The Credit You Could Not See

How institutional lending worked on Maple

Maple Finance launched in May 2021 as an on-chain marketplace for institutional credit. The model placed a financial firm, the “pool delegate,” at the center of each lending pool. The delegate underwrote loans, ran due diligence on borrowers, set the terms, and managed the book. Lenders, ranging from retail depositors to institutions such as Nexus Mutual and Sherlock, supplied capital to specific pools in exchange for yield. Verified

The defining feature, and the source of the eventual loss, was that loans were undercollateralized or fully uncollateralized. Borrowers posted little or no collateral. Credit was extended on the strength of the delegate’s judgment about whether a borrower was good for the money. This is ordinary in traditional private credit, where it is supported by audited financials, covenants, legal recourse, and continuous monitoring. On Maple in 2022, much of that supporting apparatus was thin or absent, and the borrower’s own representations carried far more weight than they should have. Reported

Thin protection beneath the loans

Each pool held a “pool cover” fund, a layer of first-loss capital meant to absorb defaults before lenders took losses. The protection was modest relative to the loans it backed. Reporting at the time put M11 Credit’s pool cover at under $1.2 million against roughly $74 million in loans across its pools, and part of that cover was denominated in Maple’s own MPL token, which fell sharply during the crisis. First-loss capital that shrinks at the exact moment it is needed is not much of a buffer. Reported

One firm, two roles

The governance problem at the heart of this case was a conflict of position. The Orthogonal group operated two arms inside the same small ecosystem. Orthogonal Credit acted as a pool delegate, underwriting and managing its own lending pool. Orthogonal Trading acted as a borrower, drawing loans from pools run by M11 Credit. One corporate parent therefore wore both the gatekeeper’s hat and the borrower’s hat in a market that depended on gatekeepers to be independent. Verified

Eight loans, five days, thirty-six million dollars

The default did not build slowly in public. It arrived in a single weekend, after weeks in which the loans looked healthy and the borrower kept paying.

The sequence

FTX filed for bankruptcy on 11 November 2022. Through the rest of that month, M11 Credit pressed its borrowers to confirm their financial positions. Orthogonal Trading stated, more than once, that its FTX exposure was around $2.5 million, and it continued to make interest and principal payments on schedule. Reported Those payments did the quiet work of reassurance. A borrower who pays on time looks like a borrower who is solvent.

On Saturday 3 December, Orthogonal Trading told M11 Credit that it had taken much larger losses than it had disclosed, tied to funds caught on FTX, and that it could not repay. Reported A $10 million principal payment came due on 4 December and was missed. On 5 December, M11 Credit issued a notice of default across all active loans, and Maple terminated Orthogonal Trading as a borrower and Orthogonal Credit as a delegate. Verified The defaulted loans totaled roughly $36 million across eight positions, which the reporting at the time put at close to 30 percent of all active loans on the protocol. Verified

Where the loss landed

About $31 million of the default sat in M11 Credit’s USDC pool, four loans that represented roughly 80 percent of that pool’s active book. The remainder, around $5 million in wETH, struck a second M11 pool. Verified for the dollar split; the 80 percent loss figure traces to a Maple spokesperson and has been repeated across coverage since, so we present it as an attributed estimate rather than an independently audited number. Reported

The lenders were real institutions with real losses. Nexus Mutual stated it expected to lose around 2,461 ETH, which it described as roughly 1.6 percent of its assets. VerifiedSherlock’s co-founder said the firm anticipated a loss of about 4 million USDC, roughly 35 percent of its staking pool, and apologized publicly to those affected. Verified

The misrepresentation

M11 Credit’s public statement was direct. It said it believed Orthogonal Trading had “purposefully misstated” its exposure, a serious breach of the loan agreement, and that rather than disclose its position the firm had tried to trade its way out of the hole and lost more capital in doing so. The statement noted that interest and principal payments had been used to “maintain the appearance of solvency” until a missed repayment exposed the truth. Alleged / ContestedThese are the counterparty’s characterizations. Orthogonal Trading entered provisional liquidation in the British Virgin Islands later that month and did not publicly rebut them, but intent has not been adjudicated in any public record we located, and we do not assert it as fact.

Maple’s founder said he was shocked and disappointed, and acknowledged that undercollateralized lending needed more stringent due diligence. A Maple spokesperson told reporters there had been “no signs of the insolvency,” because the financials were misrepresented and the on-time payments gave the appearance of health. Verified That sentence is the whole problem in miniature. The signals the protocol could see were clean. The signal that mattered was one the protocol had no independent way to check.

How a managed risk became the whole pool

The most instructive number in this case is not the size of the loss. It is the path the exposure took to get there, because that path was visible on-chain the entire time and still told no one what was coming.

As of 1 September 2022, Orthogonal Trading represented 14 percent of M11 Credit’s USDC pool and 18 percent of its wETH pool, which M11 described as well within its risk framework. Reported By 3 December, the same borrower had grown to a significant majority of the remaining pool, on the order of 80 percent. Reported The firm did not borrow dramatically more. The pool shrank around it.

Look closely at what happened here. After FTX collapsed, the healthy borrowers did exactly what prudent firms do in a panic. They repaid and exited. New lending halted. Each healthy exit raised the share of the pool represented by the one borrower who could not leave, because it could not repay. The de-risking of the pool by its good participants is what concentrated the risk onto its bad one. A risk that sat comfortably inside the framework in September became existential by December without any single decision that looked reckless at the time. Reported

The on-chain dashboard showed this line climbing in real time. Anyone watching could see the concentration rise. What the dashboard could not show was that the remaining borrower was already insolvent, so the rising line read as a balance-sheet curiosity rather than an alarm. Concentration is a flow metric. Insolvency is a credit fact. The protocol had perfect sight of the first and no sight of the second.

