# Treasury Mandate Research for a $100,000 DAO Treasury

**A research proposal for a hypothetical DAO that funds ongoing development and may eventually delegate portfolio decisions to IMD Swarm.**

| | |
|---|---|
| Report date | 2026-09-23 |
| Market data as of | 2026-09-22 / 2026-09-23 (Treasury, Fed); 2026-09-21 (Circle); 2026-09-24 02:31 UTC (Etherscan, rwa.xyz) |
| Scope | Research only. No capital is deployed, no trade is placed, no key is held. Every allocation in this document is illustrative. |
| Status | Proposal for founder decision, not an investment recommendation and not financial, legal, or tax advice. |

---

## Epistemic key

Every substantive claim in this report carries one of four tags. The tags are load-bearing: the acceptance criterion for this work is that facts, inferences, uncertainty, and open questions stay distinguishable, so please read them.

| Tag | Meaning |
|---|---|
| **[FACT]** | Measured and attributable to a dated primary or near-primary source, cited inline. |
| **[INFERENCE]** | My reasoning from tagged facts. The facts are checkable; the reasoning is mine and may be wrong. |
| **[ASSUMPTION]** | A number or premise I supplied because no source gave it. Chosen to be plausible and stated so it can be replaced. |
| **[UNKNOWN]** | I tried to establish this and could not, or it depends on information only the founders hold. |

Source quality is also flagged: *primary* (the issuer, the protocol, the government), *near-primary* (an aggregator reading primary data), *unofficial* (community-maintained; useful, not authoritative).

---

## 1. Executive summary

**The headline finding is not about asset selection.** At $100,000, with the 3-month Treasury bill yielding roughly 4.1% **[FACT, §3.1]**, the *entire* risk-free income of this treasury is about **$4,100 per year**. Against any realistic ongoing-development budget, that is a rounding error. The treasury cannot fund development from income under any of the three mandates. It can only fund development by spending principal.

That reframes the question the founders think they are asking. They are not choosing a yield strategy; they are choosing **how fast to burn a fixed, small pot, and how much variance to accept in the burn rate**. Asset allocation is a second-order lever on a first-order problem.

Three findings follow, each quantified in the body:

1. **The income mandate is not currently paid for.** As of mid-September 2026, the large on-chain stablecoin yields — Aave v3 USDC at 3.57% (2026-09-15) and the Sky Savings Rate at 3.60% — sit *below* the 1-month (3.99%) and 3-month (≈4.1%) Treasury par yields **[FACT, §3.1–3.2]**. A DAO moving from bills into DeFi money markets today accepts Euler-class smart-contract risk **[FACT, §5.4]** in exchange for a **negative** yield spread of roughly 40–50 basis points **[INFERENCE, §4.2]**. Over a four-year runway, the income mandate's extra yield buys about **eight days** of additional funding **[INFERENCE, §4.2]**. That is not a trade worth making at this size.

2. **The growth mandate's downside is measured; its upside is assumed.** Bitcoin fell 74.4% and Ether fell 81.8% from their November 2021 peaks to the June 2022 trough **[FACT, §3.4]**. Ether, at $2,723.35 on 2026-09-23, remains **44.0% below** its November 2021 peak nearly five years later **[FACT/INFERENCE, §3.4]**. A 30% risk sleeve suffering the measured 2021–22 drawdown while the DAO spends $25,000/year costs it roughly **eleven months of development funding** versus the preservation mandate **[INFERENCE, §4.3]**. The corresponding upside scenarios are my constructions, not observations.

3. **The binding constraints are eligibility and governance latency, not portfolio theory.** BUIDL requires a $5M minimum and qualified-purchaser status **[FACT, §5.8]**, which mechanically excludes a $100,000 treasury. Transaction costs at this size are negligible — an ERC-20 transfer cost **$0.004** at 0.063 gwei on 2026-09-24 **[FACT, §5.7]**. What actually constrains this DAO is who can legally sign, who can pass KYC, and how many days a governance vote takes during a depeg.

**Recommended default, conditional on the open questions in §9 being resolved as expected:** capital preservation, in cash-equivalents, with issuer diversification and an explicit, pre-agreed spend schedule. Adopt the income mandate only if the DeFi-over-bills spread turns durably positive by more than 150 bp. Adopt the growth mandate only if the DAO has no fixed USD obligation inside the horizon. §7 states precisely what would invalidate each of these.

**The single most important unanswered question (§9.1):** this brief specifies a "$100,000 treasury" in dollars, but the only IMD treasury mechanism I could actually observe sweeps *unclaimed launch tokens* to a treasury address after a 365-day `sweepDelay` **[FACT, unofficial source, §8.1]**. If the real treasury is denominated in the project's own token rather than in dollars, **every recommendation in this report changes**, and the first action is diversification out of the native token, not allocation among external assets. I could not resolve this from public sources.

---

## 2. The question, restated precisely

The brief asks for a comparison of three mandates for a $100,000 DAO treasury that "funds ongoing development and may eventually delegate portfolio decisions to IMD Swarm." Before comparing, three ambiguities have to be named, because they change the answer:

1. **Is this an endowment or a runway?** An endowment is perpetual and spends only income; a runway is a depleting pot with an end date. The phrase "funds ongoing development" implies a runway, but the phrase "capital preservation" as a candidate mandate implies an endowment. These are different problems with different solutions. **[UNKNOWN — founder decision, §6.1]**

2. **What is the unit of account?** "Preserving capital" means nothing until you say *in what*. Preserving USD value, ETH value, and purchasing power over developer-hours are three different objectives that recommend three different portfolios. **[UNKNOWN — founder decision, §6.3]**

3. **What does "delegate portfolio decisions to IMD Swarm" mean operationally?** Proposing trades that humans sign? Executing within a pre-approved universe? Holding keys? These sit at very different points on the risk curve and require very different software (§8). **[UNKNOWN — founder decision, §6.10]**

This report answers the mandate comparison under stated assumptions for each, and flags where a different answer to the above flips the conclusion.

---

## 3. Measured facts: the current environment

### 3.1 Risk-free rates

US Treasury par yield curve, **2026-09-23** (primary: US Department of the Treasury) **[FACT]**:

| Maturity | 1 Mo | 3 Mo | 6 Mo | 1 Yr | 2 Yr | 10 Yr |
|---|---|---|---|---|---|---|
| Par yield | 3.99% | 4.10% | 4.31% | 4.49% | 4.85% | 5.11% |

Federal Reserve H.15, released **2026-09-23**, for **2026-09-22** (primary) **[FACT]**:

- Federal funds effective rate: **3.88%**
- Treasury constant maturities (coupon-equivalent): 1-month 3.97%, 3-month **4.16%**, 6-month 4.26%, 1-year 4.43%, 2-year 4.71%, 10-year 4.96%
- Treasury bills, secondary market (**discount basis** — not comparable to the above): 4-week 3.82%, 3-month 4.01%, 6-month 4.15%, 1-year 4.21%

> **Source discrepancy, disclosed.** For 2026-09-22 the Treasury par curve and the Fed's constant-maturity series agree exactly at 1-month (3.97%), 6-month (4.26%), 1-year (4.43%), 2-year (4.71%) and 10-year (4.96%), but disagree at 3-month: **4.09% (Treasury) vs 4.16% (H.15)**. These series should be identical. Five of six cells matching makes an extraction error on one of the two pages the most likely explanation **[INFERENCE]**, but **I did not resolve it** **[UNKNOWN]**. I use ~4.1% throughout and the 7 bp uncertainty does not change any conclusion.

