Assessing Stablecoin Credit Risks

Ethan Jones; Yuval Boneh

REVIEW 4 major objections 6 minor 3 references

Assessing Stablecoin Credit Risks

T0 review · 4 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Decentralized stablecoin credit risks form a cumulative spectrum, and the paper derives a formula that sizes a safe line of unbacked credit to an external lending market.

desk verdict The risk-spectrum framework is genuinely useful, but the paper's headline $500k credit line rests on an equation that contradicts its own scaling rationale—worth refereeing, but the number shouldn't be trusted as is. read the letter →

arxiv 2411.13762 v1 pith:IUW73Y7G submitted 2024-11-21 q-fin.RM q-fin.GN

Yuval Boneh , Ethan Jones This is my paper · ORCID

classification q-fin.RMq-fin.GN

keywords DecentralizedfinanceStablecoinCreditriskspectrumOvercollateralizedlendingAlgorithmicmarketoperationsUnderwritingLiquidation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper argues that credit risk in decentralized stablecoin issuance is a cumulative spectrum: each layer, from overcollateralized lending to business-to-system credit, inherits the risks below it and adds one new risk. It assigns each layer a quantified likelihood–consequence grade, with mitigated grades of A2, A2, B2, and A3 for the four layers. The paper also derives a concrete underwriting rule: a $2,000,000 cdxUSD liquidity pool can safely extend a 500,000 cdxUSD line of credit to an Aave market when the transfer function r = 0.15E/(1-E) and a 10% target rate are used. If the rule holds, stablecoin issuers have a computable method for sizing unbacked credit to external protocols, which matters because such credit lines are a proposed path to scaling decentralized stablecoins.

What carries the argument

The machinery is the layered risk spectrum with the discretized risk nomenclature (A/B/C likelihood, 1/2/3 consequence), plus the transfer function r = 0.15*E_controller/(1-E_controller) imported from a companion paper on control-theoretic money supply. This transfer function maps controller utilization to interest rate. The underwriting comparison in Equation 5 sets the external Aave interest rate curve against this transfer function at the instant of utilization, and Equation 6 algebraically reduces that comparison to Rslope1 ≥ 3X/(25-20X). The risk matrix in Table A1 is the summarizing device that assigns final grades to each layer.

What would settle it

Observe the actual Cod3x Lend interest rate for a cdxUSD pool as utilization rises from 0 to 100%: if the realized curve deviates materially from r = 0.15E/(1-E), or if a real Aave market extended a 500,000 cdxUSD line sees the borrow rate drop below the Lend rate before 500,000 is borrowed, the paper's cap is falsified. Concretely, a single utilization snapshot where the Aave rate at 80% utilization is below 10% with Rslope1 = 0.1 would break Equation 6.

Watch

Extended reading notes

Core claim

The central claim is that decentralized stablecoin issuance is not a single risk but a spectrum of layered credit risks, with a distinct dominating risk at each layer: Liquidation Risk for overcollateralized lending, Operations Risk for algorithmic market operations, Cost of Borrowing Risk for business-to-function credit, and Unbacked Circulation Risk for business-to-system credit. The paper quantifies these risks with a three-by-three likelihood–consequence matrix and assigns mitigated grades: Liquidation Risk A2, Operations Risk A2, Cost of Borrowing Risk B2, and Unbacked Circulation Risk A3. The quantitative core is an underwriting inequality, costexternal(x) ≥ yieldfacilitator(x), which the paper resolves into a closed-form cap on credit size. Using the transfer function r = 0.15E/(1-E), a 10% target rate, and Aave's optimal utilization of 80%, the cap solves to X = 0.5, giving a 500,000 cdxUSD line for a $2,000,000 pool; the same logic applied to a perpetuals exchange with a 6% worst-case historical drawdown yields a $6.67M credit cap.

Load-bearing premise

The sizing result depends on the transfer function r = 0.15E/(1-E) being the true cost of liquidity and on every borrowed coin being sold one-for-one into the core pool; if either is wrong, the 500,000 credit cap collapses.

