Self-managed JAAA loop on Aave Horizon

Latest economics refresh — 7 September 2026

The refreshed $1M / 4× table supersedes the older economics below: 4.87% trailing 30-day JAAA APY, 3.00% GHO borrow APR, and 10.36% modeled equity APY / $103,556 annually before swaps, gas and entry delays. Capacity covers the scenario. The current NOCA Safe is not a valid JAAA member on-chain, so wallet onboarding remains outstanding. This is an economics/capacity/permission refresh; the risk assessment below retains its original review date.

Bottom line

The position is a self-managed leveraged JAAA carry, run from a NOCA wallet that is KYC’d and allowlisted by Anemoy: subscribe USDC into the JAAA fund, supply JAAA to Aave Horizon (the Aave Labs RWA instance) in eMode “JAAA GHO” (LTV 88% / liquidation threshold 90% / bonus 5%, verified), borrow GHO at its governance-set flat 3% rate, swap to USDC, re-subscribe, and loop to a chosen leverage. Net return convention: leveraged spread between the JAAA NAV accrual (5.36% trailing-30d, on-chain) and the GHO borrow rate, minus swap costs (~5–15bp per loop leg) and zero venue fees — at 4x roughly 12.4% gross ≈ 12.2% net, versus ~14.5% net at 8x on 3F with materially thinner liquidation distance. LlamaRisk’s own Horizon reporting quoted “9.63% at 5.26x” in April 2026, before the carry re-widened.

This is the same trade 3F sells (see ../3f-jaaa-leveraged-carry/memo.md), with the middleman removed: no performance fee, no bridge spread, no keeper who alone can execute our exit, no 2-of-3-behind-24h upgrade risk, a Chainlink OCR2 feed (16 transmitters) instead of Chronicle’s 2-of-25, and a 90% liquidation threshold instead of 98%. What we take back in exchange: entry mechanics (~5 sequential subscription rounds to loop to 4x — roughly a week in total at the fund’s measured T+0–T+3 fulfillment, or a single ~1–3-day round if the treasury self-bridges the full notional upfront), self-managed liquidation risk, and Horizon’s own control plane (Aave Labs Safes with no timelock on parameters).

Immediately available exit: none at the position level — debt can be repaid from treasury instantly (GHO is liquid), but the JAAA collateral itself always exits through the fund’s async redemption gate (Anemoy single-EOA approval, no forcing function; measured fulfillment: median 0.27d, max 3.22d over six months, n = 43/43 — fast in practice, guaranteed by nothing). Dominant risk: a NAV markdown while levered, priced by a guarded feed that freezes at the last accepted value during anomalies, in a venue whose three whale co-borrowers sit at health factors 1.03–1.07 and would hit the same redemption pipe simultaneously.

Capacity today is the binding fact: the JAAA supply cap is 40M shares, 32.7M already supplied — headroom ≈ 7.3M JAAA (~$7.6M of collateral, ~$1.9M of equity at 4x) unless Horizon raises the cap again (it has gone 9.9M → 40M in a year, and the Resolv/Centrifuge $100M loop program implies more raises — discretionary, not guaranteed).

Mechanism and true exposure

One unit of USDC travels: Anemoy subscription → JAAA → Horizon supply (eMode 10) → GHO borrow → swap → repeat.

