One asset.
One box.
Zero spillover.
A box is a lending market with exactly one thing in it. Lenders put dollars in and hold shares of that box and no other. Borrowers post one tokenized stock and draw dollars against it. The box has a price feed, a rate curve, a loan-to-value and a liquidation bonus. All six are fixed the moment it is stacked, and there is nobody who can change them afterwards.
Pooled lending shares risk the way a warehouse shares a fire. Stackyard stacks crates instead. If a name gaps through its collateral on a Monday open, the loss is written off inside that crate, in that transaction, against the lenders who chose it. The crate beside it does not know anything happened.
Anyone can stack a box: any tokenized stock on the chain, any oracle module, any rate module, any loan-to-value under the cap. A different opinion about NVDA is a different box with its own lenders, not an argument in a governance forum. The yard keeps the list and touches nothing.
The boxes
Price is the Chainlink print for one token, which already carries the issuer’s dividend multiplier. Print is how old that print is; equity feeds hold through the weekend without updating, and a box accepts a print up to four days old. Supplied, Borrowed and the rates read from the box itself and stay blank until it exists.
Gap one crate. Watch the other.
One lender, one hundred thousand dollars in each of two crates. A borrower posts 10,000 GME at $25 and draws every dollar in that crate. Drag the price. Below $22.22 the position is liquidatable; below $11 the collateral no longer covers the debt plus the bonus, and the rest is written off against the GME lenders, right then. Look at the NVDA crate while you do it.
Stack your own box
Pick a tokenized stock, an oracle, a rate curve, a loan-to-value and a bonus. The yard checks that the collateral is really a stock (it answers the issuer’s interface, which eight fake AAPLs on this chain cannot), computes the box’s address from its parameters, and stacks it. One spec, one box, one address, forever.