Stablecoins: From Pegged IOUs to Ungovernable Money

The design space, the graveyard, the fork-choice problem, and the flatcoin frontier — and why the most successful stablecoins gave up nearly everything crypto was for.

tl;dr

• The dream was bearer dollars without a banker. We mostly built bankers without the bearer dollars. The two coins that won — USDT ($189.5B) and USDC ($76.8B) — are 82.7% of a $321.9B market (DefiLlama, 2026-05-22) and are permissioned IOUs with corporate freeze switches, bank-dependent reserves, and KYC-gated redemption. Fiat-backed stables are a distribution success, not an innovation. The innovation lives in the immutable, crypto-collateralized line: SAI → LUSD → RAI → the flatcoin frontier.
• The design space has three families and one trilemma. Crypto-overcollateralized (BitUSD, SAI/DAI, LUSD), fiat-backed (USDT, USDC), and undercollateralized/algorithmic (Basis, ESD, UST). You can pick at most two of {decentralized, capital-efficient, robustly pegged}. Every blowup is a system that pretended otherwise.
• MakerDAO is the cautionary arc, not a hero story. ETH-only SAI (2017) → multi-collateral DAI (2019) → the Peg Stability Module (2020) → ~60% USDC-collateralized by 2022. The most successful “decentralized” stable progressively became wrapped USDC, and after the SVB depeg its governance voted ~79% to keep USDC anyway. Decentralization decays under the gravity of capital efficiency.
• The failure modes are a clean taxonomy, not random bad luck. Algorithmic reflexivity/death spiral (UST, IRON/TITAN, the seigniorage wave), collateral risk (fUSD, Frax↔SVB), bank-run/coordination failure (IRON), governance/oracle exploit (Beanstalk, IRON’s TWAP), and yield-subsidy unsustainability (Anchor’s 20%, OHM’s rebase APY). Durable pegs need exogenous uncorrelated collateral + a redemption mechanism that closes the arbitrage under stress + organic demand. Every casualty was missing at least one.
• Haseeb called the fork-choice problem in 2019; the Merge proved it in 2022. The canonical essay is Qureshi & Lee, “Ethereum is now unforkable, thanks to DeFi” (Oct 31, 2019) — written about a hypothetical ProgPoW fork, not ETHPoW. The claim: in a contentious fork, Circle/Tether decide which chain their stable is redeemable on, DeFi composability forces everyone to follow, and that becomes the Schelling point. The Sept 2022 ETHPoW fork was the live experiment: Circle and Tether pre-committed to PoS-only, and ETHW collapsed to single digits. Two private companies’ redemption policies ratified canonical Ethereum.
• The incumbents’ failings are now empirically proven on both ends. USDC inherited a bank run (SVB, March 2023, traded to $0.8774 with $3.3B trapped). Tether has never published a full audit and was fined by the CFTC for being backed only 27.6% of the time in a 2016–2018 sample. And on censorship: the courts found Tornado Cash’s code couldn’t be sanctioned (Fifth Circuit 2024 → delisting March 2025), yet the centralized stables flowing through it were frozen instantly — on a sanction that didn’t survive review. The protocol was more censorship-resistant than the “dollars.”
• Regulation is entrenching the centralized duopoly, not correcting it. The GENIUS Act (signed July 18, 2025) rewards the USDC/Tether model — 100% cash/Treasury reserves, licensed issuers, monthly disclosure — and bans paying yield to holders, which structurally disadvantages decentralized designs. MiCA did the same in the EU and pushed USDT off regulated venues.
• The flatcoin frontier is where the original dream survives — but the scaling question is unsolved. “Flatcoin” has a strict sense (pegged to purchasing power / a CPI basket — Balaji 2021 → Coinbase 2023 → Vitalik) and a loose sense (decentralized, non-fiat-pegged stable-value money — RAI). The genuinely immutable, censorship-resistant designs today are LUSD (Liquity v1) and RAI; everything else is decentralized-but-governed (BOLD, crvUSD, GHO) or not decentralized at all (frxUSD, Ethena USDe). The frontier exemplar is ZAI on Zcash — an oracle-free, ZEC-collateralized, shielded-native CDP — but it is a simulation/research proposal, not deployed. The open problem all of them share: you cannot have decentralized + capital-efficient + scalable at once.

Contents

1. The Design Space: What Were We Actually Trying to Build?
2. The Graveyard: Failure Modes as a Taxonomy
3. The Fork-Choice Problem: Who Decides Canonical Ethereum?
4. The Failings of the Incumbents (USDC / Tether)
5. The Flatcoin Frontier: Ungovernable Money That Actually Works
6. Sources
7. Data Notes & Caveats

1. The Design Space

What were we actually trying to build?

A stablecoin is an attempt to put a stable unit of account on a permissionless ledger. The disagreement — the entire history of the field — is over what you trust to keep it stable. That trust choice sorts every design into one of three families.