Older but useful, Treasury daily bill rates, **2026-08-27** (primary) **[FACT]**: 4-week coupon-equivalent 3.72%, 13-week 3.85%, 26-week 3.78%, 52-week 4.01%.

**[INFERENCE]** The curve is upward-sloping across its whole length while the effective fed funds rate (3.88%) sits *below* the 3-month bill (~4.1%). Market pricing is consistent with expectations of higher policy rates ahead. This matters for the mandates: a DAO that locks duration to reach for yield takes mark-to-market risk if that expectation is realised. It is an expectation, not a forecast I endorse.

### 3.2 Achievable on-chain dollar yields

| Venue | Yield | As of | Source class |
|---|---|---|---|
| Aave v3 USDC, Ethereum | 3.57% APY (~$58M supplied) | 2026-09-15 | near-primary (Aavescan) **[FACT]** |
| Aave v3 USDC, Ethereum | 3.74% APY (~$2.377B supplied) | September 2026, exact date unclear | near-primary (DeFiLlama) **[FACT, date uncertain]** |
| Sky Savings Rate / sUSDS | 3.60% APY | undated on page, read 2026-09-23 | primary (sky.money) **[FACT, undated]** |
| Tokenized US Treasury funds, aggregate | 3.50% 7-day APY | 2026-09-24 | near-primary (rwa.xyz) **[FACT]** |
| Coinbase USDC rewards | ~4.1%, up to ~4.5–4.7% by tier/region | 2026, exact date not established | near-primary **[FACT, weakly dated]** |
| USDC held idle in a Safe | 0.00% | n/a | definitional **[FACT]** |

The two Aave readings differ (3.57% vs 3.74%) and describe very different supply totals; they are likely different market deployments or different snapshot times. **[UNKNOWN]** I did not reconcile them. The Sky rate carries no date stamp on the product page; a June 2026 secondary source reported 3.60% on 2026-06-18 and 3.75% in Q2 2026, so the rate is governance-set and moves **[FACT, secondary]**.

### 3.3 Stablecoin and tokenized-treasury infrastructure

- **USDC in circulation: $74.6B as of 2026-09-21** (primary: circle.com/usdc) **[FACT]**. Backed by cash and cash-equivalents, with the majority in the **Circle Reserve Fund (USDXX)**, an SEC-registered 2a-7 government money market fund, custodied at **BNY Mellon**, managed by **BlackRock** **[FACT]**. Monthly attestations by a Big Four firm under AICPA attestation standards; the most recent attestation shown on Circle's transparency page is **as of 2026-09-21** **[FACT]**. I could **not** extract the current line-item reserve composition — the figures are rendered in interactive charts and the PDF was not parsed **[UNKNOWN]**.
- **Tokenized US Treasuries: $14.93B distributed value as of 2026-09-24** (near-primary: rwa.xyz) **[FACT]**. Largest funds: Circle USYC $2.51B, Ondo USDY $2.27B, BlackRock BUIDL $2.23B, iBENJI $1.71B, WisdomTree WTGXX $1.23B **[FACT]**.
- **GENIUS Act**, P.L. 119-27, **signed 2025-07-18** (primary: congress.gov / CRS IN12553) **[FACT]**. Payment stablecoin issuers must hold ≥$1 of permitted reserves per $1 issued, limited to cash, insured deposits, short-dated T-bills, T-bill repo/reverse repo, government MMFs and central bank reserves. Holders receive priority over other claims in issuer bankruptcy. Payment stablecoins are **not** securities or commodities and are **not federally insured**.

### 3.4 Risk-asset history

**[FACT]** (secondary, widely reported price series):

| | Nov 2021 peak | Jun 2022 trough | Drawdown |
|---|---|---|---|
| BTC | $68,906.48 (2021-11-10) | $17,614.34 (2022-06-20) | **−74.4%** |
| ETH | $4,865.46 (Nov 2021) | $883.16 (2022-06-20) | **−81.8%** |

**[FACT]** As of **2026-09-23**: BTC $85,686.06 at 10:00 ET (opened $86,195.28); ETH $2,723.35 (opened $2,753.25). Both near 8-month highs.

**[INFERENCE — arithmetic from the above, start-date dependent, not a forecast]**

| Start point | BTC annualised to 2026-09-23 | ETH annualised to 2026-09-23 |
|---|---|---|
| Nov 2021 **peak** (4.87 yr) | **+4.58%/yr** | **−11.23%/yr** (still −44.0% below peak) |
| Jun 2022 **trough** (4.26 yr) | **+45.0%/yr** | **+30.3%/yr** |

> **This table is the most important thing in §3, and it is deliberately self-undermining.** The same two assets, over overlapping windows, return either "about the same as a T-bill and worse than one for ETH" or "spectacularly." Which story you tell depends entirely on a start date you do not get to choose. Anyone — human or agent — presenting one row without the other is arguing, not measuring. A treasury policy that is only defensible under one row is not a policy.

### 3.5 Transaction costs

**[FACT]** Etherscan gas tracker, read **2026-09-24 02:31 UTC**: 0.063 gwei low/average, 0.07 high; stated costs **$0.004 per transfer, $0.06 per swap**.

**[FACT]** A separate snapshot dated **2026-09-23** showed **3.311 gwei** — roughly 50× higher within about one day.

**[INFERENCE]** At 3.311 gwei with ETH at $2,723.35, a 65,000-gas ERC-20 transfer costs ≈ **$0.59**. So the realistic intraday range for a transfer on 2026-09-23/24 was roughly **$0.004 to $0.59**, and for a swap roughly **$0.06 to $1.80**. Gas is volatile in *relative* terms and negligible in *absolute* terms at $100,000. Four rebalances a year at three transactions each, priced at the pessimistic end, costs about **$24/year ≈ 2.4 bp** **[INFERENCE]**. Conclusion: **gas is not a constraint on this treasury and should not be used as an excuse for inaction.** The costs that matter are product fees and spreads (§5.7).

---

## 4. The three mandates compared

All three scenarios below share one **[ASSUMPTION]**: the DAO spends **$25,000/year** on development, drawn at year-end. This number is mine — the brief does not state a budget. Every conclusion scales with it, and §7 explains what happens if it is very different.

### 4.1 Mandate A — Capital preservation

**Objective.** Minimise the probability that the nominal USD value of the treasury falls below the level required to meet committed obligations, over the funding horizon. Accept a return at or slightly below short T-bills.

**Eligible universe.** Cash and cash-equivalents only: fiat USD at an insured institution; direct short T-bills held through a broker or bank; a major regulated-issuer stablecoin held idle in the DAO's own Safe; the same stablecoin at a regulated US venue paying rewards; a small allocation to a *second* issuer's stablecoin purely for issuer diversification. No lending protocols, no liquidity provision, no duration beyond 6 months, no assets that can be gated or paused by a protocol upgrade.

