Eight Modules,
One Framework

1. The Structural Gap

Traditional credit rating architecture was designed for a different universe. Fitch, S&P, and Moody's methodologies depend on audited financial statements, private issuer disclosures, and decades of default data from traditional finance. That architecture serves sovereign debt and corporate bonds. It was never designed for BTC-collateralised lending, stablecoin yield protocols, BTC treasury preferred equity, or on-chain real-world assets.

The risk factors that govern outcomes in digital asset yield-bearing positions — on-chain reserve verification, collateral liquidation mechanics, stablecoin peg behaviour, DeFi governance attack surface — are structurally absent from traditional agency methodologies.

2. Empirical Calibration

ACI's calibration is anchored to 38 reconstructed historical stress events and public failure cases from January 2020 through December 2025. The dataset spans four asset classes.

• CeFi lending and yield platform insolvencies: Celsius, BlockFi, Voyager, FTX, Genesis, Babel Finance.
• DeFi protocol exploits and stablecoin failures: Euler, Cream, Beanstalk, Mango Markets, Wormhole, TerraUSD, Iron Finance, USDN.
• Bitcoin mining infrastructure failures: Core Scientific, Compute North, Argo Blockchain, Iris Energy, Stronghold, Greenidge, Mawson, Marathon Digital, Bitfarms.
• Market-neutral strategy failures: Three Arrows Capital, Alameda Research, Genesis Trading.

For each event, public evidence was reconstructed approximately 30 days prior to failure and scored retroactively under the ACI framework. Historical reconstruction testing identified all reconstructed failures as ELEVATED or HIGH at T-30.

3. Framework Architecture

3.1 Eight Modules

3.2 Universal Scoring Formula

3.3 Risk Band Framework

4. Module-by-Module Methodology

4.1 Module 1A — BTC Collateral Lending

Module 1A scores custodial and non-custodial credit facilities collateralised by Bitcoin. Core question: if the platform fails, are clients' pledged BTC recoverable? Five criteria are evaluated.

• C1 — Transparency and Disclosure (20%): Proof of Reserves frequency. Monthly third-party attestation scores 100; no PoR scores 0. Disclosure quality has limited standalone discriminative power — several platforms with active PoR subsequently collapsed due to undisclosed structural risks.
• C2 — Collateral Control and Liquidation Mechanics (35%, composite): Custody model (70% of C2) is the dominant single predictor of failure outcomes. Every major CeFi platform that collapsed with material client losses operated commingled or undisclosed custody models. Undisclosed rehypothecation was the central mechanism in the Celsius collapse ($4.7B in client losses).
• C3 — Jurisdiction and Legal Enforceability (15%): Four-tier categorical scoring. Tier 1 jurisdictions (Switzerland, US, UK, Singapore, EU MiCA) score 100; offshore anonymous/unregistered score 0.
• C4 — Structural and Custody Risk — Rehypothecation (25%): Rehypothecation status is the second-strongest predictor of platform failure. No rehypothecation: 100. Undisclosed rehypothecation: 0.
• C5 — Track Record and Operational Resilience (5%): 5+ years established and regulated scores 100; under one year or no regulatory history scores 15.

Representative providers: AMINA Bank (CH), Sygnum Bank (CH), Unchained (US), Ledn (KY), Nexo, Matrixport (SG).

4.2 Module 1B — Treasury Preferred Shares

Module 1B scores BTC-treasury preferred share instruments covering STRF, STRK, STRC, STRD, SATA, and STRE. The primary credit signal is the Senior-Debt-Adjusted Coverage Ratio (SDACR), which accounts for senior secured and unsecured debt ranking ahead of preferred equity in a wind-down scenario.

Five criteria: BTC Coverage Ratio/SDACR (30%), Income Mechanism (25%), Market Risk Composite (20%), Convertibility Risk (17%), Issuer Maturity (8%).

4.3 Module 1C-A — CeFi Stablecoin Yield

Module 1C-A assesses centralised stablecoin yield platforms — not stablecoin issuers. Core question: if this platform fails, do clients recover their stablecoins? Solvency Verification carries the highest weight (35%) as the strongest single predictor of CeFi platform failure. TVL drawdown preceded CeFi platform failures by a median of approximately 47 days in the historical dataset — incorporated via the Liquidity criterion (25%).

