying it: the Slope facility reports a 47-day average remaining term across 1,523 loans. If 3Jane stopped redeploying, cash would rebuild in weeks rather than years, and 95% utilisation on a 47-day book is a very different animal from 95% on a five-year book. Second, 7,326,490 is sitting as bank cash at Erebor, which is real cash even though a USD3 holder cannot reach it by pressing withdraw. The anti-mitigant is that this staged cash is staged for a reason: if it lands in facilities, the buffer stays near 4.7% and the off-chain share goes higher still.

Yield: real versus emission. Essentially unchanged and that is the point. 3Jane's own stats print USD3 at 6.773% and sUSD3 at 13.886%, and Pendle independently measures the underlying at 7.006%. Those agree, and the interest is genuinely paid by borrowers rather than printed: the Slope facility carries a 10.76% APY and the LendSwift senior note 16.08%, so there is real spread between what the book earns and what USD3 pays. It is also a thin senior coupon for an unrated book with one dominant originator and no first-loss under 91% of it. The gap to the marketing numbers is still locked JANE against a supply that is not fixed and not transferable, which is a claim, not a yield.

For the PT buyer the arithmetic favours you and by more than the previous draft said. PT trades at 0.9427 for an implied 13.89% to the 17 December expiry, 134.5 days out, at a 4.68% discount to par, against a 7.01% underlying. YT buyers farming JANE are paying that spread. On the leveraged version, we have to retract an escalation. Morpho's PT-USD3/USDC borrow APY is 6.46%, not 10.10%, which is below the 7.30% it carried in July, and the market sits at 83.21% utilisation with 2,322,172 of free USDC rather than 91.5% with 1.07M. The loop's gross carry is therefore about 7.43 points before leverage, wider than July's 5.81, not compressed to 3.79. The reason to avoid that loop is not the carry. It is the oracle described above, which is a clock that cannot see credit.

Holder concentration. Worse on the side that decides whether you get paid. Slope is 55,252,798 of a 60,767,395 facility book, 90.93%, and its absolute size grew 9.09M since July while the buffer fell 11.65M. The protocol's own X post puts cumulative purchases above 60M against a 50M hold level. LendSwift is 5,514,597 drawn against a 10M limit, 55.15% utilised, with 7,517,385 of collateral behind it across 18,992 loans, and remains the only slice with real overcollateralisation.

On the liability side, verified by balanceOf: Pendle's SY holds 25,862,373 USD3 (40.26%), Morpho Blue 17,737,508 (27.61%), sUSD3 6,798,157 (10.58%). Three contracts, 78.45%. Most USD3 is either wrapped into fixed-rate paper or pledged as leverage collateral, so a markdown would propagate through liquidations and PT repricing rather than through quiet redemptions. And the PT overhang is bigger than the AMM suggests: total PT supply is 27,469,261 with 22,800,609 floating outside the pool, all of which converts to USD3 at expiry. Wallet-level holders behind those three contracts remain unenumerated.

Track record. Still clean everywhere we can look, and this run we could look at the thing that actually matters.

The Slope tape rendered. It reports 96.9418% current, 2.1710% late 1 to 29 days, 0.5742% late 30 to 59, 0.2601% late 60 to 89, 0.0529% late 90 plus, zero charged off and zero static loss to date. Everything past 30 days sums to 0.8872%. That is marginally better than the 96.78% current recorded on 19 July. We are correcting the previous draft on this directly: the number was not missing, and it does not support the darker reading.

Both Morpho markets report bad debt and realised bad debt of zero. DeFiLlama's hacks list, 610 entries, has no match for 3Jane or USD3, and we checked the ticker collision deliberately because USD3 is a common symbol. The share price compounded through the entire deployment period without a stumble. USD3 trades at a slight premium on its only DEX pair. A targeted X scan across 20 July to 4 August found no post, complaint or thread about slow, queued, gated or failing redemptions, none about the buffer or utilisation, and none alleging delinquency or credit-quality problems.

What holds this at yellow rather than green is age, not evidence. The fintech sleeve dates from the Erebor facility on 19 May, LendSwift on 29 May and the first Slope purchase on 8 June, so it is about eleven weeks old on a 47-day book. That is roughly two turns of the assets. Credit losses in this class show up on a lag of quarters, and a back-book purchase with a 30-day-plus bucket under 1% at eleven weeks tells you the vintage has not aged, not that it will not.