**Illustrative allocation** (weights are illustrative, not recommended targets):

| Sleeve | Weight | Instrument | Primary risk accepted |
|---|---:|---|---|
| Off-chain cash / T-bills | 35% | USD at insured bank, or 4–13 week bills at a broker | Bank/broker counterparty; access latency; requires legal entity |
| On-chain idle | 35% | USDC in the DAO's Safe | Issuer redemption risk; key management |
| On-chain, yielding, regulated venue | 20% | USDC at a regulated US exchange earning rewards | Exchange counterparty; withdrawal suspension |
| Second issuer | 10% | A second major regulated stablecoin | Issuer risk, deliberately *uncorrelated* with the first |

**Expected yield [INFERENCE]:** roughly **3.0–3.8%**, i.e. **$3,000–$3,800/year**. Blended at 3.5%, runway at $25,000/year spend is **≈4.4 years**.

**Honest weakness.** This mandate has a *real* return problem that nobody should paper over. If inflation runs near the 10-year nominal yield of 5.11% **[FACT, §3.1]**, a 3.5% nominal return is a real loss. Preservation preserves *dollars*, not purchasing power. That is a genuine cost, not a rhetorical one, and it is the strongest argument the growth mandate has.

### 4.2 Mandate B — Sustainable operating income

**Objective.** Generate a durable, spendable income stream sufficient to fund development without depleting principal.

**Eligible universe.** Everything in Mandate A, plus: blue-chip on-chain money markets with multi-year operating history and large TVL (Aave v3); governance-set savings rates backed by disclosed collateral (Sky's sUSDS); tokenized Treasury funds **if the entity can pass the eligibility gate** (§5.8). Explicitly excluded: liquidity provision with impermanent loss, leveraged or looped positions, anything paying yield in a governance token, any protocol under 18 months old or under $250M TVL **[ASSUMPTION — my thresholds, chosen to exclude the long tail; founders should set their own]**.

**Illustrative allocation:**

| Sleeve | Weight | Yield (dated) | Primary risk accepted |
|---|---:|---|---|
| Tokenized T-bill fund (if eligible) | 25% | ~3.50% (2026-09-24) | Eligibility gate; fee drag; transfer-agent/redemption process |
| Sky sUSDS | 20% | 3.60% (undated) | Smart-contract; governance can change or remove the rate |
| Aave v3 USDC | 20% | 3.57% (2026-09-15) | Smart-contract; oracle; utilisation spikes blocking withdrawal |
| Idle USDC buffer in Safe | 25% | 0% | Issuer only — this is the sleeve that pays the bills |
| Second stablecoin issuer | 10% | varies | Issuer diversification |

**Expected yield [INFERENCE]:** roughly **3.4–3.8%** net, call it **3.7%** — i.e. **$3,700/year**.

**The finding that decides this mandate.** Compare like for like, all dated within nine days:

| | Yield | Date |
|---|---|---|
| US 1-month Treasury par | **3.99%** | 2026-09-23 |
| US 3-month Treasury par | **~4.10%** | 2026-09-23 |
| Sky Savings Rate | 3.60% | undated |
| Aave v3 USDC | 3.57% | 2026-09-15 |
| Tokenized Treasuries (aggregate) | 3.50% | 2026-09-24 |

**[INFERENCE]** The spread of on-chain money-market yield over the 3-month bill is **negative by roughly 40–50 bp**, and negative by roughly 39–42 bp against the 1-month bill. The DAO would be **paying** about half a percent a year for the privilege of taking smart-contract, governance and oracle risk. The ~60 bp gap between the bill yield (4.10%) and the tokenized-treasury aggregate (3.50%) is the observable cost of putting bills on-chain **[INFERENCE — it plausibly also reflects fund mix and shorter paper, which I did not decompose; treat 60 bp as an upper bound on pure fee drag]**.

**And the payoff is trivially small.** Running the two mandates side by side at $25,000/year spend **[INFERENCE, arithmetic verified]**:

| End of year | Mandate A (3.5%) | Mandate B (3.7%) | Difference |
|---|---:|---:|---:|
| 1 | $78,500 | $78,700 | $200 |
| 2 | $56,247 | $56,612 | $365 |
| 3 | $33,216 | $33,707 | $491 |
| 4 | $9,379 | $9,954 | **$575** |

$575 over four years is **about eight days** of development funding. That is what the DAO is buying with its exposure to a Euler-class exploit (§5.4). **[INFERENCE]** At $100,000, the income mandate is not a strategy; it is an unpriced option sold to the protocol layer for eight days of runway.

This conclusion is **rate-regime-dependent and explicitly reversible** — see §7.2.

### 4.3 Mandate C — Long-term growth

**Objective.** Maximise expected terminal real value over 5–10 years, accepting large interim drawdowns.

**Eligible universe.** Mandate A's universe as the reserve sleeve, plus BTC and ETH as the risk sleeve, held in self-custody or at a regulated custodian. Excluded: the DAO's own token (that is a concentration, not an investment, §5.6), governance tokens, anything illiquid, anything requiring active management the DAO cannot staff.

**Illustrative allocation:** 70% cash-equivalents (structured as Mandate A) / 30% risk assets (20% BTC, 10% ETH).

**Why 30% and not 100%.** A treasury with a spend obligation cannot hold 100% risk assets, because withdrawals during a drawdown crystallise losses — the sequence-of-returns problem. The measured 2021–22 drawdown makes this concrete **[INFERENCE, arithmetic verified, using the measured −74.4% BTC figure applied to the whole risk sleeve in year 1]**:

| Year-1 outcome | Risk sleeve | Cash sleeve after spend | Total | vs Mandate A ($78,500) |
|---|---:|---:|---:|---:|
| Risk sleeve −74.4% (**measured** magnitude) | $7,680 | $47,450 | **$55,130** | **−$23,370 ≈ 11 months of funding** |
| Risk sleeve +100% (**assumed**) | $60,000 | $47,450 | **$107,450** | +$28,950 ≈ 14 months of funding |

Note the asymmetry in *evidential status*, not in magnitude: the −74.4% is something that actually happened to this exact asset; the +100% is a number I chose. Anyone comparing these two rows as if they were equally supported is making a category error.

Note also a second-order effect **[INFERENCE]**: in the adverse row the risk sleeve falls to **13.9%** of the portfolio. A mechanical rebalancing rule would then require the DAO to **buy more BTC after a 74% crash**, using cash it needs for payroll. Whether that is the intended behaviour is a founder decision (§6.9), and it must be decided *before* the crash, not during it.

**Honest strength.** Over the Jun-2022-trough window, this mandate would have massively outperformed **[FACT, §3.4]**. The growth case is not stupid. It is simply not compatible with a fixed near-term USD obligation.

### 4.4 Side-by-side

| | A: Preservation | B: Income | C: Growth |
|---|---|---|---|
| Primary risk | Inflation / real erosion | Smart-contract, governance | Market drawdown |
| Expected yield (dated) | 3.0–3.8% | 3.4–3.8% | Reserve sleeve only |
| Worst realistic 1-yr outcome | Depeg loss on one issuer | Total loss of one protocol sleeve | −22% portfolio (measured drawdown × 30%) |
| Runway at $25k/yr spend | ≈4.4 yr | ≈4.4 yr | 3.3 yr to 5+ yr, path-dependent |
| Governance latency tolerated | High | Low — exploits move in minutes | Medium |
| Eligibility/KYC burden | Low–medium | Medium–high | Low |
| Suitable when | Fixed USD obligations, short horizon | Yield spread is genuinely positive | No near-term obligation, long horizon |
| Fails when | Inflation exceeds yield | Spread is negative (**now**) | Cash is needed during a drawdown |

---

## 5. Risk evaluation across the required axes

### 5.1 Liquidity
**[FACT]** USDC is redeemable 1:1 with Circle subject to their process; sUSDS is withdrawable to USDS "at any time" with no lockup or maturity (primary: sky.money). **[FACT]** Ondo OUSG supports $5,000 minimum instant mints/redemptions, but non-instant minimums are $100,000 in and $50,000 out. **[INFERENCE]** For a $100,000 treasury, a $100,000 non-instant subscription minimum means the product is either the whole portfolio or nothing — an all-or-nothing gate that is itself a liquidity risk. **[INFERENCE]** Aave withdrawal liquidity depends on pool utilisation, which is *highest exactly when everyone wants out*; "withdrawable any time" is a property of the contract, not of the market.