Fund leg: identical to the base memo — USDC into the ERC-7540 vault (0x4880…780B), JAAA (KYC-gated) minted at NAV. Measured fulfillment over the six months to 2026-09-03: deposits median 0.61d / max 4.18d (n = 83), redemptions median 0.27d / max 3.22d (n = 43/43 fulfilled) — near-continuous business-day processing, not batched cycles. The Anemoy pool-manager EOA remains the sole manager (re-verified); its NAV publication cadence has degraded to 22–94h lags even as settlement stays fast.
Horizon supply: pool 0xAe05…32C8, a licensed Aave v3.3 instance operated by Aave Labs with Aave DAO holding upgrade rights. Supplying JAAA requires no Horizon-side permission — the gate is JAAA’s own transfer hook (verified: a non-allowlisted supply reverts at the token; supply onBehalfOf another account is disabled). The received aJAAA is non-transferable (transfer/approve/transferOnLiquidation/transferUnderlyingTo all revert in RwaAToken); positions cannot be moved, sold, or used elsewhere — accounting liquidity only.
eMode 10 “JAAA GHO”: LTV 88%, LT 90%, bonus 5% (verified). Base mode (81/86, bonus 5%) allows USDC/RLUSD borrowing instead; eMode restricts borrowing to GHO (verified: RLUSD borrow reverts in eMode, GHO borrow simulates clean from an existing supplier).
GHO leg: flat 3% borrow rate — a governance-set parameter, not a utilization curve (verified IR data: base 3%, slopes 0; changed five times in twelve months: 4.75% → 4.25% → 3.25% → 2.75% → 3.00%). GHO is oracle-priced at a hardcoded $1.00, so debt never marks above par — but unwinding requires buying GHO at market. Borrow cap 50M, 32.5M drawn (headroom ≈ 17.5M). Swap leg GHO↔USDC via Curve/Balancer, UNVERIFIED depth at our size.
Oracle: Horizon prices JAAA via “JAAA NAV — Aave LlamaGuard”: a Chainlink OCR2 aggregator (16 transmitters) behind Aave’s ScaledPriceAdapter, publishing the same accrued NAV as Chronicle’s feed (1.047726 on both, vs the vault’s own 1.047512). LlamaGuard (LlamaRisk + Chainlink + Aave Labs) applies off-chain dynamic bounds and rejects out-of-bounds NAV updates before publication; on anomaly the feed freezes at the last accepted value and circuit breakers can block supply/borrow. Bound widths and staleness tolerances for JAAA are unpublished (UNVERIFIED); the bounds registry (0x69D5…318b) is owned by a 3-of-4 Safe. The NAV root is still Anemoy’s single EOA — shared with the base JAAA position, with 3F’s Morpho market, and with every other Horizon JAAA borrower.
Liquidation: liquidationCall with receiveAToken=true is disabled; the working path burns aJAAA and transfers underlying JAAA to the liquidator — so the liquidator set is exactly the Anemoy-allowlisted addresses (per the Horizon design doc and the RwaAToken source). Thin, but permissionless within that set, and a seizing liquidator still exits through the fund’s redemption path (measured T+0–T+3 in calm markets; unobserved under stress). Liquidation protocol fee 0 (required, verified). Separately, an ATOKEN_ADMIN forced-transfer power exists (RwaATokenManager 0x803e…0EB9, administered by the Aave Labs 4-of-6 Safe; no per-token grants made yet): it can move any user’s aJAAA without approval, health-factor-enforced, designed for sanctions/lost-key cases.

Label-versus-exposure gap: “JAAA collateral on Aave” is AAA-CLO credit at chosen leverage where (i) collateral pricing is a guarded copy of a NAV one EOA produces, (ii) collateral mobility is zero (non-transferable aToken over a KYC-gated token over an async fund), and (iii) the venue’s other side is three whale loopers at HF 1.03–1.07 whose stress behavior converges on our exit pipe. Leverage layers: the CLO waterfall (inherent) plus one on-chain loop. No socialization layer — the position is ours alone, which also means nobody manages it but us.

Net economics, benchmark and capacity

Return legs:

Organic: JAAA NAV accrual, 5.36% (30d) / 4.29% (90d) / 4.23% (180d) annualized from pricePerShare() (pinned reads). Floating: SOFR + 100–150bp minus the 0.50% expense ratio.
Cost, borrow: GHO flat 3.00% today; governance-adjusted within a 2.75–4.75% observed range. The spread (2.36% today) is policy-dependent on both legs, but unlike 3F’s utilization-driven market it cannot spike intra-week — it moves by governance action.
Cost, friction: GHO→USDC swap on each loop leg and USDC→GHO on unwind (~5–15bp each, UNVERIFIED at size); subscription/redemption at NAV (no fund entry/exit fee); gas trivial. No management fee, no performance fee, no bridge spread.