Worst case

The shape is unchanged and the odds are, if anything, slightly better than we thought. What changed is your ability to leave before it resolves.

Slope's book deteriorates or Slope fails as servicer. Because 91% of the facility principal is a whole-loan purchase carried with no first-loss and no minimum overcollateralisation, and Accountable now confirms that on the record with an explicit cushion of zero, there is no originator equity beneath that slice. Note also that 3Jane's own facility record names Slope as the servicer of the loans it sold, so origination, servicing and collections all sit with one counterparty. Your defences in order are the SPV true-sale and repurchase mechanics, the excess spread the vintage has earned, and then roughly 7,941,348 of sUSD3 junior, which absorbs the first 14.4% of a Slope-book loss before USD3 is touched. The insurance fund does not help: the 19 July FAQ read said it covered the crypto credit lines only, and those are now 4.30% of the pool, and the current suppliers page does not mention an insurance fund at all.

Now the part that got materially worse. In July, a holder who saw trouble early could redeem into a 15.2M buffer. Today that holder redeems into 3.56M, in a pool that cannot accept new deposits, where 90.74% of the assets are off-chain and the fastest of them still takes about 47 days to turn. Everyone reaches for the same 4.7% at once, and the ones who do not get it wait for receivables to amortise while the mark they are waiting on is the same mark they stopped trusting. Nothing has to break for that queue to hurt; it only takes enough holders deciding at once that they would rather not find out.

The second-order version is the one the oracle work exposed. The Morpho USD3/USDC market values 17.7M of USD3 collateral at the vault's own share price with no external feed. A markdown does not get discovered by that market; it gets published to it by 3Jane. Until it is published, borrowers there are neither liquidated nor warned, and when it is published, 17.7M of collateral reprices at once into a market with 2.6M of free USDC.

The levered PT position remains the worst way to hold this, for the right reason. The oracle is a 30% per year time-based discount curve that mechanically converges to par regardless of credit, so you are never liquidated as the book deteriorates. You simply arrive on 17 December with PT redeeming into impaired USD3, against USDC debt that did not shrink, in a market with 2.3M free. And at that expiry 22.8M of floating PT converts to USD3 and joins the same 4.7% queue.

Bottom line

Caution, and we are going to give the reasoning rather than decide it quietly, because the previous draft of this report came close to avoid on numbers that did not survive checking.

The case for avoid is real. Instant redemption capacity is 4.7% of the pool, down by a factor of four. Deposits are shut, so no new capital can fund an exit. Nine tenths of the assets are off-chain paper and bank cash. The junior is 10.6%. The dominant originator got bigger, not smaller, and is also the servicer. And the collateral price for the largest on-chain use of USD3 is the issuer's own mark with no external input, which we verified rather than inferred.

The case against is that avoid should mean do not hold this at any size, and the evidence for that got weaker on inspection, not stronger. The delinquency tape rendered and is clean at 96.94% current with zero charge-offs and under 0.9% past 30 days. The share price compounded straight through the deployment. Both Morpho markets carry zero bad debt. The assets are short-dated and self-liquidating on a verified 47-day average term. The exit path at showcase size still costs nothing today. And the transparency moved the right way: there is now a live outside feed that confirms the facility balances and states the zero first-loss on Slope in the protocol's own shop window.

So it stays caution, and two of the previous draft's action calls need adjusting.

Unlevered PT-USD3 held to 17 December is still defensible at a size you would accept losing, and it is still the only leg where you are paid a genuine premium, 13.89% implied against a 7.01% underlying, by JANE farmers on the other side. Understand that the redemption-at-expiry leg now lands into a 4.7% buffer instead of a 20% one, alongside 22.8M of other floating PT, so hold it because you are content to hold it, not because you expect to exit at will.

The Morpho PT loop stays at do not run it, but not for the reason previously given. The carry did not compress; it widened to about 7.43 points and borrowing there got cheaper, not dearer. Do not run it because the oracle underneath it is a clock. It converges to par on a fixed 30% per year schedule no matter what happens to the loans, which means the position gives you no liquidation signal and no early exit, and hands you the entire credit outcome in a single lump on 17 December.

Named triggers that would flip this to avoid: the buffer falling below about 2% of the pool or maxWithdraw returning zero; any delinquency bucket beyond 30 days moving above roughly 1% on the Slope tape, which is close, at 0.89% today, or the tape ceasing to be published; the Accountable Slope feed going stale by more than a week or the collateral ratio printing below 100%; USD3 trading more than 1% under its share price on Curve for a sustained period; Slope crossing 95% of the facility book; or LCC going live in August without a published audit while the pool is still above 95% utilisation.

