Consider the moment when a protocol that once minted billions of dollars in synthetic currency was shut down by two wallets.
I opened the Abracadabra governance page on a Tuesday evening in Tallinn, coffee going cold, and stared at the participation count on the proposal to wind down MIM. Two addresses. That was the entire electorate. One of them — the proposer — held roughly 99.5% of the votes cast. The poll opened September 29 and would close Wednesday at 13:24 ET, and barring something genuinely extraordinary, the outcome had been decided before the first signature was collected.
That is the detail that will never make the press release. Not the $22 million in circulating MIM. Not the roughly 96% loss facing holders. Not the legal opinion that quietly zeroed out the governance token. The story is the two wallets, and what they reveal about who actually controls a DAO when the collateral runs dry.
Abracadabra was, for a stretch, one of the more genuinely interesting experiments in the collateralized-debt-position design space. The model is worth restating plainly. Users lock collateral into isolated lending markets the protocol calls cauldrons, and against that collateral they mint MIM — Magic Internet Money — a stablecoin targeting one dollar. Same family tree as DAI. Different branches, similar roots.
I have been reading whitepapers and liquidation logic since the 2017 ICO mania, when I screened fifty projects in a single quarter and found twelve with economic models I would defend in writing. Abracadabra was not on that list, but it belonged to the same intellectual lineage: overcollateralized, on-chain, composable, auditable. At its peak, MIM's float was measured in the billions, and for a lot of retail users it was the first stablecoin they ever minted rather than bought.
Then the attacks arrived, and they never really stopped. January 2024 brought roughly $6.5 million in losses. March 2025 took about $13 million through a GMX-linked cauldron. Later that year, an abandoned cauldron was used to mint 1.79 million MIM out of nothing. Three incidents across three years, and one uncomfortable conclusion: the protocol patched the symptom of the week and never touched the root cause.
By the time the wind-down proposal landed, the arithmetic had stopped being ambiguous. Roughly $22 million of MIM circulating outside the protocol. About $21 million in bad debt. Effective collateralization below four cents on the dollar. More than 95% of outstanding supply unbacked by anything at all.
The team's own language was unusually blunt: MIM is severely undercollateralized, with no viable path to restore parity with the dollar. For the Estonian builders I ran workshops with back in 2020, this is precisely the scenario we drilled — not how to farm yield, but how to recognize the moment a protocol's promise quietly expires.
Here is where technical detail stops being trivia and becomes the entire story.
The wind-down mechanism is simple in outline. Collateral is recalled from the cauldrons, converted to ETH, and distributed through Merkl, the distribution infrastructure built by Angle Protocol. Borrowers hold first claim: they receive their collateral value minus their MIM debt, computed at one dollar per MIM. MIM holders receive whatever remains. That remainder is estimated at roughly four cents per token.
Read that waterfall twice, because the ordering is the whole economic argument. MIM holders are residual claimants in a liquidation where the senior tranche is already deeply underwater. Four cents is not a floor. It is an optimistic midpoint calculated before slippage, before distribution costs, and before the weeks of execution risk sitting between the snapshot and the payout.
Underneath all of it sits a mismatch the CDP model never solved. MIM is a liability that must settle at one dollar. The collateral backing it is an asset whose value moves with the market and, in this case, with the success of attackers. Rigid liability, elastic asset. When the elasticity runs in one direction for long enough, the structure is insolvent by construction, and no amount of parameter tuning repairs a balance sheet whose assets are worth four cents against a dollar of claims.
There is also a structural trap buried inside the cauldron architecture itself. One Arbitrum WETH cauldron is immutable — its interest rate parameters cannot be changed. Roughly $300,000 sits inside it. Because nobody can adjust the economics to force borrowers to close, and because the team has already stated it will not maintain the protocol, that capital is effectively stranded. The team's accessible collateral pool comes to about $900,000.
Sit with that, because it is the most instructive detail in the whole case. The immutability was a feature. It was marketed as censorship resistance, as a guarantee that no admin key could ever rewrite your terms mid-loan. In the terminal moment, that same immutability converted $300,000 of user collateral into a sunk cost nobody can retrieve. Code binds, but people break or build — and when the builders walk away, the code keeps binding the people who stayed.