Crypto-overcollateralized (the immutable line). Lock volatile crypto worth more than the stable you mint; defend the peg with liquidations and redemption arbitrage.

• BitUSD on BitShares — July 21, 2014, the first stablecoin (Larimer/Hoskinson). Behaved more like a volatile margin instrument and lost its peg in 2018.
• MakerDAO Single-Collateral Dai (SAI) — whitepaper Dec 2017; ETH-only collateral. This is the genuine innovation in the lineage: a dollar you could mint against ETH with no issuer, no bank, no redemption desk.
• Multi-Collateral Dai (MCD) — Nov 18, 2019; the original token renamed SAI.

Fiat-backed (the custodial line). Hold dollars (or T-bills) in a bank; issue a token that’s a claim on them.

• Tether (USDT) — launched as “Realcoin” July 2014 (renamed Tether Nov 2014). USDC (Circle/CENTRE) — announced May 2018, launched Sept 26, 2018.
• These are the winners by size, and the point worth making bluntly: they are not a crypto innovation. They are a wrapper around the existing banking system. The interesting properties of crypto — bearer settlement, censorship resistance, permissionlessness — are exactly the ones they give up.

Undercollateralized / algorithmic (the seigniorage line). Back the coin with nothing exogenous — use a reflexive sister token and supply expansion/contraction to chase the peg.

• Robert Sams’ “Seigniorage Shares” paper (2014) is the intellectual root; Basis/Basecoin the famous early attempt (raised $133M, shut down Dec 13, 2018 over securities concerns); Empty Set Dollar, Frax (fractional) the DeFi-era descendants; UST the catastrophic apex.

The trilemma

You can have at most two of: decentralized, capital-efficient, robustly pegged.

• Fiat-backed: capital-efficient + robustly pegged, not decentralized.
• Overcollateralized CDPs: decentralized + robustly pegged, not capital-efficient (you must lock $150+ to mint $100).
• Algorithmic: decentralized + capital-efficient, not robustly pegged (and the graveyard is the proof).

Haseeb Qureshi’s foundational taxonomy essay frames the peg itself as a Schelling point — “if enough people believe the system will survive, that belief can lead to a virtuous cycle.” That insight is the hinge of the whole analysis: pegs are confidence games, and the design question is whether the mechanism adds confidence under stress or consumes it.

The MakerDAO arc — the cautionary tale, not the victory lap

This is the most important thread in the history, because it shows decentralization decaying under economic gravity:

• Sept 2020: even before the PSM scaled, DAI was already ~61% backed by centralized assets (~52% from centralized stablecoins), because stablecoin collateral needs only ~$101 to mint 100 DAI vs ~$150 in ETH.
• Late 2020: the Peg Stability Module (MIP29) lets anyone swap USDC↔DAI 1:1. Great for the peg; mechanically pulls USDC onto Maker’s balance sheet.
• 2021–22: USDC-backing climbs past 50% via the PSM, peaking around ~60% USDC-collateralized by Aug 2022 (Messari). Critics start calling DAI “wrapped USDC.”
• March 2023 (the test): USDC depegs in the SVB crisis with ~$3B sitting in Maker’s PSM. Governance votes ~79% to keep USDC as the primary reserve anyway. Rune later concedes it’s “almost inevitable” DAI abandons the strict USD peg given this dependence.

The lesson: capital efficiency is a one-way ratchet toward centralization unless you deliberately weld the door shut — which is exactly what LUSD does (§5).

Vitalik’s two tests (the evaluation lens)

From “Two thought experiments to evaluate automated stablecoins” (May 25, 2022, written days after Terra):

• Test 1 — graceful wind-down: “can the stablecoin, even in theory, safely ‘wind down’ to zero users?”
• Test 2 — negative rates: “what happens if you try to peg the stablecoin to an index that goes up 20% per year?” — i.e., can the system express a negative interest rate honestly rather than papering it over with token printing?

He praised RAI as the “ideal type” — ETH-only, floating target — and, in “What in the Ethereum application ecosystem excites me” (Dec 5, 2022), planted the flatcoin seed: a governance-minimized coin could track “a global average CPI index” and advertise “abstract best-effort price stability,” with lower regulatory risk because it isn’t pretending to be a digital dollar.

2. The Graveyard

Failure modes as a taxonomy, not a list of disasters

Don’t read these chronologically — sort them by failure category. There are five, laid out at the end.