What the chain showed, and what it hid

The central thesis of this piece is simple to state and easy to underrate. On-chain data is excellent at showing money moving and poor at showing whether the entity moving it is sound.

This matters because the transparency of a blockchain is genuine, and it invites a specific mistake. An allocator can watch loans originate, watch interest arrive, watch pools fill and drain, and feel informed. The feeling is real and the information is real. It is also the wrong information for the question that decides whether the loan gets repaid. Repayment depends on the borrower’s balance sheet, its leverage, its hidden exposures, and its honesty, none of which appear on-chain. The Orthogonal default is a clean demonstration that a fully transparent flow ledger can sit on top of a completely opaque credit reality.

Same shock, opposite outcome: the Auros contrast

The cleanest evidence that conduct and credit quality drove this outcome, rather than the mechanics of tokenization, comes from a second borrower in the same pools. Auros Global was also caught by FTX, with reporting describing roughly $20 million frozen on the exchange. It also missed payments to M11 Credit in late November 2022. Verified

The response was different because the borrower was different. M11 Credit did not declare default. It pursued a restructuring. Auros entered a light-touch provisional liquidation, agreed a workout on its outstanding exposure, and by November 2023 had repaid its lenders in full, including interest, in what was described as the first successful restructuring of a DeFi credit default. Verified

The five channels, and their limits

A monitoring framework earns trust by being honest about what it cannot do. Here is what disciplined credit monitoring would have surfaced ahead of this default, and what it would have missed.

Would have flagged, in advance

• Transparency: the reliance on borrower self-reporting, with no audited financials and no independent attestation, was a standing structural weakness visible from the start. A framework that scores disclosure quality would have marked this pool down before any stress arrived.
• Governance and structure: one firm holding both a delegate role and a borrower role inside the same network is a conflict that can be identified on day one, not discovered at default.
• Collateral: the undercollateralized nature of the loans meant recovery depended entirely on borrower solvency. That is a knowable, quantifiable fragility.
• Concentration: the climb from 14 percent toward a dominant share was observable on-chain throughout. A monitoring rule with a hard concentration ceiling would have fired weeks before the default, regardless of what the borrower claimed.
• First-loss adequacy: thin pool cover, partly denominated in a volatile protocol token, was measurable in advance and clearly inadequate to the loans it backed.

Would not have caught

• The specific falsehood in the borrower’s self-reported FTX exposure. A claim of $2.5 million against a far larger reality is an off-chain misrepresentation that no on-chain tool can detect on its own.
• The precise size of the loss, which depended on the borrower’s full balance sheet and on post-default recovery.
• The exact timing of the collapse. Monitoring reduces exposure to fragile credits. It does not predict the day the music stops.

The honest conclusion is that a credit-monitoring discipline would not have known the borrower was lying. It would have known the pool was structurally fragile, dangerously concentrated, thinly protected, and dependent on a borrower it could not independently verify. Faced with that profile, the rational action is to reduce exposure before the default, not to wait for confirmation that arrives only as a missed payment. The value of monitoring here is not clairvoyance. It is the discipline to act on knowable fragility instead of trusting clean-looking flows.

A larger market with the same underlying gap

Maple absorbed the lesson at the protocol level. Within weeks it released a set of reforms: faster default declarations, first-loss capital provided by delegates to align their incentives, and the removal of the long withdrawal lockups that had let faster lenders exit ahead of slower ones. Through 2023 it moved away from undercollateralized lending entirely, brought underwriting in-house, and built out cash-management and yield products. By late 2025 and into 2026 the protocol had grown substantially, reporting several billion dollars in assets under management and describing itself as the largest on-chain asset manager of its kind. Reported

The broader category grew with it. Tokenized private credit became the largest real-world-asset segment on-chain, with active value measured in the tens of billions depending on how the figure is defined. Reported Definitions vary widely here, and the gap between transferable on-chain value and total represented value is large enough that any single headline number should be read with care. We flag that variance rather than pick a convenient figure.

Flows are not credit

The Orthogonal Trading default was not a failure of blockchain technology. The chain did its job perfectly, recording every loan and every payment with complete fidelity. The failure was a category error built into the lending model, the assumption that a transparent record of money moving is the same as knowledge of whether the borrower can pay it back. It is not. The loans were current until they were catastrophic, the dashboard was accurate until it was irrelevant, and the loss formed entirely in the off-chain space that the on-chain data never touched.

For an allocator or issuer in tokenized private credit, the operative lesson is that credit assessment cannot be outsourced to the ledger. The structural fragilities in this case, undercollateralization, self-reported financials, a conflicted gatekeeper, thin first-loss capital, and runaway concentration, were all knowable before the default. The borrower’s specific lie was not. A serious monitoring discipline does not promise to catch every lie. It promises to refuse to treat clean flows as a clean bill of health, and to act on fragility that is visible while there is still time to act.

What remains unresolved

Orthogonal Trading’s true FTX exposure was never publicly quantified with precision, so the gap between the $2.5 million it reported and its actual loss remains an estimate. The 80 percent pool loss was an at-the-time figure from a protocol spokesperson rather than a final audited recovery number. Whether the misrepresentation was deliberate fraud or catastrophic internal failure has not been adjudicated in any public record we found. This document will be updated if liquidation filings or recovery records quantify these with primary sources.

References & sources

Collateral Risk Series.No. 01: the Stream Finance / xUSD collapse. No. 02 (this piece): tokenized private-credit credit quality. Forthcoming: recursive-leverage and rehypothecation failure modes; oracle design and liquidation failure in tokenized-collateral lending. Arunights Research · Prepared by Dhruv Aggarwal. Figures and tags current as of the date of writing and subject to revision as primary findings emerge. © Arunights.