Editorial extensions

If this is right

For an Aave market with optimal utilization 80% and Rslope1 = 0.1, a $2,000,000 cdxUSD core pool can safely extend a 500,000 cdxUSD credit line, giving issuers a concrete starting point for underwriting business-to-function credit.
If the same rate-parity logic is applied to a perpetuals vault using Gains Network's 6% worst-case undercollateralization, a $6.67M line of credit is defensible for a cdxUSD counterparty vault.
Endogenous yield from both the external line and the core pool can be redirected to support liquidity, and at the figures in Section 4.3.1 it sustains 6.6% yield on the pool, further mitigating Cost of Borrowing Risk.
Mitigations reduce Liquidation Risk from C2 to A2, Operations Risk from B3 to A2, Cost of Borrowing Risk from C2 to B2, and Unbacked Circulation Risk from B3 to A3, according to Table A1.
The same underwriting inequality can be applied to other external lending protocols and perpetuals exchanges, making the method a general template for sizing unbacked stablecoin credit, not just the specific Aave and Gains Network examples.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same rate-parity inequality could be inverted to size credit for other external lending protocols, not just Aave, by plugging in their specific interest-rate curves and utilization targets; the paper only demonstrates the Aave case.
Equation 5's single-instant comparison suggests a dynamic version: instead of checking the rate curves once, an underwriter could require the inequality to hold over a rolling window of utilization observations, which would address the paper's own caveat that market conditions over time invalidate the static comparison.
The 0.15 constant in the transfer function is a calibration choice; if Cod3x Lend's actual controller uses a different constant or functional form, the 500,000 result would shift proportionally, and a sensitivity analysis over plausible values of the constant would reveal how robust the cap is to that choice.
The business-to-system example relies on Gains Network's 6% historical drawdown transferring to a cdxUSD vault; a straightforward extension is to run the same sizing exercise with drawdown distributions from other perpetuals exchanges to see how the cap varies with exchange-specific worst-case losses.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Quick take: the conceptual framework is real, but the paper's flagship numeric example has a specific, correctable error that cuts the headline credit line by about 29%.

What's genuinely new is the five-layer risk spectrum—overcollateralized lending, AMOs, B2F, B2S, B2B—with a shared likelihood/consequence grading scale. That's a useful synthesis that I haven't seen laid out this cleanly elsewhere. The B2F/B2S distinction (whether the issuer controls the cost of accessing the stablecoin) is informative, and the risk matrix in Table A1 is a good communication tool. The paper is also honest about its limits: it admits the Eq 5 comparison is only valid at a single instant, and it says outright that structurally mitigating Unbacked Circulation Risk is impossible. The arithmetic in the worked examples checks out.

Now the soft spots. The big one is Eq 4. The paper sets E_controller = X*U and justifies it by saying the credit line should 'scale down with the smaller number of counterassets in the core pool' once cdxUSD has been sold. But that's not what the equation does. If the credit line is X*C0 and a fraction U is borrowed and sold, the amount sold is X*U*C0 and the remaining counterassets are (1-X*U)*C0. The ratio of sold cdxUSD to remaining counterassets is X*U/(1-X*U), not X*U. At the paper's own boundary (X=0.5, U=0.8), the correct E_controller is 0.667, not 0.4, which pushes r in Eq 3 from 0.10 to about 0.30. Solving Eq 6 with the corrected E_controller gives X≈0.357, so a $357k line on the $1M counterasset base, not $500k. The stress-test note is correct, and the problem is independent of where the 0.15 constant came from. The equation contradicts the paper's stated scaling rationale.

Secondary issues: Eq 3 is imported from the first author's SSRN paper without derivation here, Eq 1 is asserted, and several empirical premises (traders lose money over time, liquidation losses are usually covered, Gains' 6% worst case transfers to a new vault) are uncited. No sensitivity analysis. These are all fixable and don't bring down the qualitative framework.

Who this is for: anyone working on stablecoin risk or DeFi credit. The framework deserves to be in the literature. The quantitative example currently isn't reliable, but the flaw is clear and correctable. I'd send it to peer review. I wouldn't cite the 500k number as is, and I'd want to see the corrected version before using the cap in anything.