4.4 Module 1C-B — DeFi Stablecoin Yield

Module 1C-B scores on-chain stablecoin yield protocols. The risk is code risk and governance risk.

• Protocol Security (40%): Audit depth and battle-test duration are the two most predictive criteria for protocol survival. The Beanstalk ($182M) and Mango Markets ($117M) exploits were governance attacks, not code exploits.
• Governance Risk (30%): Immutability tier (50%) and regulatory record (50%). Fully immutable protocols — no upgrade path, no admin key — score 100.
• Peg Stability (5%): Triggers HIGH_DEPEG_OVERRIDE when 90-day maximum deviation exceeds 1.5%, forcing the HIGH band regardless of composite score.

4.5 Phase 2 Modules — M4 through M8

M4 — Infrastructure. Module M4 scores Bitcoin mining operators and validator infrastructure. Core question: does this operator generate positive cashflow under a sustained BTC price shock, and does the custody and legal structure protect client assets if the operator fails? Twelve criteria span unit economics (30%), reliability (25%), counterparty discipline (20%), transparency (15%), and scalability (10%). The primary criterion is Breakeven BTC Price (12%) — the distance between current BTC price and the operator's monthly cashflow zero. The reliability category requires above 99% uptime for a strong tier; below 95% scores inadequate. Representative providers: Marathon Digital, Riot Platforms, Cipher Mining, Hut 8, Core Scientific.

M5 — Market-Neutral Strategies. Module M5 scores basis, funding-rate, and hybrid market-neutral strategies. Because these strategies continuously rebalance, the duration multiplier short-circuits to 1.000. Fifteen criteria span strategy robustness (30%), risk management (25%), execution quality (20%), liquidity access (15%), and transparency (10%). Three hard caps cannot be overridden by strong performance elsewhere: leverage above 5× without a stop-loss, unknown counterparty, and no position visibility each cap the final score at 30. Each reflects a structural failure mode observed in the historical dataset — Three Arrows Capital, Alameda Research, and Babel Finance each exhibited one or more of these conditions in the months preceding collapse. Representative providers: Wintermute, GSR, Cumberland, Amber Group, B2C2.

M6 — Digital Asset Venture. Module M6 scores early-stage digital asset venture allocations — pre-listing token positions and nascent protocol equity. The return distribution is power-law in nature: a small number of outcomes drive most of the category's aggregate return. The convexity classifier defaults to POSITIVE — M6 is the portfolio convexity anchor. Fourteen criteria span team quality (30%), tokenomics expertise (20%), market potential (20%), investor base (15%), and liquidity path (15%). An anonymous team caps the final score at 30 (VENTURE_ANON_TEAM_CAP), regardless of how well the protocol scores elsewhere — anonymous teams have zero accountability surface in the event of rug-pull, exploit, or misappropriation. Unlock risk penalises short vesting schedules heavily: no vesting with immediate unlock scores 5. Representative providers: a16z crypto, Polychain, Pantera, Paradigm, Dragonfly.

M7 — Tokenised Real-World Assets. Module M7 scores on-chain RWA platforms covering sovereign bills, corporate credit, private credit, and structured products. The core question differs from Phase 1: the risk here is primarily legal and structural, not operational. Can a token holder enforce their claim on the underlying asset if the platform fails? Twelve criteria span asset quality (30%), legal structure (25%), custody (20%), liquidity (15%), and transparency (10%). Two hard caps apply: offshore jurisdiction with no audit and self-custody with no asset segregation each cap the final score at 30. M7 applies subtype-driven duration multipliers: tokenised T-bills 0.95 (the only module below 1.0), private credit 1.10, real estate 1.15, structured products 1.05. Representative providers: Securitize, Backed Finance, Ondo Finance, Maple Finance, Centrifuge.

M8 — Volatility / Options. Module M8 scores options structures, structured products, and DeFi option vaults. Unlike every other ACI module, M8 applies segment-adjusted weights across three venue types: institutional OTC, exchange venue, and on-chain DeFi option vaults. Counterparty Quality is weighted most heavily in institutional OTC (30%); Transparency is weighted most heavily in on-chain DeFi vaults (30%). The convexity classifier defaults to NEGATIVE for short-volatility positions — M8 is the only module where tail-risk profile is surfaced separately from the composite score. Large-capital positions with NEGATIVE convexity trigger a forced ELEVATED band floor. Representative providers: Deribit, Wintermute (OTC), Ribbon Finance, Friktion, Opyn.