### 5.2 Volatility
**[FACT]** Measured maximum drawdowns of −74.4% (BTC) and −81.8% (ETH) over Nov 2021 → Jun 2022. **[FACT]** ETH remains 44.0% below its Nov 2021 peak as of 2026-09-23. **[INFERENCE]** Cash-equivalent sleeves have near-zero price volatility but non-zero *event* volatility: their risk shows up as a discrete jump (a depeg), not as a standard deviation. Treating stablecoins as "zero volatility" in a risk model is the single most common modelling error in DAO treasury management, and any agent-generated risk report must be checked for it.

### 5.3 Custody
**[FACT]** A Safe smart account stores an array of owner addresses and a `uint256` threshold; execution requires at least `threshold` valid, correctly ordered owner signatures (primary: Safe docs). **[FACT]** Safe Foundation's Q1 2026 report put assets secured at roughly **$35.25B across 61M+ deployed accounts** (near-primary).
**[INFERENCE]** The dominant custody risk for a small DAO is not contract failure — it is *signer* failure: too few signers (single point of loss), too many (no quorum reachable in a crisis), signers in one jurisdiction, or signers who cannot be reached within the hours an incident allows. A 3-of-5 threshold with geographically distributed signers and a documented succession plan is a reasonable starting point **[ASSUMPTION — my judgement; no source prescribes this]**.
**[INFERENCE]** Blind signing is the live threat: a signer approving a transaction they cannot independently decode gets the security of a 1-of-1 wallet with the inconvenience of a 3-of-5. Any delegation to an agent must make transactions *human-verifiable*, not merely human-approved.

### 5.4 Smart-contract risk
Two dated incidents define the shape of this risk.

**[FACT] Euler Finance, 2023-03-13, $197M.** The `donateToReserves` function, added to the eToken contract in an earlier upgrade, let a caller reduce their own collateral without reducing their debt and — unlike every other balance-changing function — **never checked caller solvency**. An attacker flash-loaned ~$30M DAI from Aave, donated to force an artificial insolvency, and self-liquidated. Losses included ~$135.8M stETH, ~$33.85M USDC, ~$18.5M WBTC, ~$8.75M DAI. The attacker later returned substantially all the funds.

**[FACT] Curve / Vyper, 2023-07-30, ~$70M** (reported figures range ~$45–70M across sources; the spread is itself a fact about incident reporting). The **Vyper compiler** versions 0.2.15, 0.2.16 and 0.3.0 mis-assigned storage slots, breaking the `@nonreentrant` lock. Affected pools included crv/eth, aleth/eth, mseth/eth, peth/eth.

**[INFERENCE] These two incidents teach opposite and equally important lessons.** Euler was an *audited, mature, well-reviewed* protocol with a logic gap in a recently added function — so protocol maturity is not protection. Curve was a *compiler* bug — so auditing the contract source would not have found it, and the risk was shared silently across every protocol using those compiler versions. Together they mean smart-contract risk is **neither diversifiable by splitting across protocols** (shared dependencies) **nor eliminable by audit** (audits miss things, and the bug may be below the audited layer). It can only be *sized*. For a $100,000 treasury being paid a negative spread to take it (§4.2), the correct size is small or zero.

### 5.5 Stablecoin depegging
**[FACT]** On 2023-03-10/11, Circle disclosed **$3.3 billion** of USDC reserves — approximately **8% of total reserves** — trapped at the failed Silicon Valley Bank. Circle's 2023-03-12 release stated the reserve then stood at 77% T-bills ($32.4B) and 23% cash ($9.7B), and that the SVB funds would be available when US banks opened on **2023-03-13**. USDC traded below $1.00 during the episode and the de-peg subsequently closed.
**[UNKNOWN]** Circle's own releases do **not** state the low print. Third-party reports commonly cite ~$0.87–0.88, but I did not verify this against a primary source and **do not rely on it**.

**[INFERENCE] The generalisable lesson is about proportionality.** An 8% exposure to one failing bank was sufficient to break the peg intraday, even though the reserve was overwhelmingly sound and holders were ultimately made whole. Depeg magnitude is driven by *uncertainty about redemption*, not by the arithmetic size of the impairment. Therefore: (a) a DAO holding one issuer has 100% exposure to that issuer's redemption path regardless of how diversified the issuer's own reserves are; (b) the correct response to a depeg is usually **not to sell** — selling at a 10% discount **[ASSUMPTION]** on a $70,000 stablecoin sleeve realises a **$7,000 loss**, about 3.4 months of funding, precisely when the fundamental thesis (reserves are sound) is most likely still intact. This is why "do nothing" must be an explicit, pre-authorised state (§6, §7).

**[FACT]** The GENIUS Act (2025-07-18) improves the structural picture — 1:1 permitted reserves, holder priority in issuer bankruptcy — but explicitly provides that payment stablecoins are **not federally insured**. Regulation reduces the probability of impairment; it does not eliminate it, and it does not prevent a market de-peg driven by fear.

### 5.6 Concentration
**[INFERENCE]** For this treasury the relevant concentrations are, in descending order of importance:
1. **Issuer concentration** — one stablecoin issuer is one redemption path, one banking stack, one regulator. This is the concentration that actually bites (§5.5), and it is invisible on a portfolio report that lists "USDC, USDC, USDC" as three different positions because they sit in three different venues.
2. **Native-token concentration** — if any part of the treasury is the DAO's own token, treasury value and the DAO's need for treasury value are perfectly correlated: both collapse together. See §9.1, which is unresolved for this DAO.
3. **Chain concentration** — all assets on one chain share one halting risk and one bridge risk.
4. **Signer concentration** — see §5.3.
5. **Dependency concentration** — the Curve/Vyper incident (§5.4) shows that protocols which look independent can share a compiler.

**[INFERENCE]** Conventional position-size limits address only the least important item on this list.

### 5.7 Transaction costs
**[FACT]** $0.004 per transfer / $0.06 per swap at 0.063 gwei (2026-09-24 02:31 UTC); ~$0.59 per transfer at 3.311 gwei (2026-09-23).
**[ASSUMPTION]** Spread plus slippage on $10,000–$25,000 stablecoin clips in major venues: roughly 1–5 bp. I found no primary source for depth at these sizes and this number should be replaced with measured fills before any policy relies on it.
**[FACT/INFERENCE]** Product fee drag, measured as the gap between the 3-month bill (4.10%) and the tokenized-treasury aggregate APY (3.50%), is about **60 bp/year** — an upper bound, since mix effects are not decomposed.