Data appendix

  • TVL / pool: totalAssets() = 75,042,961.85 USDC, totalSupply() = 64,240,219.32 USD3, rate = 1.168162, asset = USDC (0xa0b8...eb48), 6 decimals. Compare 2026-07-19: 75,021,206.46 / 64,409,496.12 / 1.164754. Pool flat (+0.03%), share price +0.29% in 16 days (about 6.7% annualised). Supply drifts intraday with redemptions; we read 64,240,038.85 an hour later and the app's own snapshot printed 64,240,194.48. DeFiLlama slug 3jane reports 3,655,616 today vs 15,377,318 on 2026-07-20 (-76%) and an all-time peak of 50,037,444 on 2026-01-28; that series counts only the liquid leg, with borrowed rising from 59,793,541 to 71,474,441 over the same window, and its token breakdown showing WAETHUSDC going from 12,893,845 to 3,014,407 tokens.
  • Redemption capacity (the crux): maxWithdraw() = 3,559,316 USDC, returned identically for Morpho Blue, SY-USD3 and sUSD3, so it is a global ceiling. That is 4.74% of the pool, against 15,212,591 and 20.28% on 2026-07-19. maxRedeem(MorphoBlue) = 3,046,929.17 shares. Deployed per the protocol's own stats = 71,489,142 (95.26%), against 59,808,646 (79.72%) on 19 July. No gate found: paused(), isPaused(), withdrawalsEnabled(), cooldownDuration() and totalPendingWithdrawals() all revert or are absent. Docs add a structural cap: "USD3 can be redeemed up to the tranche ratio."
  • Deposits: closed. maxDeposit(0x0) and maxMint(0x0) both return 0 on-chain; the app's supply-cap query returns currentSupplyUsd 75,042,961.85 against supplyCapUsd 75,000,000. The cap was raised from the 50M in the docs to 75M and is full.
  • Pool composition (3Jane's own stats snapshot, 2026-08-04 10:59 UTC): crypto money markets 3,559,316 (4.74%); bank cash at Erebor 7,326,490 (9.76%); LendSwift warehouse, consumer installment, 5,514,597 (7.35%) with 7,517,385 of collateral, 136.32% backing; Slope forward flows, SMB LOC, 55,252,798 (73.63%); crypto credit lines 3,226,999 (4.30%). Residual 162,762 (0.22%) is accrued interest not itemised. Off-chain paper plus bank cash = 90.74% of the pool, against 73.82% on 2026-07-19. Slope = 90.93% of the 60,767,395 facility book, up from 90.03% of 51,274,815. Facility verified value 62,770,182, weighted LTV 0.707.
  • Slope facility detail (new this run): APY 10.76%, commitment 62,312,839, deployed 55,252,798, utilisation 88.67%, 1,523 loans, average remaining term 47 days, servicer Slope, back-book, status Active, static loss to date 0.
  • Slope delinquency tape (previously unverified, now read): current 96.9418%, late 1-29 2.1710%, late 30-59 0.5742%, late 60-89 0.2601%, late 90+ 0.0529%, charged off 0. All buckets past 30 days = 0.8872%. Compare 96.78% current on 2026-07-19.
  • LendSwift facility detail: senior tranche, advance rate 75%, APY 16.08%, drawn 5,514,597 against a 10,000,000 limit, 55.15% utilised, 18,992 loans, facility maturity 2027-11-26, status Active.
  • Buffer drain timeline (DeFiLlama daily liquid leg / borrowed): 07-20 15,377,318; 07-21 12,204,266; 07-24 11,194,488; 07-29 11,455,023 / 63,679,863; 07-30 6,969,770 / 68,139,073; 07-31 7,042,282 / 68,049,676; 08-01 3,513,061 / 71,622,381; 08-04 3,655,616 / 71,474,441. Four steps, largest on 30 July and 1 August.
  • Third-party attestation: accountable.3jane.xyz, live, announced by @3janexyz on 2026-07-30; the dashboard shell was deployed 2026-07-29. Data pulled from its /dashboard JSON endpoint. Total reserves 75,048,710.43 vs on-chain liabilities 75,042,999.79, net 5,710.63, collateral ratio 1.000076, payload timestamp 2026-08-04 11:22:14 UTC. Loan receivables: Slope forward flows 54,656,674.14 par outstanding, backing "Self-collateralized, owned whole-loan", cushion current 0 / min_oc 0, labelled "first-loss"; LendSwift warehouse 5,314,597.25 lender senior position against 