Terra UST / LUNA — the canonical death spiral (May 2022)

• Mechanism: $1 algorithmic peg with no exogenous collateral; maintained by mint/burn arbitrage with the volatile sister token LUNA (burn $1 of LUNA → mint 1 UST, and vice versa). “Backing” = the market cap of LUNA itself.
• Demand driver: Anchor Protocol’s ~19.5% APY on UST, structurally subsidized (deposits » borrowing); LFG topped up the reserve with $450M in Feb 2022.
• Failure: when UST depegged and stayed there, redemptions printed astronomical new LUNA — supply went from ~343M (May 9) to ~6.53 trillion in about a week — hyperinflating and destroying the very backstop meant to absorb the run. LFG’s BTC reserve was drained from 80,394 BTC to 313 BTC defending it.
• Dates: depeg May 9; chain halted May 13, 2022. LUNA fell from a $119.51 ATH to ~zero.
• Scale: ~$40–50B notional value destroyed (value wiped, not user clawback).
• Lesson: a peg “backed” by its own reflexive token has no floor; the stabilizer accelerates collapse. A 20% yield that manufactures demand is a liability, not demand.

OlympusDAO (OHM) — not a stablecoin (2021–2022)

• Framing correction: OHM was never pegged. It was a free-floating “reserve currency” with a floor of 1 DAI of backing, routinely trading at a large premium above backing.
• Mechanism: bonding (sell assets/LP to the protocol for discounted vesting OHM → builds Protocol-Owned Liquidity) + staking (rebase rewards at thousands-of-percent APY, funded by dilution). The “(3,3)” meme: if everyone stakes, everyone wins; the reflexive loop pumps on the way up and reverses violently on the way down.
• History: launched March 23, 2021; ATH $1,415 (Apr 25, 2021) per CoinGecko; fell to a low of ~$7.54 (Nov 26, 2022); ~$19 today. (Heavy rebasing acts like repeated stock-splits, so raw price comparisons are distorted — but the CoinGecko raw figures here are the cleanest reference.)
• Lesson: the cleanest example that this isn’t a peg problem at all — it’s a reflexive-Ponzi problem that afflicts any emissions-funded “money.” The 1 DAI floor protected backing, not market price; everyone who bought the premium lost.

Fantom fUSD — overcollateralized, yet depegged and stayed depegged

• The name: it’s fUSD — Fantom’s overcollateralized, DAI-style CDP minted against staked FTM, in the Andre Cronje–adjacent ecosystem. Not an algorithmic mint/burn coin.
• Failure (the instructive part): the redemption/liquidation path back to $1 didn’t function effectively — there was no reliable way to redeem fUSD for $1 of value — so when FTM collapsed (post-Terra contagion, 2022) the peg broke and had no force pulling it back. It fell to roughly $0.68–0.69 during the May 2022 depeg and traded below peg for an extended period after. Cronje proposed an fUSD “optimization” on May 20, 2022 specifically to address the entrenched depeg.
• Lesson: overcollateralization is necessary but not sufficient. If the arbitrage-closing mechanism is broken or undercapitalized, a coin can sit below peg indefinitely. And backing a stable with a single, highly-correlated ecosystem token means collateral and demand collapse together.

Iron Finance — IRON / TITAN: “DeFi’s first large-scale bank run” (June 16, 2021)

• Mechanism: partially-collateralized stable on Polygon, ~75% USDC + 25% TITAN (the reflexive share token), governed by dynamic target/effective collateral ratios. Redeeming IRON mints fresh TITAN.
• Failure: as TITAN fell, holders redeemed and dumped TITAN; the 10-minute TWAP oracle lagged the crashing spot price, so redeemers got TITAN valued at the stale high price → more minting → lower price → panic. Rational holders were incentivized to run even though “the system worked as designed.”
• Scale: TITAN → ~$0 (supply hyperinflated into the tens of thousands of billions); ~$2B value wiped (estimate). Mark Cuban was an investor, confirmed losses, and called for stablecoin regulation.
• Lesson: partial collateral + reflexive share token = the same spiral as pure-algo, just delayed by the USDC cushion. Oracle design is part of the attack surface. This was the dress rehearsal for Terra 11 months later — the Fed wrote it up alongside UST.

The seigniorage wave — Basis Cash, ESD, DSD (late 2020 – 2021)

• Mechanism: rebase/bond designs. Above $1, mint new coin to shareholders (seigniorage); below $1, sell discount “bonds/coupons” (burning supply) redeemable later if the peg recovers.
• Why they all died: the contraction mechanism is a pure confidence bet — coupon buyers only show up if they believe the peg will recover; once it stays broken, they vanish, supply can’t contract, peg stays dead. ESD issued ~$550M at peak → ~$0.23; DSD ~$300M → ~$0.24; Basis Cash never durably held peg (its co-founder was later revealed to be Do Kwon — a tidy bit of foreshadowing).
• Lesson: uncollateralized seigniorage needs perpetual net new demand; the stabilizer evaporates exactly when needed. NuBits (2014–2018) is the proof this pattern predates DeFi — same two-token design, broke in 2016, terminally failed March 2018.