**[INFERENCE] Ranking, which is the useful output:** fee drag (~60 bp/yr, recurring) ≫ spread/slippage (~1–5 bp per trade) ≫ gas (~2.4 bp/yr at four rebalances). **The costs that matter are the ones embedded in products and charged annually, not the ones visible at the moment of clicking.** An agent optimising for gas while ignoring a 60 bp management fee is optimising the wrong quantity by a factor of about twenty-five.

### 5.8 Eligibility and the legal wrapper — the constraint that actually binds
**[FACT]** BUIDL carries a **$5 million minimum** and requires **qualified purchaser** status, typically evidenced by a letter from a CPA, attorney or broker-dealer attesting to $5M+ in investments.
**[FACT]** OUSG requires **accredited investor** status (SEC thresholds: $200K income, $300K joint, or $1M net worth excluding primary residence), with a $5,000 instant-mint minimum.
**[INFERENCE]** A $100,000 treasury is **categorically ineligible for BUIDL** — not merely too small to bother, but excluded by rule. This single fact eliminates the largest and most-cited institutional product from the universe, and it is the kind of constraint that an agent optimising on yield and AUM would silently violate.

**[FACT]** Wyoming's **DUNA** Act (SF0050) was signed **2024-03-07**, effective **2024-07-01**. The following are quoted from the enrolled bill text, which I downloaded and parsed directly (primary):

- **Membership floor.** A decentralized unincorporated nonprofit association must "`Consist[] of at least one hundred (100) members joined by mutual consent under an agreement, that may be in writing or inferred from conduct, for a common nonprofit purpose`."
- **Falling below the floor is a dissolution trigger.** The statute provides for dissolution "`If membership in the nonprofit association falls below one hundred (100) members and the decentralized unincorporated nonprofit association does not meet the requirements of a Wyoming unincorporated nonprofit association under W.S. 17-22-101 through 17-22-115`."
- **Liability shield.** "`A person is not liable for a breach of a decentralized unincorporated nonprofit association's contract merely because the person is a member, administrator, authorized to participate in the management of the affairs of the nonprofit association or considered as a member`," with a parallel provision for "`a tortious act or omission`."
- **Smart contracts are recognised governance instruments.** The Act defines "smart contract" by reference to W.S. 40-21-102(a)(ii) and contemplates governance by "`association agreements, consensus formation algorithms, smart contracts or enacted governance proposals`."

**[INFERENCE]** A wrapper is a precondition, not an afterthought: without a legal entity the DAO cannot open the bank or brokerage account that Mandate A's 35% off-chain sleeve assumes, cannot complete KYC for tokenized funds, and leaves members personally exposed in a way the DUNA liability shield would otherwise close. **If the founders will not form an entity, Mandate A's illustrative allocation is not implementable as written and collapses toward a fully on-chain portfolio with higher issuer concentration.**

**[INFERENCE] The 100-member floor is a live treasury risk, not a formality.** A small project DAO that drifts below 100 members does not merely lose a tax status — it hits a statutory dissolution trigger, at which point custody of a $100,000 treasury sits with an entity whose legal existence is in question and whose members may have just lost their liability shield. This belongs in §7.3 as a human-review event, and it is the kind of constraint that no portfolio-optimisation process would ever surface.

**[UNKNOWN]** Whether a DUNA is the right wrapper for this specific DAO, how the "reasonable compensation" allowance interacts with paying contributors from the treasury, and the tax treatment of investment income are questions for counsel, not for this report.

---

## 6. Decisions the founders must make before any asset is selected

These are ordered by how much they constrain everything downstream. None can be delegated to an agent, because each is a statement of preference, not a computation.

1. **Endowment or runway?** Perpetual capital spending only income, or a depleting pot with a target end date. At $100,000 the honest answer is almost certainly runway (§1), and saying so out loud changes the whole design.
2. **Spend rate, and whether it is fixed or flexible.** $25,000/year fixed, or 8% of NAV recalculated quarterly? A percentage-of-NAV rule makes the treasury self-stabilising and makes the growth mandate survivable; a fixed-dollar rule makes drawdowns compound into insolvency.
3. **Unit of account.** USD, ETH, or developer-hours. This determines what "risk-free" means. A DAO whose costs are USD invoices but whose treasury is ETH is running an unhedged FX book whether or not it calls it that.
4. **Liability profile.** Are obligations contractual (invoices with dates) or discretionary (grants that can pause)? Contractual liabilities inside the horizon rule out Mandate C.
5. **Legal wrapper and tax.** Entity or not; jurisdiction; who signs contracts; whether trading gains create a tax liability the treasury must reserve for. Gates §5.8.
6. **Custody design.** Threshold, signer identities, jurisdictional and organisational diversity, key-loss and succession procedures, and whether signers can actually decode what they sign.
7. **Eligibility ceiling.** Can the entity pass accredited-investor or qualified-purchaser checks? Answering "no" removes a large part of the universe immediately and honestly.
8. **Governance latency.** How many hours or days from "we should act" to "the transaction is executed"? This is a **risk parameter, not an administrative detail**: a protocol exploited in ten minutes cannot be escaped by a seven-day vote. Latency should be measured, in a drill, before it is relied on.
9. **Rebalancing intent.** Does the DAO *want* a rule that buys more of an asset that just fell 74%? Decide before, not during (§4.3).
10. **Scope of agent delegation.** Advisory (propose, humans decide) / bounded execution (act inside a pre-approved universe, humans can veto) / discretionary (act freely). And the question that must be answered before any of these: **who is accountable when an agent loses money?**
11. **Stated maximum acceptable drawdown**, in dollars, written down and signed before deployment. Not a volatility target — a number that, if breached, triggers a pre-agreed response.
12. **Reporting and benchmark.** What the treasury will be compared against, how often, and who publishes it (§7.4, §8.2).

---

## 7. Policy: rebalancing, no-action, human review, and falsification

### 7.1 Rebalancing rules
**[ASSUMPTION — a proposed policy, not a derived result]**

- **Bands, not calendars.** Rebalance a sleeve only when it deviates by more than **±5 absolute percentage points** of total NAV *or* **±25% relative** to its target, whichever binds first.
- **Evaluate weekly, act monthly at most.** Cadence limits decision churn, which at this size costs more in error than gas costs in fees.
- **Cash-flow rebalancing first.** Direct inflows and spending to the most-underweight and most-overweight sleeves respectively before placing any trade. At a $25,000/year spend on a $100,000 book, this alone handles most drift for free.
- **Minimum trade size $2,500** **[ASSUMPTION]**, to keep per-trade fixed costs and review effort proportionate.
- **Never rebalance and change policy in the same transaction.** Policy changes get their own vote and their own record.

### 7.2 Conditions requiring no action
This section exists because **"do nothing" is a decision that must be explicitly authorised in advance**, or it will be overridden by whoever feels the most urgency. No action is the correct output when:

- Deviation is inside the band.
- Estimated round-trip cost exceeds the estimated benefit of correcting the deviation.
- **A stablecoin is trading below peg and the solvency thesis is intact.** Selling into a panic bid converts a temporary discount into a permanent loss (§5.5). Overriding this requires evidence that reserves are actually impaired, not that the price is low.
- A price feed is stale by more than one hour, or two independent feeds disagree by more than 50 bp.
- Within 48 hours of a governance parameter change in a protocol the DAO holds — let the change settle and be observed.
- During an active, unresolved security incident in a protocol the DAO holds, *unless* the action is a withdrawal to the safe sleeve.
- When the proposed action originates from an agent whose reasoning a human signer cannot follow. Illegibility is a veto.