Measure	4x (LTV 75%)	5x (LTV 80%)	8.3x (LTV 88%, eMode max)
Gross carry (today’s legs)	12.4%	14.8%	22.6%
NAV markdown to liquidation (LT 90%)	16.7%	11.1%	2.2%
Equity loss per 1% markdown	~4%	~5%	~8.3%

Benchmarks: unlevered JAAA ~5.4%; 3F’s 8x product ~14.5% net at a 5.7% liquidation distance. The decision-relevant comparison: 4x here earns within ~2 points of 3F’s 8x while tolerating three times the markdown.

Capacity and marginal capital: supply cap 40M JAAA, 32.72M supplied (82%): headroom ~7.3M shares ≈ $7.6M collateral ≈ $1.9M equity at 4x ($1.5M at 5x). GHO headroom $17.5M covers the borrow leg. Marginal capital does not move the GHO rate (flat curve) — the constraint is purely the cap, which Aave Labs’ risk track (LlamaRisk/Chaos) has raised four times in a year. A NOCA-scale position needs a cap raise conversation first; the Resolv program suggests Horizon wants the growth, but the raise authority is a no-timelock operator function.

Exit, cost and latency

Fast unwind (treasury self-bridge — the path 3F cannot offer us): repay the full GHO debt from treasury in one transaction (buy GHO at market; debt is par-priced), withdraw all JAAA, requestRedeem the entire position. Latency: ~1–3 days at the measured fulfillment cadence (median 0.27d, max 3.22d, n = 43 — including Grove’s $318.5M redemption fulfilled in 1.92d); temporary treasury outlay ≈ 3× equity at 4x. This also works as a fast entry: front the full notional, subscribe once (deposit fulfillment: median 0.61d, max 4.18d, n = 83), borrow back 75% — leveraged in days with no bridge counterparty.

Looped unwind (no external capital): per round — redeem the withdrawable excess collateral, wait for settlement (typically hours, up to ~3 days), repay, withdraw more. ~5 sequential rounds from 4x — roughly a week in total at measured cadence, symmetric with looped entry. During the whole unwind the position stays levered and exposed to markdown.

What can block or degrade exit: the Anemoy redemption gate (single EOA, no forcing function — the base memo’s R1/R3 verbatim); a Horizon reserve freeze/pause (operator Safes, no timelock — freezing blocks new borrows/supplies; a pause also blocks withdraw/repay); a LlamaGuard feed freeze that holds the price while true NAV moves (borrow headroom and liquidation both computed on the frozen price); GHO trading above par when we must buy it back (paid at the swap, not the oracle). Concurrent exits: the three whales ($55M+ combined collateral at HF 1.03–1.07) redeem through the same operator-gated fund path; under a markdown, liquidators seizing their JAAA also join that queue ahead of or beside us — and the fund’s measured T+0–T+3 record contains no episode of concurrent large redemptions under a markdown.

Stress scenario (Mar-2020 shape, 3–5% markdown): at 4x we lose 12–20% of equity on paper, remain ~12 points above liquidation, and can choose to hold through recovery — the position survives without forced realization, which is the entire argument for 4x over 3F’s 8x/10x. At 8.3x the same event is a near-certain liquidation at the marked-down NAV plus 5% bonus. Between those bounds, leverage is a dial the committee controls at entry and we control thereafter (repay-down is always available at treasury speed).

Risks

Verified on-chain this session (2026-09-02, blocks 25,890,991 and 25,891,011): reserve config and history, eMode categories, caps and headroom, IR data and GHO rate history, oracle chain (adapter → EACAggregatorProxy → OCR2, 16 transmitters), RwaAToken source and disabled paths, ACL role map, RwaATokenManager grants (none), whale account data, borrow/withdraw/supply simulations. External: Horizon design doc (aave-v3-horizon repo), ARFC + governance forum, LlamaRisk weeklies, LlamaGuard release post — all 2026-09-02. Base-memo JAAA risks import wholesale.