7,517,384.54 net collateral, cushion 28.4189% current, 25% minimum. Four connections: "USD3 Minted Liabilities" ERC4626 15 MIN last 2026-08-04 11:15:00 UTC; "USD3 On-Chain Reserves" ERC4626 15 MIN last 11:15:01; "Slope - Forward Flows" Document Report DAILY last 2026-07-30 00:00:00 UTC (5.47 days stale, config carries staleAfterDays: 3); "LendSwift - Warehouse Senior Note" Document Report WEEKLY last 2026-08-02 00:00:00 UTC (2.47 days, within cadence). Config: DFID 100000026, dvnNode dvn.accountable.capital, disableBlockchainAnchoring: true, proofs displayed as merkle-root, enclave-attestation and zk. Caveats: the asset-allocation view reports all 75.05M as "Deployed For Borrowing To Morpho" with "Idle In Aave" and "Undeployed USDC" at 0, contradicting both the 4.74% buffer and its own loan-receivables block; the Merkle root rotates each cycle and is not chain-anchored in this deployment; and its Slope and LendSwift figures sit 596,124 and 200,000 below the protocol's own live numbers. It attests existence at reported par, not value, not credit quality.
  • Audits: unchanged, five PDFs in github.com/3jane-protocol/audits, pushed_at 2026-06-15T04:46:17Z (veridise-audit, sherlock-audit, sherlock-2-audit, electisec-audit, yaudit-usd3-susd3-may-2026-audit). Electisec and yAudit are the same organisation (yAudit, rebranded Electisec, reverted to yAudit November 2025), so these five files represent three independent firms. Sherlock's engagement is independently listed on audits.sherlock.xyz as a 3Jane contest. SPV, banking rail, underwriting, servicing and collections remain outside all of them. No LCC audit published.
  • Admin/governance (re-read on-chain today): Safe 0x33333333Bd7045F1A601A1E289D7AB21036fB5EF, VERSION "1.3.0", getThreshold() = 3, five owners (0x2086...5631, 0x5a51...95d1, 0x751a...f64f, 0xa84b...6d33, 0x1226...d06b). TimelockController 0x1dCcD4628d48a50C1A7adEA3848bcC869f08f8C2, getMinDelay() = 86400 (24h). owner() returns that timelock on MorphoCredit 0xDe6e...EcBc, ProtocolConfig 0x6b27...f34e and CreditLine 0x2638...c6A9. Role wiring verified this run via hasRole: PROPOSER, EXECUTOR and CANCELLER all TRUE for the Safe; DEFAULT_ADMIN FALSE for the Safe and TRUE for the timelock itself; EXECUTOR and PROPOSER both FALSE for the zero address. Signer identities remain unknown.
  • Morpho markets: USD3/USDC 0xe3df58f9...26d7, LLTV 91.5%, supply 20,456,950, borrow 17,810,102, free 2,646,848, utilisation 87.06% (87.23% on re-read), borrow APY 6.18%, supply APY 5.37%, APY at target 6.33%, collateral 17,737,507.76 USD3, bad debt 0, realised bad debt 0. Oracle 0x68b4c2B2b2e245AB54a3bD55DfD5A9d84f029C06: BASE_VAULT = the USD3 vault, all four price feeds = zero address, price() = 1.168165 = totalAssets/totalSupply. PT-USD3/USDC 0xf8c5aa31...ba89, LLTV 86%, supply 13,829,048, borrow 11,506,876, free 2,322,172, utilisation 83.21%, borrow APY 6.46% (was 7.30% in July, so it fell), supply APY 5.35%, APY at target 6.86%, collateral 16,604,939.74 PT, bad debt 0. Oracle 0xe5E6Ec063E63A9D0c6Ac041d0AD2bd03F7f08d72: single BASE_FEED_1 = 0xdb87eb32a784286312de3a73b2446bb83467ef8d, no vault, no other feeds; that feed returns roundId/startedAt/answeredInRound all 0 with updatedAt = block timestamp and an answer of 0.889432, exactly a 30.0%/yr linear discount to par at 134.5 days out. price() = 0.889432, which is 5.65% below Pendle's market price for PT.
  • Pendle: market 0x4a5067c3ff1abb7449244025b0e37feaf77d8e3e, expiry 2026-12-17 (134.5 days out). Liquidity 7,008,068 (was 7,301,909, -4.0%), 24h volume 128,248 (was 337,002, -62%), implied APY 13.8917% (was 13.11%), underlying APY 7.0063% (was 6.60%), PT discount 4.6812% (was 4.96%), PT price 0.9427, YT 0.0463. Pool reserves totalPt 4,668,652 / totalSy 2,256,459. Now resolved: total PT supply 27,469,261, floating PT 22,800,609, SY total supply 25,862,373. SY contract 0xea3bc608f32847b97965c5e1648bdfcd4c2c40d0, PT 0x7f47c3e6b2c00fc4eb4d5ae50d0ab0ab6888eb4d.