Honorable mentions (one-liners that expand the taxonomy)

• Beanstalk (April 2022): ~$1B Aave flash loan → bought >67% of governance in one block → passed a malicious proposal draining ~$182M (attacker netted ~$80M). Governance is part of the attack surface; unaudited, no flash-loan/timelock protection.
• Wonderland / MIM / “Sifu” (Jan 2022): the “decentralized” treasury manager was revealed to be a QuadrigaCX co-founder. Key-man and founder-fraud risk in supposedly trustless systems; correlated-token contagion.
• Frax near-miss (March 2023): heavily USDC-collateralized, depegged to ~$0.877 when USDC did; subsequently moved to full collateralization. Even “careful” fractional-algo inherits its reserve’s weakest asset.

Synthesis — the five failure categories

1. Algorithmic reflexivity / death spiral — UST, IRON/TITAN, ESD/DSD/Basis, NuBits. The dominant killer.
2. Collateral risk — fUSD (correlated, broken redemption), Frax↔SVB (off-system shock).
3. Bank-run / coordination failure — IRON, the textbook case.
4. Governance / oracle exploit — Beanstalk (flash-loan governance), IRON (lagging TWAP), Wonderland (treasury control).
5. Yield-subsidy unsustainability — Anchor’s 20%, OHM’s rebase APY.

The through-line: durable pegs need (a) exogenous, uncorrelated collateral; (b) a redemption mechanism that closes the arbitrage under stress; (c) organic demand not propped by unsustainable yield. The algorithmic casualties were missing all three.

3. The Fork-Choice Problem

Who decides canonical Ethereum?

The actual source — and the correction

The canonical essay is Haseeb Qureshi & Leland Lee, “Ethereum is now unforkable, thanks to DeFi” (Oct 31, 2019). Two things to get right:

1. It’s co-authored with Leland Lee.
2. It was written ~3 years before the Merge, framed around the then-active, hypothetical ProgPoW mining dispute — not ETHPoW. The argument is structural, and the Merge later became its real-world confirmation. That is the sharper framing: a 2019 prediction about a hypothetical fork, validated by a 2022 fork nobody was thinking about when he wrote it.

The argument, precisely

Thesis: “Ethereum will never again have a meaningful minority fork, in large part because of DeFi’s inherent fragility.” The mechanism is composability + centralized chokepoints, with USDC as the keystone:

• CENTRE announces USDC will only be redeemable on one fork.
• “All DeFi operators are forced to now follow USDC’s lead. They cannot defy CENTRE… Composability both rules and constrains everything.”
• “DeFi operators would have no choice but to side with CENTRE and throw all of their weight behind the USDC-blessed fork, regardless of where community opinion came down.”
• Supporting cascade: oracles stop posting on the minority fork (“There are no more prices”), and Maker would “simply trigger global settlement” on it.

The issuer’s redemption choice becomes the Schelling point. Coordination incentives do the rest.

Vitalik conceded the same point (Aug 2022)

At BUIDL Asia (Seoul, ~Aug 4–5, 2022), weeks before the Merge: you’ll have “100 billion of USDT on one chain and 100 billion of USDT on the other chain… and so, they [Tether] need to stop respecting one of them.” He called it a 5–10-year concern and suggested diversifying stablecoins (USDC and DAI) as mitigation. (Reported quotes from a conference talk, not a written primary.)

The live experiment — the Merge / ETHPoW (Sept 2022)

• Circle (~Aug 8–9, 2022): “intends to fully and solely support” the PoS chain; USDC “can only exist as a single valid version.”
• Tether (~Aug 8–9, 2022): committed to support PoS Ethereum “in line with the official schedule”; Ardoino said stablecoins “should act responsibly and avoid disruption.”
• Outcome: ETHPoW (ETHW) went live ~Sept 15–16, 2022. With USDC/USDT unredeemable on it and oracle-dependent DeFi worthless there, ETHW collapsed from ~$50 to a low of ~$4.2 by Sept 19, 2022 (after a replay-attack scare), and later ~$3.13 during the Nov 2022 FTX collapse. The fork survived as a marginal chain and captured essentially none of the economic activity.

Why this is the deepest indictment

The sharpest framing: the canonicity of “real” Ethereum — the supposedly credibly-neutral, unstoppable settlement layer — was effectively ratified by the redemption policies of two private companies. This isn’t a peg risk or a counterparty risk; it’s a sovereignty risk at the base layer. It is the strongest argument for why decentralized stables aren’t a nice-to-have but a precondition for Ethereum’s neutrality. (Related: Lyn Alden, “Proof-of-Stake and Stablecoins: A Blockchain Centralization Dilemma.”)

