Freeboard
Your loan has a price nobody has shown you.
Pledge NVDA on Robinhood Chain, borrow dollars against it, and there is an exact price — on chain, today — at which any stranger in the world may repay your loan and walk off with 12.68 cents of your collateral for every dollar they put in. No email, no margin call, no limit on how much of it they take. Freeboard measures how far away that price is, and moves it further away in one transaction, using none of your own money.
The deck is where the water is allowed to come to. Fall 36.0% and it arrives: $4,507 of NVDA leaves to repay $4,000, and $507 of that is the fee for not having done it yourself.
One line of arithmetic, and nothing tells it to you
Morpho Blue is a good lending protocol precisely because it is small: it has no keeper, no notification, no grace period and no dashboard. What it does have is one inequality, and the price at which that inequality flips.
Where the line is
A position is healthy while collateral × price × lltv ≥ debt. Rearranged, the price at which it stops being healthy is debt / (collateral × lltv) — and for every stock market on this chain lltv is 62.5%, so you lose the loan while your collateral is still worth substantially more than you owe.
- Read from Morpho Blue itself, not from an indexer
- Priced by the market's own oracle, which is what Morpho uses
- 30 stock markets, 11 of them with debt
There is no close factor
Aave caps a liquidator at half your position. Morpho Blue has no such cap: the first person to act may repay all of your debt and seize the collateral that goes with it. So the penalty behind a loan is not a slice of it — it is 12.68% of the whole thing, every second the position is underwater.
- Proved by doing it: the suite closes a whole position in one call
- $507 on the $4,000 example above
And it is not a wipeout
This is the part nobody says. At the liquidation price your collateral is worth debt / lltv, and a liquidator takes debt × 1.1268 — so on the example, $1,893 is still handed back. A liquidation is not ruin. It is a 12.68% fee for a sale you did not choose, at a moment you did not choose.
- Which is exactly why it is worth paying a fraction of that to avoid
Room costs money. Some of it costs less.
Every route to more freeboard is one of three things, and Freeboard prices all three against each other and against doing nothing.
Repay
Dollars off the loan. Costs you nothing but the dollars, and it is the cheapest route there is — so it is offered first whenever the wallet holds any.
- Two calls, no flash loan
- Permissionless: needs no signature from you at all
- Your holding does not change
Ballast
Pledge more of the same stock you already hold. Costs nothing, and keeps your exposure exactly where it was — which is usually the whole reason the position exists.
- Two calls, no swap, no slippage
- Also permissionless, so it cannot leave a permission standing
- Needs stock in the wallet
Trim
Sell part of what is already pledged. This is the one that needs none of your own money: a flash loan repays the debt, the freed collateral is sold to pay the flash loan back, and what is left stays pledged.
- One transaction, through Morpho's own Bundler3
- Costs the trading cost — measured, not estimated
- Reduces your exposure, which is the point and the price
One transaction, and none of your money in it
Six calls inside one call to Morpho's own Bundler3. Your wallet signs once.
A signature, for a minute
Taking collateral out acts for you, so Morpho wants an authorisation. The bundle carries your signed grant and your signed revocation over consecutive nonces, so it is gone again before the transaction ends.
The dollars are borrowed
Morpho flash-lends exactly the repayment. Not a cent of it is yours and not a cent of it survives the transaction.
Part of the loan is repaid
A fixed number of dollars, chosen by the plan — not "all of it", so nothing here depends on what the debt happens to be at the second it is mined.
Only what is sold comes out
The collateral that is not being sold never moves. The rest is withdrawn to the adapter and sent to Uniswap's Universal Router.
Sold, with one floor
The router's last command sweeps the dollars out and reverts if they are less than the flash loan. A trim that cannot raise what it borrowed does not half-happen.
The slippage that did not happen
Surplus dollars go back onto the loan rather than into your wallet: you asked for freeboard, so they buy more of it. Nothing is left in the adapter, the bundler or the router.