  • Secondary USD3 venue: still exactly one DEX pair, Curve USD3/frxUSD 0x7BA89Bc658c07569cfa6d7947adAA80181a24568, liquidity 1,879,629 (was 2,568,164, -27%) composed of 348,814 USD3 and 1,471,514 frxUSD, 24h volume 299,166 (was 595,659, -50%), 27 buys and 19 sells in 24h, price 1.1705 frxUSD against a 1.168162 NAV, a 0.2% premium.
  • Reachable liquidity, corrected: true exit today is primary redemption 3,559,316 plus at most the Curve frxUSD side 1,471,514, about 5.03M, or 6.7% of the pool. Morpho's 2,646,848 of free USDC in the USD3 market is borrowing capacity, not an exit, and adding it gives about 7.68M on the most generous accounting.
  • Holder concentration: SY-USD3 holds 25,862,373 (40.26%), Morpho Blue 0xBBBB...FFCb 17,737,508 (27.61%), sUSD3 0xf689...64a7 6,798,157 (10.58%). Three contracts, 78.45%. Wallet-level distribution behind them: unverified.
  • Tranche structure: sUSD3 totalAssets() = 6,798,156.72 USD3 against 6,144,279.32 shares; the protocol values it at 7,941,348 USDC-equivalent, 10.58% of the pool, absorbing the first 14.4% of a Slope-book loss before USD3 is impaired. Docs: 1-month lock then a cooldown and withdrawal window (durations not stated), "Losses always hit sUSD3 before USD3", tranche ratio caps subordination with a 15% example. The current suppliers page does not mention an insurance fund.
  • Yield: USD3 6.773%, sUSD3 13.886% (3Jane stats, 2026-08-04 10:59 UTC). Pendle-measured underlying 7.006%. Facility-side: Slope 10.76% APY, LendSwift senior 16.08% APY. JANE token supply not fixed and not transferable.
  • Announcements 2026-07-20 to 2026-08-04: @3janexyz on 2026-07-22, "3Jane has purchased a total +$60M in loans from @slopepay, crossing the $50M hold-level on the forward flow announced in June"; @3janexyz on 2026-07-30, integration with @AccountableData "to provide 3rd party proof of solvency for USD3's backing across offchain warehouse & forward-flow facilities". 3jane.xyz/reports index confirmed: nothing published after "Introducing Levered Callable Capital" on 2026-07-09. A targeted X scan over the window found no post, complaint or thread about slow, queued, gated or failing redemptions, none about the buffer or utilisation, and none alleging delinquency or credit-quality problems.
  • Levered Callable Capital: published 2026-07-09, "LCCs are currently in audit and will be live for deposits in August". Users post 7.5% margin against a callable commitment (75,000 supporting 1,000,000 in the worked example), earn base yield on margin plus a commitment fee on notional, have multiple weeks to fund a call, and have margin slashed and auctioned on a miss. Not live, no audit in the repo, not assessed.
  • Incident scan: DeFiLlama hacks list (610 entries) returns zero matches for 3jane or USD3. Ticker collision checked deliberately: no incident found that belongs to this USD3 at 0x056b269eb1f75477a8666ae8c7fe01b64dd55ecc. No depeg, exploit, emergency pause, default or gating found this run.
  • Not verified this run: Safe signer identities; wallet-level holders behind the three large contracts; LendSwift's loan-level average remaining term (the facility matures 2027-11-26, which is a different quantity); whether the 7,326,490 staged at Erebor is committed to a named facility; findings and severity counts inside four of the five audit PDFs; the current size and scope of any insurance fund; and the "112.50% | 11%" overcollateralisation readout previously seen on the app, which did not appear in this run's data.

Maintained per DeFi Research Instruction v2.

← all protocols