4. The Failings of the Incumbents

USDC and Tether — the price of winning

USDC and the SVB depeg (March 2023) — banking risk, imported wholesale

• Circle confirmed $3.3B of USDC reserves stuck at Silicon Valley Bank (~8% of ~$40B reserves), disclosed March 11, 2023.
• USDC traded as low as $0.8774 (CoinGecko all-time low, March 11, 2023); sub-$0.87 wicks on individual venues were thin-order-book artifacts, not the aggregate low.
• Recovery came only after the Treasury/Fed/FDIC backstop on Sunday March 12 (6:15pm ET) guaranteed all depositors; full repeg once Circle resumed redemptions Monday March 13.
• Takeaway: the “1:1 dollar” was only as good as the solvency of the banks holding the cash. Absent a discretionary government bailout, the $3.3B haircut would have been real. A fiat-backed stable imports the entire counterparty risk of the banking system.

Tether — reserves, transparency, and the audit that never comes

• NYAG settlement ($18.5M, Feb 23, 2021): found Tether’s “backed 1:1” claim false; for periods it “held no reserves to back tethers in circulation at the rate of one dollar for every tether.”
• CFTC settlement ($41M, Oct 15, 2021): Tether held sufficient fiat reserves only 27.6% of the days in a 26-month sample (2016–2018).
• Now: reserves shifted from commercial-paper-heavy to Treasuries. FY2025 attestation (BDO, released Jan 30, 2026): net profit >$10B, total direct+indirect U.S. Treasury exposure >$141B, excess reserves $6.3B. USDT in circulation $189.5B (DefiLlama, May 2026; Tether’s own Q4’25 figure was >$186B).
• The persistent failing: Tether has never published a full financial audit — only point-in-time attestations. A ~$189B liability issuer running on attestations is the transparency gap to name plainly.

Censorship / freeze power — the core CROPS failure

• Both USDT and USDC contracts contain centralized blacklist/freeze functions; the issuer can immobilize any balance on-chain.
• Tether: has frozen >$4.4B cumulatively (>$2.1B tied to U.S. authorities); e.g., $344M across two Tron wallets in April 2026. (Notably declined to proactively freeze Tornado Cash in 2022, calling unilateral freezing “reckless.”)
• Circle / Tornado Cash (the canonical case): OFAC sanctioned Tornado Cash Aug 8, 2022, listing 44 addresses (38 ETH + 6 USDC). Within hours Circle blacklisted ~75 addresses, freezing ~$75,000 of USDC. (Three distinct numbers — 44 sanctioned, ~75 blacklisted, ~$75K frozen — that frequently get conflated.)
• The proof of the point: the Fifth Circuit ruled (Nov 2024) that immutable smart contracts aren’t sanctionable “property”; Treasury delisted Tornado Cash March 21, 2025. The code couldn’t be censored — but the centralized stables flowing through it were, instantly, on a sanction later found unlawful. Decentralized protocol > “decentralized” dollar.

Counterparty / regulatory capture — structural, not incidental

• Issued by regulated entities (Circle is NYSE: CRCL) that comply with sanctions/freeze requests as routine.
• Reserves sit inside the banking + Treasury system — the exact channel that produced the SVB depeg.
• Redemption is permissioned: direct 1:1 redemption only for KYC’d institutions. Retail are price-takers on secondary markets — which is why USDC could trade to $0.87 instead of being arbitraged to par instantly.

Scale + regulatory status (early–mid 2026)

• Total stablecoin market $321.9B; USDT $189.5B (58.9%), USDC $76.8B (23.9%), combined 82.7% — a near-duopoly (DefiLlama API, 2026-05-22). For scale, the next tier is far behind: USDS (Sky) $8.8B, USD1 (World Liberty) $4.8B, DAI $4.6B, Ethena USDe $4.4B.
• GENIUS Act (signed July 18, 2025): 100% reserves in cash/short-term Treasuries, monthly disclosures, licensed-issuer regime — and bans paying yield to holders. Effective the earlier of Jan 18, 2027 or 120 days after final rules; rulemaking in progress as of early 2026.
• MiCA (EU): stablecoin rules enforceable from mid-2024; Tether did not seek authorization, so regulated EU venues delisted/restricted USDT through early 2025, shifting EU volume toward USDC.

The irony

The most “successful” stablecoins are the most centralized — and regulation is entrenching that, not fixing it. GENIUS rewards the fiat/T-bill model and bans yield (protecting bank deposit franchises and Treasury demand), while effectively excluding algorithmic and disadvantaging decentralized designs that can’t post fiat reserves through a licensed depository. The state is blessing a duopoly that gives up every CROPS property — Censorship-Resistant, Open-source, Private, Secure — that motivated crypto in the first place. CROPS is the rubric used to score the flatcoins in §5.