### 7.3 Events requiring human review (agent may not act autonomously)
- Any stablecoin held trades below **$0.995** for more than 60 minutes.
- An issuer's monthly attestation is late by more than 5 business days, or an attestation is qualified.
- Any exploit above **$10M** in any protocol the DAO holds, or in any protocol sharing a known dependency with one it holds (the Curve/Vyper lesson, §5.4).
- A governance proposal in a held protocol touching collateral composition, oracles, upgrade admin, or the savings rate.
- Single-day NAV move greater than **3%** under Mandates A or B, or **10%** under Mandate C.
- Any change to the signer set or threshold.
- Any redemption gate, withdrawal suspension, or pause event at any venue.
- Regulatory action against an issuer, custodian, or venue.
- **Membership falls below the legal wrapper's statutory floor** — for a Wyoming DUNA, below 100 members, which is a dissolution trigger (§5.8). Review the wrapper before reviewing the portfolio.
- An agent proposes anything outside the eligible universe — this is treated as a **defect report**, investigated, and not merely rejected.
- Cumulative drawdown breaches the stated maximum from §6.11.

**Kill switch.** Any **two** signers may move the entire treasury to idle cash-equivalents without a governance vote. Re-deployment always requires a full vote. The asymmetry is deliberate: de-risking should be fast and cheap, re-risking should be slow and deliberate.

### 7.4 What would invalidate each recommendation
Stated as falsifiers, so this report can be checked rather than believed.

**Mandate A (preservation) is invalidated if:**
- Short bill yields fall below ~1%, making the opportunity cost of on-chain alternatives immaterial and the real-erosion problem dominant.
- The DAO's liabilities turn out not to be USD-denominated (§6.3).
- Persistent inflation above the nominal yield makes real capital loss certain, and the DAO's horizon is long enough for that to compound meaningfully.
- The DAO cannot or will not form a legal entity, which removes the off-chain sleeve (§5.8).

**Mandate B (income) is invalidated — or rather, becomes valid — if:**
- The on-chain yield spread over duration-matched bills turns **durably positive by more than 150 bp** with unchanged protocol risk. **[ASSUMPTION — 150 bp is my threshold for "paid enough to take exploit risk at this size"; founders should set their own, but they must set it in advance.]** As of 2026-09-23 the spread is about **−45 bp**, so this mandate currently fails its own test.
- Equally: it is *further* invalidated if a major money-market protocol suffers a loss event, or if the treasury grows large enough that a 40 bp spread is worth real money — note this cuts *both* ways depending on which happens.

**Mandate C (growth) is invalidated if:**
- The DAO has any fixed USD obligation inside the horizon (§6.4).
- Governance latency (§6.8) exceeds the time in which de-risking would need to happen.
- The DAO cannot honestly affirm it would hold through an 80%+ drawdown — and the test of that affirmation is behavioural, not verbal.
- Conversely, it is *strengthened* if the DAO adopts a percentage-of-NAV spend rule (§6.2), which largely defuses the sequence-of-returns objection in §4.3.

**All three are invalidated if** the treasury turns out to be denominated in the DAO's own token (§9.1). In that case the entire analysis is replaced by a diversification problem, and the questions become: over what schedule, through what venue, with what market impact, and with what disclosure.

**This report as a whole is invalidated if** the rate environment shifts materially. Every quantitative conclusion here rests on the 2026-09-22/23 curve. **Re-run it if the 3-month bill moves more than 100 bp in either direction.**

---

## 8. The paper-portfolio experiment

### 8.1 Design principles
The experiment exists to answer one question: **does delegating portfolio decisions to IMD Swarm produce better decisions than doing nothing, after honest costs?** It is designed so that the answer can come back "no."

**Pre-registration.** Hypotheses, benchmark, allocation rules, cost model, duration and stopping rules are published and hash-committed on-chain **before** the first decision. Nothing may be changed mid-flight without a dated, published amendment that is itself part of the record. Without pre-registration, a 26-week experiment on three mandates will always find something that looks like a win.

### 8.2 Benchmarks
- **Primary benchmark: 100% USDC held idle in the Safe, zero yield, zero transactions.** This is the true counterfactual — it is literally what the DAO does if nobody does anything. Any strategy that cannot beat doing nothing, after costs and after the cost of running the swarm, has failed regardless of its absolute return.
- **Reference benchmark: rolling 13-week Treasury bill**, marked from the Treasury par yield series. This is the "you cannot beat this without taking risk" line.
- **Cost-of-operation benchmark.** The swarm's own cost (agent compute, oracle panels, human review hours) is subtracted from the strategy's return. **[INFERENCE]** This is the benchmark most treasury experiments omit, and it is the one most likely to flip a modest outperformance into a loss.

### 8.3 Protocol
| Parameter | Value |
|---|---|
| Duration | 26 weeks, 2026-09-28 → 2027-03-29 **[ASSUMPTION]** |
| Notional | $100,000 paper, three parallel mandate portfolios plus two benchmarks |
| Decision cadence | Weekly evaluation; action only when §7.1 bands trigger |
| Capital at risk | **Zero.** Paper only. |

**Timestamped decisions (commit–reveal).** Before any paper execution, the swarm publishes `keccak256(decision ‖ rationale ‖ nonce)` on-chain and records the block height. The plaintext is revealed after the execution price is fixed. This makes look-ahead bias and after-the-fact rationalisation **structurally impossible rather than merely discouraged** — which matters more when the decision-maker is a language model that can produce a fluent justification for any position.

**Execution pricing.** Paper fills are struck at the **next hourly close after the reveal**, never at the decision-time price. Fills are priced at mid ± half the observed spread, plus a slippage estimate scaled to clip size, plus gas at the actual base fee of the fill block. **Mid-price fills are prohibited** — they are the standard way paper portfolios flatter themselves.

**Disagreement record.** Every proposal is assessed by an N-of-M oracle panel. The record retains, for each panel member: the verdict, the stated reasoning, and the dissent if any. **Dissent is published even when overruled.** Over the 26 weeks, dissenter calibration is tracked: when the panel split, who was right? **[INFERENCE]** This is the highest-value artefact of the whole experiment. Whether a swarm can pick assets is doubtful and barely matters at $100,000; whether a swarm's disagreement structure carries information about decision quality is both testable here and valuable at any size.

**Recorded per decision:** decision ID, commit hash and block, reveal timestamp, proposing agent ID, mandate, action, size, rationale, stated confidence, expected cost, panel verdicts, dissents, realised fill price, realised cost, and the NAV before and after.

### 8.4 Declared success and failure criteria
The experiment is declared a **success** only if all four hold:
1. The audit trail is complete and independently reconstructible from public data.
2. Every decision has a commit predating its reveal and its fill. (A single violation fails the experiment outright — the integrity property is binary.)
3. At least one mandate beats the idle-USDC benchmark by more than the cost of operating the swarm.
4. The disagreement record shows measurable signal in dissent.

The experiment is declared a **failure** if it produces returns without a clean audit trail. **[INFERENCE]** A good number from an unverifiable process is worse than a bad number from a verifiable one, because it will be believed.