ID	Risk	Mechanism and evidence
R1	Leveraged liquidation into a thin, KYC-gated liquidator set, beside whales at the edge	eMode LT 90%, bonus 5%. Working liquidation path transfers underlying JAAA — only Anemoy-allowlisted addresses can liquidate, and they then exit through the same operator-gated fund path themselves. Top three suppliers run HF 1.027 / 1.038 / 1.073 ($14.3M / $31.2M / $10.2M collateral): a ~3–4% markdown makes tens of millions liquidatable at once. The Horizon design doc itself flags that slow liquidations accrue bad debt. Our control: enter ≤4–5x and hold repay-capacity in treasury. Trigger: own HF < 1.5; any whale HF < 1.01; aggregate market HF deterioration.
R2	Guarded-NAV oracle freezes exactly when it matters; root is one EOA	LlamaGuard rejects out-of-bounds NAV updates and freezes at last-good; bound widths/staleness for JAAA unpublished; bounds registry owned by a 3-of-4 Safe. A freeze during a real markdown lets us borrow against stale collateral value and delays liquidations market-wide, then gaps. Feed = Chainlink OCR2 (16 transmitters — materially stronger than Chronicle’s 2-of-25), but the published value is the same Anemoy-EOA NAV accrual (identical 1.047726 on both venues). Shared-root exposure across: this position, the base JAAA position, 3F, and all Horizon JAAA whales (D6). Trigger: feed vs vault `pricePerShare` divergence >25bp; feed age >36h; params-registry change.
R3	Operator control plane at Safe speed, no timelock	Aave Labs-side Safes (4-of-6 `0x13b5…e1d3` and executors it owns) hold POOL_ADMIN / RISK_ADMIN / EMERGENCY_ADMIN: can freeze/pause the reserve, cut LTV/LT, and change caps with no timelock (BGD Labs’ unadopted objection in the ARFC). The same Safe administers the ATOKEN_ADMIN forced-transfer power over aJAAA (HF-enforced; no grants made yet; recipients not validated against the allowlist). Aave DAO holds upgrade rights via Executor Lvl-1 (slow, standard). Open item: deployment contract `0x3eaf…d51e` (AaveV3SetupBatch, owned by an EOA) still holds DEFAULT_ADMIN + POOL_ADMIN. Trigger: any ACL role change; any RwaATokenManager grant; reserve parameter change; the 0x3eaf roles being exercised.
R4	GHO legs: policy rate and peg-at-exit	Borrow rate is set by governance (five changes in 12 months, 2.75–4.75%): a 150bp hike at 4x costs ~4.5 points of carry with no utilization signal preceding it. Debt is oracle-priced at $1.00 but repayment buys GHO at market — a GHO premium at unwind is an exit tax; depth of GHO/USDC at our size `UNVERIFIED`. Trigger: GHO rate-strategy update event; GHO market price >$1.005 sustained.
R5	Self-managed operations; fund cadence is habit, not contract	Looped entry to 4x ≈ 5 sequential subscription rounds (~1 week at measured T+0–T+3 fulfillment), exposed to NAV and rate moves mid-build; fast entry/exit needs treasury fronting ≈ full notional for days. The measured cadence (median 0.27d redeem / 0.61d deposit over six months, no stress observation) is operator behavior with no on-chain enforcement — plan operations assuming T+3 or worse. No keeper exists to fail — and none exists to save us either: liquidation-avoidance (repay-downs on markdown) is a NOCA operational duty, including through the fund’s observed 94h NAV-publication lags. Trigger: NAV staleness >72h while levered; markdown >1.5% in 24h; any redemption request pending >5d.
R6	Underlying JAAA and venue concentration imports	Single-EOA NAV/redemption/KYC gate, custody opacity, CLO mark-to-market, anchor concentration — base memo R1–R5 unchanged. Venue-side: JAAA reserve is 82% of cap, dominated by three addresses (likely the Resolv program, attribution `UNVERIFIED`); Horizon TVL has drifted down (~$557M Nov-2025 peak → ~$350M Jul-2026). Trigger: supply-cap utilization >95%; any whale redeeming >25% of the reserve.