Then there is the external dependency. LayerZero's V1 relayer is being retired, and funds held in the Stargate USDC/USDT cauldron — roughly $1 million — must be withdrawn before December 15. That deadline has nothing to do with Abracadabra. It is a third party's infrastructure migration landing squarely inside somebody else's liquidation. This is the composability tax that almost nobody prices in advance: your protocol's terminal value depends on other protocols' roadmaps, and those roadmaps do not wait for your wind-down to finish.
Now the part that matters most for anyone still holding a token. The team's legal counsel issued an opinion stating that MIM is a liability senior to the SPELL governance token, and that until that liability is fully discharged, SPELL carries no accounting value. That is not a market judgment or a sentiment reading. It is a formal legal characterization, and it means the equity-like layer of the protocol has been marked to zero by the protocol's own advisors. Governance tokens are supposed to be the residual claim on upside. Here they are the residual claim on nothing.
There is also a six-month claim window. Unclaimed funds after that period are redistributed to those who did claim, up to one dollar per MIM, with any excess flowing to borrowers. On paper that is a fairness mechanism. In practice it is a penalty on passive holders — the retail wallets that bought MIM two years ago, forgot about it, and will never see the announcement. Trust is the only currency that matters, and this design spends it on the assumption that everyone is watching.
There is a final layer of fragility the proposal does not advertise. The distribution runs through Merkl, which means the liquidation depends on a centralized operating entity voluntarily executing it, on schedule, with no legal obligation to finish. The team has already disclaimed any legal or technical responsibility for maintaining the protocol. If Anubis walks away mid-process, holders get the worst possible outcome: a liquidation that is approved, on-chain, and permanently unclaimed.
The interface, for what it is worth, stays online. No active maintenance, but the front end does not go dark. I find that worse than a shutdown. A dead page is honest. A live page with nobody behind it is a suggestion of continuity that no longer exists.
The governance numbers deserve their own paragraph. Only two wallets participated in the vote. The proposer controlled roughly 99.5% of the votes cast. And this was not an anomaly — the June proposal to hand operational control to Anubis also drew exactly two wallets. Two consecutive existential decisions, four wallets' worth of participation if you count generously, and zero meaningful community deliberation. Whatever that process was, it was not governance in any sense the word usually carries.
The industry will file this under responsible wind-down. Orderly liquidation. Transparent process. No sudden rug, no exit scam, just a protocol choosing to die with its paperwork in order. That framing is comfortable, and it is wrong.
An orderly wind-down run by a single entity, approved by two wallets, executed through a third party's distribution tool, with the operating team having already disclaimed any legal or technical responsibility for maintaining the protocol — that is not decentralization maturing. That is the compliance shield doing precisely what it was designed to do.
Look at the sequence. In June, operational control was transferred to an entity called Anubis. That handover proposal also drew exactly two voting wallets. Four months later, the same governance apparatus approved the shutdown. "Code is law" was never the operating principle here. The upgrade rights and the multisig always sat with a handful of people, and when the moment came, those people used them to close the door and step away from the wreckage. The DAO wrapper supplied legitimacy to a decision the community never actually made.
And the arbitrage story — MIM trading near $0.029 against a four-cent recovery estimate, a headline spread of roughly 38% — is a mirage dressed up as opportunity. You are pricing execution risk, legal risk, and a six-month claim deadline against a payout that depends entirely on a team that has publicly stopped maintaining the thing. That is not arbitrage. That is buying a lottery ticket and calling it a bond.
MIM will be filed next to TerraUSD in the industry's memory, but the lessons are not the same. UST died of a flaw in its peg mechanism. Abracadabra died of governance that was never real and code that was never fixable. Culture eats blockchain for breakfast — and the culture here was two wallets and a press release.
The next protocol facing this choice should ask one question before the crisis, not during it: when the collateral runs out, who actually holds the keys, and will they still be in the room? We are building the future, together. We should build it so the last person out does not get to decide alone.