**[INFERENCE]** The most likely outcome, stated in advance: over 26 weeks, three cash-heavy mandates will differ from the benchmark by a few hundred dollars — an amount indistinguishable from noise at this notional. **The experiment should therefore be judged primarily on process integrity, not on return.** Anyone who designs it to be judged on return has designed it to be gamed.

---

## 9. IMD Swarm: what exists today, what must be built

### 9.1 The unresolved question about the treasury itself

**[FACT — unofficial source]** The IMD swarm-ledger documentation describes launches allocating tokens to wallets via **`MerkleDistributor`** contracts, with claim status queried as `claimed(0, wallet)`, a root frozen one-shot by `openClaims()` after `setRoot()`, and a **`sweepDelay` "currently 365 days from opening"** after which unclaimed tokens are swept **to the treasury**.

**[INFERENCE]** The only IMD treasury mechanism I could observe accumulates **unclaimed launch tokens**, not dollars. If that is the treasury the brief means, then it is denominated in the project's own token and:
- Treasury value and the DAO's need for treasury value are perfectly correlated (§5.6).
- The "$100,000" is a mark-to-market on a token with unknown depth, not a spendable balance.
- The first action is a diversification schedule with disclosed market impact, and mandate selection is a *second-phase* question.

**[UNKNOWN]** I could not determine from public sources whether the hypothetical $100,000 is native-token-denominated, stablecoin-denominated, or purely notional. **This is the first question the founders should answer, and it should be answered before anything in §4 is acted on.** I have written §4–§7 for a dollar-denominated treasury because that is the literal reading of the brief, and flagged the dependency here rather than silently assuming it away.

### 9.2 Capability assessment

Evidence base: explorer.imd.fun (primary, read 2026-09-23), the Identity-md/worker README (primary), the EIP-8004 specification (primary), and two explicitly **unofficial** community repositories (swarm-ledger, imd-node-guide) that I treat as indicative only.

**[FACT]** Observed on explorer.imd.fun, 2026-09-23: 54 total jobs (3 running, 9 incomplete, 42 completed); **262 agents online**, 3 working; **2,669 submissions accepted in the last 24 hours**; 3 oracles running. Job categories include smart-contract development with "independent adversarial reviews," Sepolia deployment, IPFS publication, and — notably — **"DAO treasury proposals"** as an existing research category.

**[FACT]** From the worker README: pairing requires "the wallet that owns an eligible IdentityMD NFT," registering "that token as an ERC-8004 agent"; "An unregistered token cannot connect for work. One NFT authorizes one active device"; runtime selection via `--runtime claude` or `--runtime codex`; completed results held in a local outbox for reconnection resilience.

**[FACT]** ERC-8004 "Trustless Agents," created **2025-08-13**, status **Draft**, defines three singleton registries: **Identity** (ERC-721-based portable identifier), **Reputation** (posting and fetching feedback signals), **Validation** (hooks for independent validator checks — stake-secured re-execution, zkML, TEE oracles).

**[FACT — unofficial]** Task lifecycle observed as `accepted → working → submitted → stored` pending a verdict that arrives later; `oracle_assess` described as "the high-volume panel task" producing an `answer.json`; tiers economy / standard / premium mapping to model size and effort; "Rejected attempts do not hurt your standing. Bad reviews do."

**Supported today [INFERENCE from the above]:**

| Capability | Evidence |
|---|---|
| Job decomposition and distribution | explorer: 54 jobs across research/dev/frontend categories |
| Cryptographic agent identity | ERC-8004 registration; one NFT → one active device |
| Multi-agent parallel execution at scale | 262 agents online; 2,669 accepted submissions/24h |
| Independent adversarial review | `oracle_assess` panels; "independent adversarial reviews" in job specs |
| Structured verdicts on work | `accepted → … → verdict` lifecycle; `answer.json` |
| Reputation / track record | seats, accepted/rejected counts, hours, rank |
| Effort tiering by task value | economy / standard / premium |
| Merkle-based payout with claim tracking | `MerkleDistributor`, `claimed(0, wallet)` |
| A treasury address receiving sweeps | `sweepDelay` 365 days |
| Public provenance | read API + explorer + GitHub JSON snapshots |
| Research report production | this report is itself an instance |

**[INFERENCE]** The fit is better than it first appears, but in an unobvious way. IMD's genuine strength for this problem is **not** portfolio construction — it is the **adversarial panel plus verdict plus reputation** stack, which maps almost exactly onto the §8.3 disagreement record. The swarm is already an apparatus for producing, contesting and scoring opinions with attributable identity. That is the scarce component of a trustworthy treasury process. Asset picking is the part that is both easy to automate and nearly worthless at $100,000.

**Requires new software [INFERENCE]:**

| Gap | Why IMD does not cover it |
|---|---|
| Price/NAV oracle with staleness and deviation guards | Oracles here assess *work quality*, not market data |
| Portfolio state and NAV accounting | No position ledger exists |
| Benchmark computation | No external return series ingestion |
| Execution-cost model | No concept of fills, spreads or slippage |
| Policy/band engine incl. explicit no-action | No policy representation |
| Commit–reveal decision timestamping | Verdicts are timestamped; *decisions* are not committed pre-execution |
| Structured disagreement ledger | Reputation registry stores feedback signals, not per-decision dissent with reasoning |
| Safe Module + Guard enforcing the eligible universe on-chain | No custody integration |
| Circuit breaker / kill switch | No such primitive |
| Human-review queue with escalation SLAs | Review is of submissions, not of live risk events |
| KYC / accreditation attestation | Nothing addresses §5.8 eligibility gates |
| Post-trade reconciliation | No settlement layer |

**[INFERENCE — a risk of the approach, not of the assets]** ERC-8004 remains a **Draft** standard (created 2025-08-13, no Last Call or Final date established **[UNKNOWN]**). Its Validation Registry is the natural home for oracle verdicts in this design, so building the treasury system on it is a bet on a moving specification. That is acceptable for a paper experiment and would need re-examination before any custody integration.

---

## 10. Open questions

1. **Is the treasury dollar-denominated or native-token-denominated?** (§9.1) Blocks everything. Unresolvable from public sources.
2. **What is the actual development spend rate?** All runway arithmetic assumes $25,000/year.
3. **Can the DAO form a legal entity and pass KYC?** (§5.8) Gates ~35% of Mandate A and all tokenized-treasury exposure.
4. **What is measured governance latency?** (§6.8) Should be established by drill, not estimate.
5. **Current USDC reserve line-item composition.** Circle's charts did not parse; the PDF attestation was not retrieved (§3.3).
6. **The 7 bp 3-month yield discrepancy** between the Treasury par curve and H.15 (§3.1). Probably extraction error; unconfirmed.
7. **Real stablecoin market depth at $10,000–$25,000 clip sizes.** §5.7 uses an assumption where it needs a measurement.
8. **Who is accountable when a delegated agent loses money?** (§6.10) A governance and legal question this report cannot answer.
9. **The exact USDC low print in March 2023.** Widely reported, not primary-sourced, deliberately not relied upon (§5.5).
10. **Are the two Aave readings** (3.57% / 3.74%) the same market? (§3.2) Not reconciled.

---

## 11. Sources

All URLs were retrieved on 2026-09-23 or 2026-09-24, and all 33 were checked programmatically on 2026-09-24.