Referee Report

4 major / 6 minor

Summary. The paper proposes a risk classification for decentralized stablecoin issuance, arranged as a spectrum from overcollateralized lending to business-to-business credit. It defines likelihood/consequence tiers and, for each layer, identifies the principal risk and mitigations, summarized in Table A1. The quantitative centerpiece is a B2F underwriting example (Section 4.3): using a transfer function imported from Boneh (2024), an inequality comparing Aave and Cod3x Lend rates is reduced to Rslope1 ≥ 3X/(25−20X), yielding a 500,000 cdxUSD line of credit for a $2,000,000 pool. A B2S example (Section 5.3) scales the same absorbable amount by a 6% worst-case undercollateralization to obtain a $6.67M cap.

Significance. If correct, the framework offers a transparent, computable rule for sizing unbacked stablecoin credit, which is a genuinely useful addition to a young risk-management literature. The risk matrix is clearly structured and the examples give concrete reference points. The strengths are the explicit inequality conditions and the reproducible numerical arithmetic in Section 4.3; the weaknesses are the reliance on an underived transfer function and the static nature of the comparisons. The contribution is therefore conditional and would benefit from a derivation and sensitivity analysis.

major comments (4)

[§4.3, Eqs. (3)-(6)] E_controller is never defined in the manuscript. Eq. (3) is imported from the first author's SSRN paper (Boneh 2024) without derivation, and Eq. (4) is asserted with only a verbal rationale. Because Eq. (6) and the headline 500,000 cap are algebraic consequences of these equations, the paper must define E_controller in terms of pool state (e.g., normalized imbalance), derive or justify Eq. (4) from the one-for-one selling assumption, and state the status of the 0.15 constant (calibrated or fitted).
[§4.3, Eq. (5) and the paragraph following it] The comparison is evaluated at a single utilization instant, and the authors note that 'market conditions over time are inherently invalidated.' This means the derived 500,000 cap is not a time-valid underwriting limit; it holds only at the chosen point. The paper should either provide a dynamic model of utilization and pool state, or explicitly present the result as a point-in-time baseline and justify why checking optimal utilization is sufficient.
[§5.3] The $6.67M B2S cap is obtained by dividing the $400,000 absorbable amount by the 6% worst-case undercollateralization observed historically on Gains Network. No argument is given that this historical drawdown transfers to a cdxUSD vault, nor is the sample period or data source specified. Without evidence or a stress-test analysis, the cap is an arbitrary scaling of an unrelated historical figure.
[§4.2, Eq. (1)] The central underwriting condition, costexternal(x) ≥ yieldfacilitator(x), is asserted without derivation, and the functions costexternal and yieldfacilitator are not formally defined. As written, it is unclear how x enters each function (e.g., whether it is an amount, a utilization, or a marginal rate). A formal statement of the arbitrage that motivates Eq. (1) is needed.

minor comments (6)

[Throughout] There are numerous typographical errors: 'Consequenece' (Table 1), 'utilzation' in §4.3, 'liqudity' in §4.3.1, 'satisifed' in §4.3.1, 'ingoring' in §4.3, 'benfits' in §2.1, 'inherets' in §7, 'typicaly' in §7. A proofreading pass is needed.
[§4.3] The StableSwap swap result (400,000 cdxUSD for approximately 398,132 counterassets, 70% pool balance) should be accompanied by the invariant or a reference formula so that the reader can reproduce the calculation.
[§4.3] 'VariableRateSlope1 set to 1e26' should be translated into percentage terms (10% in Aave's rate precision) to avoid confusion with a 1e26 value.
[§5.3] The caveats 'not accounting for volume of cdxUSD purchased by traders' and 'subsequent order effect' should be made quantitative if they are intended to justify increasing the cap.
[§6] The B2B section states that understanding its risks is out of scope; if so, the abstract and introduction should be adjusted so that the claimed coverage of B2B credit is not overstated.
[Table A1 and §7] The claim that each layer 'inherits' the risk of the layer beneath it is not represented in the risk matrix; consider adding a column or arrow to show which risks are inherited versus newly introduced.