Trust and control

Funds (theft / forced loss): no single key. Paths: (i) Aave DAO governance upgrade — the standard Aave proposal + executor path, slow and public; (ii) the Aave Labs 4-of-6 Safe granting itself/an accomplice the aJAAA transfer role and force-transferring collateral — health-factor-enforced, so encumbered collateral is protected up to HF 1, but unencumbered collateral is seizable by design (sanctions tool); (iii) a false NAV needs a majority of 16 Chainlink OCR2 transmitters and passage through LlamaGuard bounds — materially harder than the 2-of-25 Chronicle path at 3F. Root NAV manipulation by the Anemoy EOA moves value slowly and visibly (bounded per push by LlamaGuard).
Liveness (freeze / blocked exit): 1 — the Anemoy pool-manager EOA (redemption gate), same as every JAAA venue. Additionally the 4-of-6 operator Safe can freeze/pause the reserve at transaction speed, and LlamaGuard can freeze the price feed by policy.
Compared with 3F: we shed the keeper monopoly, the pausable claims, the 2-of-3/24h upgrade cliff, and the quorum-0 intent surface; we accept an operator Safe without timelock and a sanctions-grade forced-transfer power held by identified, reputationally-exposed parties (Aave Labs / Aave DAO).

Evidence boundary

Pinned and reproducible: blocks 25,890,991 (my reads) / 25,891,011 (delegated reads), 2026-09-02, cast + $ETH_RPC_URL; snapshot.json beside this memo. Sources verified from Blockscout/GitHub: RwaAToken + RwaATokenManager + ScaledPriceAdapter + EACAggregatorProxy/OCR2, ACL role events, eMode configs, IR data + GHO rate-update history, reserve config history (71/78 → 81/86; cap 9.9M → 40M), whale account data, borrow/withdraw/supply and eMode-restriction simulations.

External inputs (2026-09-02): Horizon design doc (liquidation + forced-transfer semantics); ARFC: Horizon’s RWA Instance (governance split + BGD/ACI dissent); LlamaRisk Horizon weeklies; LlamaGuard release; Chaos Labs on Horizon; Resolv $100M program; Aave launch/institutions blog posts.

Unresolved (falsifiable, before any capital decision):

LlamaGuard bound widths and staleness policy for the JAAA feed (feed contract unverified; params registry layout not yet reverse-engineered).
Identity/intent of 0x3eaf…d51e (setup batch with live POOL_ADMIN + DEFAULT_ADMIN, owned by an EOA).
Whale attribution (Resolv?) and their deleveraging behavior/agreements.
GHO/USDC executable depth at $5M+ and worst observed GHO premium.
Supply-cap raise process and whether NOCA-scale headroom is realistic (requires engaging Aave Labs/LlamaRisk).
NOCA KYC + allowlisting with Anemoy (prerequisite, same as base memo) and jurisdictional eligibility for the fund.
A fork test of the full loop (supply → eMode → borrow → repay → withdraw → redeem request) from an allowlisted address; this session verified each leg by static simulation but not the sequenced flow.

Supporting evidence

Reserve history (verified): listed Sep-2025 at LTV 71 / LT 78, cap 9.9M; raised to 81/86 by Dec-2025; cap 9.9M → 10M (Apr-2026) → 22M (May-2026) → 40M (Sep-2026); supplied ~1M → 2.5M (Apr) → 16.7M (May) → 32.7M (Sep). GHO base rate: 4.75% (Aug-2025) → 4.25% → 3.25% → 2.75% → 3.00% (May-2026).

Whale cohort (block 25,890,991): 0xd5ef…4ae2b — $14.28M collateral, $12.51M debt, eMode 10, HF 1.027; 0xcf25…68fb — $31.21M, $26.61M, HF 1.038 (mixed collateral); 0xd838…823b — $10.19M, $8.23M, HF 1.073. Together ~97% of aJAAA.

Simulations (eth_call, pinned): borrow GHO from existing supplier → clean; borrow RLUSD in eMode 10 → revert (eMode restriction); withdraw JAAA → clean; supply from non-KYC address → reverts at the JAAA token. Comparison exhibit with the 3F route lives in the 3F memo’s answers thread and the table under Net economics.

Fund settlement latency (measured 2026-09-03, vault events blocks 24,606,331→25,896,331, timestamps batch-fetched): every RedeemRequest matched to the next RedeemClaimable per controller — 43/43 fulfilled, median 0.27d, max 3.22d, largest single redemption Grove’s 318.5M JAAA (2026-03-09) fulfilled in 1.92d; deposits n = 83, median 0.61d, max 4.18d. This latency series is the basis for every entry/exit timing figure above; it is observed operator behavior with no on-chain enforcement and contains no stress episode.