5. The Flatcoin Frontier

Ungovernable money that actually works

Definition first — two senses of “flatcoin”

• Strict sense (purchasing power): pegged to a CPI basket / real goods, not to $1. Coined by Balaji Srinivasan (2021) (“one coin buys a hamburger today and in five years”), made a build priority by Coinbase (Feb 2023), named a top opportunity by Vitalik (late 2022). The clearest live example is Nuon (Laguna), which markets itself as a decentralized flatcoin pegged to a goods basket via Truflation.
• Loose sense (decentralized, non-fiat-pegged stable-value): RAI lives here — it floats around a redemption target set by a controller; it is not CPI-pegged. The two should not be conflated.

LUSD, RAI, HAI, and ZAI are often grouped together — but the unifying property isn’t the peg target, it’s the trust model: decentralized, crypto-collateralized, minimal-governance, censorship-resistant. That is the real category — call it “ungovernable money” for the loose-sense umbrella.

The two critiques the category must answer

1. Volatility of the underlying collateral. ETH is volatile, so you need some shock absorber: overcollateralization + liquidations (LUSD), or a floating redemption price that absorbs volatility into the peg target itself (RAI/HAI), or a soft-liquidation curve (crvUSD’s LLAMMA).
2. Capital efficiency / scalability. This is the unsolved one. Overcollateralized CDP supply is capped by demand for leverage — you only get stablecoins when someone wants to borrow against crypto. That’s why DAI reached for USDC: to scale past the leverage ceiling, it had to centralize collateral. The honest statement: no decentralized stable has yet scaled to USDC-size without centralizing collateral. The trilemma is alive.

The protocols (sizes from DefiLlama/CoinGecko, 2026-05-22)

LUSD — Liquity v1 (the immutability gold standard). ETH-only, 0% interest, 110% minimum collateral ratio, Stability Pool + redemption arbitrage giving a ~$1.00–1.10 band. No admin keys, no governance, no upgradeability — it cannot be changed. The strongest CROPS score in the field and the cleanest counterexample to “decentralized stables inevitably centralize.” Cost: capital inefficiency and a soft ceiling on size. ($28.6M mcap; $242M Liquity v1 TVL. Note its all-time low was $0.897 in Jan 2022, so the “$1.00–1.10 band” holds in normal conditions but isn’t a hard floor.)

RAI — Reflexer (the non-pegged ideal type). ETH-only, non-pegged: an on-chain PID controller sets a floating redemption price via a redemption rate (price above target → negative rate nudges it down; below → positive). “Ungovernable money,” governance minimized. Reflexer Labs has wound down; contracts are immutable and live but static and tiny ($1.7M mcap; $2.3M protocol TVL; redemption price now ~$3.06 — it never targeted $1). The point it proves: you can build stable-value money with no peg promise and no governance — Vitalik’s “ideal type.”

BOLD — Liquity v2. Adds user-set interest rates (redemptions hit the lowest-rate loans first, not the lowest-CR ones) and multi-collateral (WETH, wstETH, rETH in isolated branches); Earn yield now comes from real borrower interest, not token emissions. More flexible, governed (so a notch below v1 on immutability). ($33M BOLD circulating; $82M Liquity v2 TVL.)

HAI & Open Dollar (OD) — the RAI descendants. HAI (“Let’s Get HAI”) is a multi-collateral RAI fork on Optimism (Ameen Soleimani; mainnet Feb 2024) with a KITE governance token — so more governed than RAI. Alive but tiny (~$0.6M circulating). Open Dollar is a distinct project (not a HAI rebrand): GEB-framework, governance-minimized, LST-backed, ~$1.00-floating on Arbitrum, by Joseph Schiarizzi, whose twist is Non-Fungible Vaults (the CDP itself is a tradable NFT) — but it is effectively dormant now (~$6K TVL), so it reads as a design idea, not a live option.

ZAI on Zcash — the frontier exemplar. An oracle-free, ZEC-collateralized CDP “flatcoin”: no Chainlink — an AMM TWAP is the only price input — and it’s shielded-transaction native (depends on Zcash Shielded Assets / ZSAs). Posted to the Zcash forum Feb 2026 by pseudonymous dev lamb356 (repo lamb356/zai-sim); claims $0 bad debt in backtested replays of Black Thursday, FTX, and Luna.

Crucial caveat: ZAI is a simulation / research proposal — not deployed. No testnet, immutability not yet addressed. It is the most on-thesis design in the field — it pushes on the two hardest things at once (oracle dependence and privacy, the P in CROPS that almost nobody else even attempts) — but it belongs in any survey as a frontier direction, not a usable product.