Assumptions & free parameters 8 free parameters · 7 assumptions · 0 invented entities

The quantitative core is carried by eight chosen or self-cited inputs: the 0.15 controller constant and the functional form of Eq 3 come from the first author's SSRN paper; the 10% target rate is tuned in the example and directly fixes X = 0.5; optimal utilization (80%), Aave slope (1e26), Curve amplification (100), the reserve factor (20%), the Gains yield proxy (10%), and the 6% worst-case undercollateralization are all set for the illustration. Seven domain assumptions, including the asserted Eq 1 condition and the static single-instant comparison, are listed in the axiom audit. The paper introduces no new physical entities; its novelties are risk categories and a calibrated example, so the invented-entities list is empty.

free parameters (8)

Controller constant in transfer function = 0.15 (Eq 3)
r = 0.15*E/(1-E) is imported from Boneh (2024), the first author's SSRN paper; no derivation or calibration data appears here, and it directly sets the credit cap X = 0.5 in Section 4.3.
Target optimal interest rate for Lend = 10% (Rslope1 = 0.1)
'The controller is tuned to output an interest rate of 10%' (Section 4.3); Eq 6 then resolves X to 0.5, so the headline credit number is determined by this choice.
Aave optimal utilization = 80% (U_optimal = 0.8)
Chosen in Section 4.3 to simplify Eq 5 into Eq 6; a standard Aave parameter but still an input choice for the example.
Aave VariableRateSlope1 = 1e26
Set in Section 4.3 so that the Aave rate equals 10% at optimal utilization, making the two arms of Eq 2 match by construction.
Curve StableSwap amplification factor = 100
Assumed in the Section 4.3 pool example; drives the quoted 70/30 pool balance and the roughly 0.995 price for a 400,000 token swap.
Worst-case vault undercollateralization = 6%
Taken from Gains Network history (Section 5.3) and assumed to transfer to a cdxUSD vault; the $6.67M B2S line of credit is 400,000/0.06, so this single number scales the entire B2S cap.
Reserve factor for endogenous yield = 20%
Assumed in Section 4.3.1 to compute the $32,000 annual interest redirected to the liquidity pool.
Gains vault average yield proxy = 10% (rounded from 11.75%)
Section 5.3: 'can be treated as 10% for the purpose of this example'; used to tie the B2S example back to the pool absorption logic of Section 4.3.

assumptions (7)

domain assumption Underwriting condition: cost_external(x) >= yield_facilitator(x) (Eq 1)
Asserted in Section 4.2 as the premise for sizing B2F credit lines; it formalizes an arbitrage-consistency intuition but is not derived from any model or validated on data.
ad hoc to paper Transfer function r = 0.15*E/(1-E) correctly models Cod3x Lend's rate response (Eq 3)
Assumed from Boneh (2024), the first author's SSRN paper; neither the functional form nor the 0.15 constant is derived here, and this is the paper's quantitative backbone.
domain assumption Borrowed funds are sold to the pool one-for-one: E_controller = X*U (Eq 4)
Stated in Section 4.3 as 'it is assumed that Aave utilization is sold to the market'; no evidence or stress case is given for partial or delayed selling.
domain assumption Static single-instant comparison of rate curves is sufficient for sizing (Eq 5)
The paper concedes the comparison 'does not account for the adaptive nature of the Cod3x Lend controller, since market conditions over time are inherently invalidated' (Section 4.3); the robustness of this static treatment is untested.
domain assumption Gains Network's historical undercollateralization of at most 6% transfers to a cdxUSD counterparty vault
Section 5.3 extrapolates Gains' early-lifecycle shortfall to a proposed vault with different collateral, traders, and market depth; this extrapolation sets the entire B2S credit cap.
domain assumption Retail traders lose money over long horizons on perpetuals exchanges
Section 5.1 asserts 'traders tend to lose money' as the basis for counterparty vault profitability; a contested empirical claim with no citation.
domain assumption The likelihood/consequence assignments in Table A1 are the correct expert calibrations
Each cell, such as Liquidation Risk having High likelihood because leverage is 'the core purpose', is a judgment, not an estimate from loss data.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Assessing Stablecoin Credit Risks." pith.science (2026). https://pith.science/paper/IUW73Y7G