A fee is a fact. A liquidation is a possibility.
So they are never added together. Freeboard shows what a trim costs, and separately what a liquidation would cost, and lets you decide which risk you are buying out of.
Measured against the pool, not the oracle
A trim's cost is the gap between what your collateral is worth at the market's own oracle and what a Uniswap pool will actually pay for it — the fee plus the price your size moves, together. It is not measured against the two venues' disagreement: a sibling site that did that produced a cost which came out negative, and it was true in one venue's eyes and false in every other sense.
- Pool state read properly: bitmap words and every initialised tick in the price window
- A sale that would walk past what was read comes back refused, not quoted
- 317 live pools across the chain's stock tokens
And the refusal that matters
Selling collateral to repay debt only helps while 1 − cost > ltv. Past that line every dollar raised gives up more equity than it retires and the sale pushes you towards liquidation, not away from it. That is not a rounding effect — it is where the arithmetic changes sign.
So Freeboard will not build one — and on this chain it never has to. The guard needs a trading cost above 1 − ltv, which on the position the fork suite built is 71.9%. Selling everything pledged into the real pools costs 0.01%. The region is real arithmetic and the pools do not reach it, so the refusal is insurance and not a feature. It was the control that proved this: it forced a trim through and the trim helped.
What we found that does not flatter this
The biggest borrower on this chain cannot use the main route
wsNET carries $715,647 of stock-backed debt — about 42% of all of it — and has no Uniswap pool at all. There is nowhere to sell its collateral, so a trim is impossible on it and the scanner marks it suspect for a second reason: with no pool, there is no price from outside Morpho to check its oracle against. 4 markets with debt are in that position. Repay and ballast still work; the flash-loaned route does not, and the app says so rather than quoting it.
Nobody here is about to be liquidated
11 loans on this chain sit within a 25% fall of being closed — and they hold $474 between them. The 10 loans that hold 99.8% of the debt have between 27.2% and 71.7% of room. The thin positions here are dust; the real ones are comfortable today.
So the claim is not that anyone is in trouble this afternoon. It is that the line exists, nothing shows it to you, and 27.2% is an ordinary quarter in a large-cap stock. $122,937 of penalty is contingent on a move the market makes regularly, and the people it is contingent on cannot see the price it happens at.
And a short would have been the better story
The first idea here was to let you bet against a tokenized stock: borrow the stock, sell it, buy it back cheaper. Morpho markets that lend NVDA, TSLA, SPY, AAPL and GOOGL all exist on this chain — and every one of them holds zero supply. Nobody lends a share, so nobody can borrow one, so the product was dead before a line of it was written. Ten minutes of measurement, and the honest version of the answer is that it cannot be built here at all.
Every claim above has something behind it
And where there is nothing behind it, this page says so in those words.
Run against a fork of this chain
- 27/27 — the whole product on a fresh fork (tools/test-trim.mjs): a position opened, measured, trimmed, and the collateral exact to the wei against the local simulation
- 22/22 — properties executed on chain, one per eth_call
- 32 seeds, 40 random operations — invariants after every call, and 8/8 deliberately broken builds each caught by the invariant named for it
- 16/16 — sabotages, each caught by a named property or declared a survivor with a reason
- 14/14 — every address read back off the chain, with the exact selectors this site calls
- The rate model agrees with the deployed Adaptive Curve IRM on all 38 markets, exactly
And what is not
- Nothing here is audited. Freeboard deploys no contract, so what you are trusting is Morpho Blue, Morpho's Bundler3 and Uniswap's Universal Router — none of which we wrote — plus this site's arithmetic, which is measured and not audited.
- Self-liquidation is not built. Once a position is already liquidatable, the cheapest move is to be your own liquidator and keep the penalty. It is real, it is possible with the same flash loan, and it is not here.
- Nothing watches your position. There is no keeper and no alert. You come back and look, or you do not.
- Oracle staleness is not modelled. Some of these oracles stop answering after a day, which changes what is possible over a long weekend.