Others worth a line each:

• crvUSD (Curve): LLAMMA soft-liquidation (collateral is gradually swapped around the liquidation zone rather than dumped) — genuinely novel mechanism; governed. ($239M.)
• GHO (Aave): overcollateralized, multi-collateral, governed by Aave. ($583M — the largest of the governed-decentralized set.)
• frxUSD (Frax): moved to RWA/custodial backing — not censorship-resistant; the “decentralized brand, centralized reality” example. (FRAX ~$198M; frxUSD is the RWA/custodial successor.)
• AMPL / SPOT (Ampleforth): AMPL targets a CPI-adjusted dollar via rebasing; SPOT is the non-rebasing low-volatility derivative — the most credible commodity-money / strict-flatcoin attempt, but both are now tiny/illiquid (AMPL ~$6M mcap; SPOT thinly traded).
• Nuon (Laguna): the clearest explicit strict-sense flatcoin (Truflation goods basket) — but effectively dormant (~$9K TVL); a concept, not a going concern.
• Ethena USDe: delta-neutral, CEX-hedged synthetic dollar ($4.44B circulating, down from a ~$14B peak — the unwind continued through early 2026). The foil: decentralized-flavored but CeFi-dependent — the contrast against LUSD that makes the “what counts as decentralized” point.

Comparison table

The honest close

Score them on CROPS and only LUSD and RAI clear the censorship-resistant + immutable bar cleanly today. BOLD/crvUSD/GHO are decentralized-but-governed (a real category, not a dismissal). ZAI is the frontier — the only design even trying to add Privacy to the stack — but it’s an idea, not a product. And the scaling question hangs over all of them: the immutable ones can’t scale to USDC-size without centralizing, and the day they try, they become MakerDAO. That tension — not any single protocol — is the real subject.

Worth sitting with: outside GHO ($583M) and crvUSD ($239M), the immutable/censorship-resistant designs worth championing — LUSD ($28.6M), RAI ($1.7M) — are rounding errors against USDT’s $189B. That scale gap is the argument.

6. Sources

Design space / history

• Vitalik, Two thought experiments to evaluate automated stablecoins (2022-05-25) — https://vitalik.eth.limo/general/2022/05/25/stable.html
• Vitalik, What in the Ethereum application ecosystem excites me (2022-12-05) — https://vitalik.eth.limo/general/2022/12/05/excited.html
• Haseeb Qureshi, Stablecoins: designing a price-stable cryptocurrency — https://haseebq.com/stablecoins-designing-a-price-stable-cryptocurrency/
• MakerDAO MIP29 (PSM) — https://mips.makerdao.com/mips/details/MIP29
• Deribit Insights, DAI is now 60% backed by centralized assets (2020-09-26) — https://insights.deribit.com/market-research/dai-is-now-60-backed-by-centralized-assets-what-does-that-mean/
• Messari, State of Maker Q2 2022 — https://messari.io/report/state-of-maker-q2-2022
• CoinDesk, MakerDAO votes to retain USDC after depeg (2023-03-23) — https://www.coindesk.com/business/2023/03/23/stablecoin-issuer-makerdao-votes-to-retain-usdc-as-primary-reserve-even-after-depeg
• Maker Dec-2017 whitepaper — https://makerdao.com/whitepaper/Dai-Whitepaper-Dec17-en.pdf
• BitUSD background — https://medium.com/the-ledger-by-spark/what-is-bitusd-everything-you-need-to-know-about-the-first-stablecoin-ever-created-72337c53fdfa
• CoinDesk, Basis shutdown (2018-12-13) — https://www.coindesk.com/markets/2018/12/13/basis-stablecoin-confirms-shutdown-blaming-regulatory-constraints

Failure modes

• LFG reserves drained — https://www.coindesk.com/business/2022/05/16/luna-foundation-guard-left-with-313-bitcoin-after-ust-crash
• Harvard, Anatomy of a Run: Terra-Luna — https://corpgov.law.harvard.edu/2023/05/22/anatomy-of-a-run-the-terra-luna-crash/
• Anchor yield — https://www.coindesk.com/markets/2022/01/28/anchor-protocol-reserves-slide-as-money-markets-founder-talks-down-concerns
• Olympus bonding primer — https://olympusdao.medium.com/a-primer-on-oly-bonds-9763f125c124 ; CoinGecko OHM — https://www.coingecko.com/en/coins/olympus
• fUSD relaunch / depeg — https://cointelegraph.com/news/fusd-stablecoin-launch-and-rumors-of-cronje-s-return-send-fantom-ftm-price-higher ; https://decrypt.co/120530/ftm-soars-week-fantom-preps-stablecoin-relaunch
• Iron Finance post-mortem — https://ironfinance.medium.com/iron-finance-post-mortem-17-june-2021-6a4e9ccf23f5 ; Fed note — https://www.federalreserve.gov/econres/notes/feds-notes/runs-on-algorithmic-stablecoins-evidence-from-iron-titan-and-steel-20220602.html
• Seigniorage algos (ESD/DSD figures) — https://aws.okx.com/learn/algorithmic-stablecoins-performance-suggests-they-have-yet-to-justify-their-models ; https://dailydefi.org/articles/what-are-seigniorage-stablecoins/
• Beanstalk — https://www.coindesk.com/tech/2022/04/17/attacker-drains-182m-from-beanstalk-stablecoin-protocol ; https://medium.com/immunefi/hack-analysis-beanstalk-governance-attack-april-2022-f42788fc821e
• Wonderland/Sifu — https://www.coindesk.com/tech/2022/01/27/how-did-a-former-quadriga-exec-end-up-running-a-defi-protocol
• Frax/SVB — https://www.coingecko.com/research/publications/stablecoins-supply-svb-impact
• NuBits — https://medium.com/reserve-currency/the-end-of-a-stablecoin-the-case-of-nubits-dd1f0fb427a9