**Verification status, disclosed.** 28 of 33 returned HTTP 200 to an automated request. Five returned **403 to automated clients** (bot protection, not dead links): congress.gov ×2, etherscan.io, defillama.com, help.coinbase.com. Of these, Etherscan was read successfully through a browser-equivalent fetch (the gas figures in §3.5 are from that read). The **congress.gov pages could not be read directly at all**, so the GENIUS Act facts in §3.3 rest on (a) the CRS product title on congress.gov, which supplies the public law number P.L. 119-27, and (b) independent corroboration from Latham & Watkins, which I did read directly and which confirms the 2025-07-18 signing date, the permitted reserve list, holder priority in insolvency, and that payment stablecoins are neither federally insured nor classified as securities or commodities. Readers wanting the statute itself should go to the bill text rather than relying on this report.

**Primary — government and issuers**
- US Treasury, Daily Treasury Par Yield Curve Rates, September 2026 — https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?type=daily_treasury_yield_curve&field_tdr_date_value_month=202609
- US Treasury, Daily Treasury Bill Rates, 2026 — https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?type=daily_treasury_bill_rates&field_tdr_date_value=2026
- Federal Reserve, H.15 Selected Interest Rates, released 2026-09-23 — https://www.federalreserve.gov/releases/h15/
- Congress.gov / CRS, "Stablecoin Legislation: An Overview of the GENIUS Act of 2025 (P.L. 119-27)" — https://www.congress.gov/crs-product/IN12553
- S.1582, 119th Congress, GENIUS Act — https://www.congress.gov/bill/119th-congress/senate-bill/1582
- Wyoming SF0050 (2024), Decentralized Unincorporated Nonprofit Association Act — https://www.wyoleg.gov/2024/Introduced/SF0050.pdf *(46-page PDF downloaded and text-extracted; §5.8 quotes the bill text directly)*
- Circle, "USDC" (circulation $74.6B as of 2026-09-21) — https://www.circle.com/usdc
- Circle, Transparency (attestation as of 2026-09-21) — https://www.circle.com/transparency
- Circle, "$3.3 Billion of USDC Reserve Risk Removed, Dollar De-peg Closes" (March 2023) — https://www.circle.com/pressroom/3-3-billion-of-usdc-reserve-risk-removed-dollar-de-peg-closes
- Circle, "An Update on USDC and Silicon Valley Bank" — https://www.circle.com/blog/an-update-on-usdc-and-silicon-valley-bank
- Sky, sUSDS product page (3.60% SSR, undated) — https://sky.money/susds
- Ondo Finance, Qualified Access Products — Eligibility — https://docs.ondo.finance/qualified-access-products/eligibility
- Securitize, BlackRock BUIDL — https://securitize.io/blackrock/buidl
- Safe, Smart Account Concepts (owners array, `uint256 threshold`) — https://docs.safe.global/advanced/smart-account-concepts
- Ethereum Foundation, ERC-8004: Trustless Agents (Draft, created 2025-08-13) — https://eips.ethereum.org/EIPS/eip-8004

**Primary — IdentityMD**
- IMD Explorer, jobs and network statistics — https://explorer.imd.fun/
- Identity-md/worker, worker CLI README — https://github.com/Identity-md/worker

**Unofficial — IdentityMD community (indicative, not authoritative)**
- johnfreeman777/swarm-ledger (MerkleDistributor, `sweepDelay`, seats) — https://github.com/johnfreeman777/swarm-ledger
- johnfreeman777/imd-node-guide (`oracle_assess`, task lifecycle) — https://github.com/johnfreeman777/imd-node-guide

**Near-primary and secondary**
- rwa.xyz, Tokenized U.S. Treasury Funds (2026-09-24) — https://app.rwa.xyz/treasuries
- Aavescan, USDC on Ethereum V3 — https://aavescan.com/ethereum-v3/usdc
- DeFiLlama, USDC (Aave V3 Ethereum) — https://defillama.com/yields/pool/aa70268e-4b52-42bf-a116-608b370f9501
- Etherscan Gas Tracker (2026-09-24 02:31 UTC) — https://etherscan.io/gastracker
- Chainalysis, Euler Finance flash loan attack — https://www.chainalysis.com/blog/euler-finance-flash-loan-attack/
- Veridise, "Insolvency Vulnerability: The Euler Finance $197M Exploit" — https://veridise.com/blog/audit-insights/insolvency-vulnerability-euler-finance-exploit/
- Vyper, "Nonreentrancy Lock Vulnerability Technical Post-Mortem" — https://hackmd.io/@vyperlang/HJUgNMhs2
- Chainalysis, Curve Finance liquidity pool hack — https://www.chainalysis.com/blog/curve-finance-liquidity-pool-hack/
- a16z crypto, "The DUNA: An Oasis For DAOs" — https://a16zcrypto.com/posts/article/duna-for-daos/
- Latham & Watkins, "The GENIUS Act of 2025" — https://www.lw.com/en/insights/the-genius-act-of-2025-stablecoin-legislation-adopted-in-the-us
- Fortune, price of Bitcoin / Ethereum, 2026-09-23 — https://fortune.com/article/price-of-bitcoin-09-23-2026/ , https://fortune.com/article/price-of-ethereum-09-23-2026/
- Safe Foundation, smart contracts and audits — https://safefoundation.org/smart-contracts
- Coinbase Help, USDC rewards overview — https://help.coinbase.com/en/coinbase/coinbase-staking/rewards/usd-coin-rewards-faq

---

## Appendix A — Reproducing the arithmetic

```python
# Runway, Mandates A and B, $25,000/yr spend drawn at year-end
def runway(rate, spend=25_000, v=100_000, years=5):
    out = []
    for _ in range(years):
        v = v * (1 + rate) - spend
        out.append(round(v))
    return out

runway(0.035)  # A: [78500, 56247, 33216, 9379, -15293]
runway(0.037)  # B: [78700, 56612, 33707, 9954, -14678]
# End-of-year-4 difference: $575 = 8.4 days at $25,000/yr

# Mandate C, adverse year 1 (measured BTC drawdown -74.4% on the risk sleeve)
risk = 30_000 * (1 - 0.744)      #  7,680
cash = 70_000 * 1.035 - 25_000   # 47,450
total = risk + cash              # 55,130  -> risk weight 13.9%
# Gap vs Mandate A: 78,500 - 55,130 = 23,370 = 0.93 years of funding

# Annualised returns, start-date dependent
cagr = lambda a, b, y: (b / a) ** (1 / y) - 1
cagr(68_906.48, 85_686.06, 4.87)  # BTC from Nov-2021 peak:  +4.58%/yr
cagr(4_865.462, 2_723.35, 4.87)   # ETH from Nov-2021 peak: -11.23%/yr
cagr(17_614.34, 85_686.06, 4.26)  # BTC from Jun-2022 trough: +45.0%/yr
cagr(883.159,  2_723.35, 4.26)    # ETH from Jun-2022 trough: +30.3%/yr

# Gas, ERC-20 transfer at 3.311 gwei, ETH = $2,723.35
65_000 * 3.311e-9 * 2_723.35      # $0.586
```

---

**Reminder on scope.** This is a research deliverable. No capital was deployed, no transaction was executed, no key was held or requested, and no allocation here should be acted on without the founder decisions in §6 and independent legal, tax and investment advice.