You Did Not Pick A Rate. You Picked Who Decides.

Two vaults sit side by side on a comparison sheet. Both hold dollars, both are run by firms you have heard of, and one pays a quarter of a point more. That quarter of a point gets a meeting. Somebody pulls the history, somebody asks about fees, the money goes in, and the sheet gets filed.

The sheet had no column for what each vault is allowed to lend your dollars against, or who can change that list. Over the last year, sixty one of the eighty eight stablecoin vaults on Morpho holding a million dollars or more changed it.

On Morpho, your dollars do not sit in one big pool. They get lent out in separate markets, each one a pair: dollars on one side, and on the other, one specific kind of collateral that borrowers post to take them. Bitcoin in one market. Staked ether in another. A slice of somebody's fixed income product in a third.

A vault sits on top of those markets. You hand it your dollars, and a professional firm decides which markets they go into and how much goes where. Morpho calls that firm the curator. The best known names are serious operators. Steakhouse describes what it does as "Risk-curated Morpho vaults for treasury and fund capital," which is the exact reader of this post. Gauntlet talks about "Quantitative, model-driven frameworks with human oversight." Neither of those is marketing fluff. This is real work, done by people who are better at it than most treasury teams would be.

The list of markets your dollars can go into is not fixed. The curator can add to it. A new kind of collateral can show up next month, and your dollars can be lent against it, and you agreed to that on the day you deposited.

Morpho knows that is a lot to hand over, and it designed a protection for it. Its own docs put it in one line: "nearly all of a Curator's significant actions are subject to a timelock, providing transparency and giving depositors a window to exit if they disagree with a proposed change."

In plain words: when a curator wants to let a new kind of collateral into your vault, they have to announce it first and then wait. During that wait, anyone can read the change, and if you don't like it, you can take your money out before it happens. The length of the wait is a setting on each vault, and Morpho's own code caps it at two weeks.

That is good design. What nobody has checked is whether it gets used: how often the list actually changes, and whether anybody leaves when it does.

Nobody had published an answer, so we went and read the whole year.

We looked at every stablecoin vault in the original generation of Morpho vaults, the ones Morpho calls MetaMorpho, on Ethereum, from mid September last year to this week. Two hundred and seventy five of them hold dollars of some kind. Most are tiny, somebody's test or a vault that never took off, so we only counted a vault from the first day it held at least a million dollars. That left eighty eight, holding $821 million between them at the end.

Over the year, sixty one of those eighty eight let in at least one new kind of collateral. The typical vault let in two. The busiest tenth let in fourteen or more. Across all eighty eight, that is four hundred and eighty one new kinds of collateral admitted, not counting the times an existing one was simply given different terms.

The ones that came in most often were not exotic. Coinbase's wrapped bitcoin came in twenty times, ordinary wrapped bitcoin thirteen, a restaked ether token twelve. But a lot of what came in were short lived fixed income tokens with an expiry date in their name, the kind that is created to last a few months and then ends.

In the typical vault, the money did not follow. When we looked at where each vault's dollars sat at the end of the year, the typical vault had none of them in a market that was not on its list when we started counting. The curators added permission, and mostly kept lending against what they already had. So for most depositors nothing has happened yet, except that the list of things that could happen got longer, and the permission is still standing.

The warning only exists if the vault has one. Morpho's first version of these vaults required at least a day. The second release lets a vault be created with no waiting period at all, and only asks for a day if somebody later decides to add one.

Eleven vaults in our count were created that way and never changed it. On 18 September, the last day we counted, they held $166.8 million between them, and $163.1 million of that was in one vault. In those eleven, a new kind of collateral goes in the moment the curator decides it should. Fifty eight new kinds went in that way over the year. There is no window in which to disagree.

That leaves the seventy seven vaults that do give notice. For every time one of them announced a new kind of collateral, we took everyone holding ten thousand dollars or more at the moment of the announcement, and measured how much of their money had left by the time the change took effect. Then we measured the same thing over a stretch of the same length just before the announcement, when nothing was pending.

Two hundred and twenty one warnings, some of them covering more than one change at once. A typical wait of just under five days. Almost sixteen thousand depositor positions.

During the warning, 18.9% of that money left. In the quiet stretch before it, 18.6% left.

People move money in and out of these vaults all the time, for their own reasons, and that baseline churn is the whole of what happened during the warning. The announcement barely moved anyone. In thirty four of the two hundred and twenty one, not a single dollar of that group left while the change was pending.

Splitting it by who was holding does not change the answer. Money held through other products, rather than in a person's own wallet, left less during the warning than before it, 11.8% against 14.9%. Money held by ordinary wallets left a little more, 29.3% against 23.9%, which is the most anybody reacted to anything, and it still describes a group where most of the money stayed put through every single change.

Paying closer attention would not have changed much. By the time a warning arrives, the curator has already decided, and the only thing a depositor can do with it is leave.

Go back to the four million dollars. Somebody on the team probably spent a week on which vault to use, and a good part of that week was the rate. If one vault paid a quarter of a point more than the other, that is ten thousand dollars a year, and it was worth arguing about.

The collateral list got no week. It was not on the comparison sheet, because it is not a number. And yet it is what decides whether the four million comes back. The quarter of a point is the upside you argued about, and what your dollars are lent against is the downside you handed to the curator.

For a lot of treasuries that is a perfectly good trade. The people running these vaults are good at it. It should still be a choice made knowingly, and one you can take back without standing guard over a warning you will not act on.

We already wrote about what happens after you choose, where every venue pays about the same today and the difference is in what happens later. And about getting out once a bad market is already draining. This one is about the day the market was let in.

Plug does programmable Ethereum transactions without smart contracts. For this problem, that means the finance lead decides what the company will and will not lend against, and that decision lives with the company, not with the vault.

Then the moving is handed off. You name the vault you are in, the market you would rather lend to yourself, and a market you never approved. From the company's own account, the rule checks whether the vault has put money into the one you never approved. If it has, it takes your whole position out of the vault and lends it straight into yours. If it has not, it does nothing, and you keep the curator's work and the curator's rate.

Nothing new gets deployed to make that happen, and the rule is not a promise from anybody else. It runs off your own list, from your own account, and it works the same in a vault with two weeks of notice as in one of the eleven with none.

Across two hundred and twenty one warnings, 18.9% of the money left while a change was pending and 18.6% left when nothing was. The notice period is there, and depositors holding ten thousand dollars or more behaved as if it was not. A list the company keeps itself does not depend on anybody reading the warning.