@misc{pith2026241113762,
  author       = {Pith},
  title        = {Pith review of: Assessing Stablecoin Credit Risks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/IUW73Y7G}},
  note         = {Machine review of arXiv:2411.13762}
}

read the original abstract

This paper delves into the spectrum of credit risks associated with decentralized stablecoin issuance, ranging from overcollateralized lending to business-to-business credit. It examines the mechanisms, risks, and mitigation strategies at each layer, highlighting the potential for scaling decentralized stablecoins while ensuring systemic health.

Figures

Figures reproduced from arXiv: 2411.13762 by the authors.

**Figure 2.** Overcollateralized Lending on the decentralized stablecoin credit risk spectrum. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** Algorithmic Market Operations on the decentralized stablecoin credit risk spec [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

Figures from the paper (4 more)

**Figure 4.** Figure 4: Representation of a simple Peg Stability Module. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

**Figure 5.** Figure 5: Representation of a Liquidity AMO. 6 [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]

**Figure 6.** Figure 6: B2F on the decentralized stablecoin credit risk spectrum. [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]

**Figure 7.** Figure 7: B2S on the decentralized stablecoin credit risk spectrum. [PITH_FULL_IMAGE:figures/full_fig_p011_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

3 extracted references · 3 canonical work pages

[1]

Aave. (2024). Interest rate strategy. https://aave.com/docs/developers/smart-contracts/ interest-rate-strategy BA Labs Team. (2024, March). Risk assessment - usde morpho lending integration. https: //forum.sky.money/t/risk-assessment-usde-morpho-lending-integration/23924

work page 2024
[2]

Boneh, Y. (2024). Autonomous money supply strategy utilizing control theory.SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4844212 Cod3x Labs. (2024a). Cod3x-labs/whitepapers/cod3x lend. https://github.com/Cod3x- Labs/Whitepapers/blob/main/Cod3x%20Lend%20-%20Lending%20Infrastructure% 20for%20the%20Next%20Generation%20of%20Onchain%20Finance.pdf Cod3...

work page doi:10.2139/ssrn.4844212 2024
[3]

Frangella, E., & Valeri, S. (2022). Gho technical paper. https://github.com/aave/gho- core/blob/main/techpaper/GHO_Technical_Paper.pdf Gains Network. (2024a). Gains stats. https://dune.com/gains/gtrade_stats Gains Network. (2024b). Gtoken vaults. https://gains-network.gitbook.io/docs-home/ liquidity-farming-pools/gtoken-vaults MakerDAO. (2024). Dss-direct...

work page 2022

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Aave. (2024). Interest rate strategy. https://aave.com/docs/developers/smart-contracts/ interest-rate-strategy BA Labs Team. (2024, March). Risk assessment - usde morpho lending integration. https: //forum.sky.money/t/risk-assessment-usde-morpho-lending-integration/23924

work page 2024

[2] [2]

Boneh, Y. (2024). Autonomous money supply strategy utilizing control theory.SSRN Electronic Journal. https://doi.org/10.2139/ssrn.4844212 Cod3x Labs. (2024a). Cod3x-labs/whitepapers/cod3x lend. https://github.com/Cod3x- Labs/Whitepapers/blob/main/Cod3x%20Lend%20-%20Lending%20Infrastructure% 20for%20the%20Next%20Generation%20of%20Onchain%20Finance.pdf Cod3...

work page doi:10.2139/ssrn.4844212 2024

[3] [3]

Frangella, E., & Valeri, S. (2022). Gho technical paper. https://github.com/aave/gho- core/blob/main/techpaper/GHO_Technical_Paper.pdf Gains Network. (2024a). Gains stats. https://dune.com/gains/gtrade_stats Gains Network. (2024b). Gtoken vaults. https://gains-network.gitbook.io/docs-home/ liquidity-farming-pools/gtoken-vaults MakerDAO. (2024). Dss-direct...

work page 2022