5. Three Operating Principles

Determinism. Every risk indicator is computed by a deterministic algorithm. Given the same publicly available inputs and the published weights, any analyst can independently replicate any ACI output. The platform's continuous integration pipeline runs SHA-256 determinism tests across all active scoring modules; a determinism regression is a critical control exception that prevents release.

Full Transparency. Every weight, sub-criterion, formula, data source, and hard rule is published at aethoncredit.com/methodology. No inputs are confidential. Criterion-level score breakdowns are surfaced in the platform's score-detail panel for every scenario.

Append-Only Persistence. Score persistence is structurally append-only. Database triggers block UPDATE and DELETE at the engine level, including for service-role keys. Every snapshot carries a SHA-256 hash over the framework version, methodology version, sorted JSON of inputs and outputs, and PDF byte stream where applicable. Snapshots may optionally be anchored on Bitcoin's blockchain via OpenTimestamps, proving a given hash existed before a given block height — without trusting ACI.

6. Regulatory Alignment

Five specific regulatory developments create measurable demand for the documentation framework ACI provides.

• DFSA Crypto Token Framework (effective 12 January 2026): Requires licensed firms to determine on a "reasoned and documented basis" whether each token meets suitability criteria. ACI's per-module scoring and methodology documentation are designed to be exactly this artifact.
• VARA Issuance Guidance (April 2026): Requires Risk Disclosure Statements ranked by materiality, with generic boilerplate explicitly prohibited. ACI's scoring framework is a direct input to that ranked, factor-level risk disclosure.
• GENIUS Act (signed 18 July 2025): Federal stablecoin framework requiring 100% reserves and monthly public reserve disclosures. Fiduciaries holding stablecoin positions must parse monthly reserve attestations into a credit-quality view. ACI's stablecoin scoring modules (1C-A and 1C-B) provide that framework.
• SEC Project Crypto and SAB 122 (January 2025): SAB 122 rescinded SAB 121, unlocking custody offerings from regulated institutions. RIAs recommending digital assets need a defensible counterparty and custody risk file. ACI is the standing version of that file.
• BCBS SCO60 (effective 1 January 2026): Prudential standard classifying crypto into Group 1 (qualifying stablecoins, tokenised traditional assets) and Group 2 with up to 1,250% risk weight. ACI's module outputs directly address the documentation requirements this creates.

7. Portfolio Aggregation and Stress Architecture

Every ACI module ships a deterministic stress projection. Phase 1 applies parametric curve analysis against published threshold tables. Phase 2 modules ship multi-scenario stress engines.

8. Methodological Limitations

• Public evidence constraints: ACI scores are computed exclusively from publicly available data. Private disclosures, undisclosed liabilities, and non-public regulatory correspondence are outside the evidence set. Low-disclosure entities activate the Worst-Case Protocol.
• Historical reconstruction constraints: Calibration rests on reconstructed historical events. The dataset is not a formal statistical sample and does not support inference beyond the observed cases.
• Outputs are not forecasts: ACI outputs reflect structural risk at the time of computation. A LOW band indicator does not imply safety; a HIGH band indicator does not imply imminent failure.
• Known gaps in Framework v1.0: Absence of a non-recourse SPV exposure criterion in M4; absence of a borrower disclosure quality criterion in M7 uncollateralised lending pools; absence of an oracle manipulation susceptibility criterion for DeFi derivatives in M5 and M8. Each will be addressed in ACI Framework v2.0.

9. Access and Verification

ACI is accessible at three tiers: Sandbox (free, read-only access to published scores), Professional ($1,500 per month, full scoring and IC package generation), and Enterprise ($5,000 per month, API access and white-label integration).

Any published score is verifiable by snapshot ID via the public score verification endpoint at aethoncredit.com/methodology/verify. The endpoint returns verification level, score and band, content hash, framework version, and Bitcoin anchor status. No authentication is required.

The complete ACI Risk Methodology — all criterion weights, bucket score maps, hard cap rules, and worked examples — is published in full at aethoncredit.com/methodology.

Download PDF