Push a market toward its open-interest cap and watch the maximum leverage on new size descend, so a position opened near the cap is over-collateralised enough to cover itself. Three views: the descent, a JELLY-style stress test, and how other venues handle the same risk.
Max leverage vs. forward-looking open interest
open interestcap
90
95
100
Forward OI
35%
of cap
Max leverage
10.0×
static 10×
Collateral for order
$0
the static minimum
Market
Current open interest 30% of cap
Your order size 5% of cap
Guided scenarios
Illustrative numbers. The mechanism, the 90 / 95 / 100% thresholds and the 0.01× floor mirror the on-chain contract; the market caps shown are round example values, not any live market's real limit.
In March 2025 a trader opened an oversized $JELLY short on another venue, then sent the price. The trigger for the blow-up was mundane: once the market's OI cap was hit, orders were rejected, including the vault's own close orders, so it couldn't de-risk, and the loss cascaded into the backstop. Both sides below hold the same oversized short and try to close as the price craters. A hard OI cap rejects the closes and traps the vault; Perpl's descent never gates closing, so it exits at will, and even a position held to liquidation is over-collateralised. Set the crash, then run the attack.
Price crash on liquidation 50%
price $1.00
Without · hard OI cap
Trapped, then liquidated
Close ShortSizeStatus
JELLY466,730pending
JELLY452,780pending
JELLY447,750pending
JELLY184,610pending
The oversized short tries to close before the crash.
the short's loss on liquidation—
what the losing short owes the traders who won. Who funds it? ↓
1 · the trader's own collateral$0.9M
just 5% at 20×, the only loss the trader takes
2 · backstop vault pays—
the shared insurance fund, drained to cover the gap
3 · bad debt: protocol + LPs eat—
unfunded loss left after the vault is empty, socialised onto everyone else
The oversized short tries to close. Run the attack.
With Perpl's descent
Closes freely, no trap
Close ShortSizeStatus
JELLY466,730pending
JELLY452,780pending
JELLY447,750pending
JELLY184,610pending
The oversized short tries to close before the crash.
the short's loss if forced to liquidate—
what the losing short owes the traders who won. Who funds it? ↓
1 · the trader's own collateral (descent-forced)$18.0M
≈ the whole position, the trader covers their own loss in full
2 · backstop vault pays—
the shared insurance fund is never touched
3 · bad debt: protocol + LPs eat—
none, no loss escapes the trader's own collateral
The oversized short tries to close. Run the attack.
Illustrative reference market. The JELLY figures are from public post-mortems and stand in for the class of risk the mechanism addresses.
CEX
Binance / Bybit
Centralised, off-chain
How size is limited
Per-user leverage tiers; leverage drops as position size grows
Keyed on
The account (KYC identity)
Enforced
Off-chain, trust the operator
APPCHAIN
Hyperliquid
Dedicated app-chain
How size is limited
Size-based margin tiers + dynamic, market-cap-scaled OI caps (post-JELLY)
Keyed on
Position size
Enforced
On its own chain, not composable EVM
DEX
dYdX
Order-book DEX
How size is limited
Open-interest-scaled initial margin; margin rises with market OI
Keyed on
Market open interest
Enforced
App-chain / off-chain matching
PERPL
Perpl
Fully on-chain, Monad EVM
How size is limited
Leverage descends toward 0.01× as forward OI approaches the cap
Keyed on
Market open interest; identity-blind, can't be split across wallets
Enforced
In the smart contract, every order
Same idea, different venue. Size-based margin tiers (Hyperliquid) and open-interest-scaled initial margin (dYdX) are the established answers to the same problem: an oversized position is a solvency risk, so make it post more margin. Perpl's leverage descent is that idea expressed on a permissionless EVM, keyed on aggregate market OI rather than per-account limits, so it can't be gamed by splitting a position across wallets, and enforced in the contract on every order rather than by a trusted operator.