Fork-choice problem

• Qureshi & Lee, Ethereum is now unforkable, thanks to DeFi (2019-10-31) — https://haseebq.com/ethereum-is-now-unforkable-thanks-to-defi/
• Circle, USDC and Ethereum’s Upcoming Merge (Aug 2022) — https://www.circle.com/blog/usdc-and-ethereums-upcoming-merge
• Fortune, Circle and Tether support PoS Ethereum (2022-08-09) — https://fortune.com/crypto/2022/08/09/stablecoins-circle-tether-support-ethereum-merge-proof-of-stake/
• Vitalik @ BUIDL Asia (reported) — https://cointelegraph.com/news/vitalik-centralized-usdc-could-decide-the-future-of-contentious-eth-hard-forks
• Lyn Alden, Proof-of-Stake and Stablecoins — https://www.lynalden.com/proof-of-stake/

Incumbent failings

• USDC SVB — https://www.coindesk.com/business/2023/03/11/circle-confirms-33b-of-usdcs-cash-reserves-stuck-at-failed-silicon-valley-bank ; Fed FEDS note — https://www.federalreserve.gov/econres/notes/feds-notes/in-the-shadow-of-bank-run-lessons-from-the-silicon-valley-bank-failure-and-its-impact-on-stablecoins-20251217.html
• NYAG settlement — https://ag.ny.gov/press-release/2021/attorney-general-james-ends-virtual-currency-trading-platform-bitfinexs-illegal
• CFTC settlement — https://www.cftc.gov/PressRoom/PressReleases/8450-21
• Tether Q4’25 attestation — https://tether.io/news/tether-delivers-10b-profits-in-2025-6-3b-in-excess-reserves-and-record-141-billion-exposure-in-u-s-treasury-holdings/
• Tornado Cash sanction — https://home.treasury.gov/news/press-releases/jy0916 ; delisting — https://home.treasury.gov/news/press-releases/sb0057
• Circle freeze (Dune-cited) — https://www.theblock.co/post/162172/circle-freezes-usdc-funds-in-tornado-cashs-us-treasury-sanctioned-wallets
• GENIUS Act — https://www.whitehouse.gov/fact-sheets/2025/07/fact-sheet-president-donald-j-trump-signs-genius-act-into-law/ ; https://www.cov.com/news-and-insights/insights/2025/07/the-genius-act-becomes-law-key-provisions-from-the-federal-stablecoin-regulatory-framework
• DefiLlama stablecoins (live data) — https://defillama.com/stablecoins

Flatcoins / decentralized stables

• Liquity docs (LUSD/BOLD), Reflexer/GEB docs (RAI), HAI / Open Dollar docs, Ampleforth (AMPL/SPOT), Nuon (Laguna), Ethena — primary protocol docs; current sizes from DefiLlama/CoinGecko.
• ZAI on Zcash — Zcash forum thread (Feb 2026) + repo lamb356/zai-sim.

7. Data Notes & Caveats

• Live data was pulled from the DefiLlama and CoinGecko APIs on 2026-05-22 and cross-checked between them. Market figures move daily.
• fUSD’s exact depeg trough is approximate: the best-documented figure is ~$0.68–0.69 (May 2022); fUSD is now delisted from CoinGecko, so a clean long-run chart is hard to source. The structural point (it depegged and stayed depegged) is solid regardless.
• Peak % of DAI backed by USDC: the best-supported dated figure is ~60% USDC-collateralized (Aug 2022, Messari); a Sept-2020 Deribit snapshot put centralized-stablecoin backing at ~52%. The precise all-time peak would want a Makerburn/daistats snapshot.
• Terra (~$40–50B) and Iron Finance (~$2B) figures are notional value wiped, not user clawback.
• OHM raw prices are distorted by heavy rebasing (repeated stock-split-like dilution); the CoinGecko ATH ($1,415, Apr 2021) and low ($7.54, Nov 2022) are the cleanest reference points.
• Vitalik’s BUIDL Asia remarks are reported from a conference talk, not an authored text.
• ZAI (Zcash) is a simulation/research proposal, not a deployed protocol.