What happened with Hyperliquid's JELLY incident — and what it means for DEX governance

• -$13.5M - HLP peak unrealized loss during the JELLY incident
• ~20% - HYPE price drop during the episode
• 560% - JELLY price spike after Binance listed the token on the same day
• $0.0095 - Forced settlement price chosen by validators, vs ~$0.50 spot price at the time

Contents

1. The short version
2. How the attack actually worked
3. How HLP absorbed the damage
4. The governance response
5. The "code is law" debate
6. What changed at the protocol level
7. What it means for active traders
8. FAQs

In late March 2026, a coordinated manipulation attempt on Hyperliquid turned a low-liquidity meme token into a live stress test for on-chain governance. The JELLY incident did not break Hyperliquid. But it exposed fault lines that every serious perp trader needs to understand - about how DEX governance actually works, who holds power when things go wrong, and what "decentralized" means in practice when a protocol is under attack.

This is not a hit piece on Hyperliquid. It is an honest account of what happened, what it revealed, and what it means for how you think about risk on any perp DEX.

The short version

A trader - or coordinated group of traders - opened a large short position in JELLY perpetuals on Hyperliquid, then deliberately engineered a liquidation that forced the protocol's liquidity provider vault (HLP) to absorb it. HLP's unrealized loss peaked at negative $13.5 million. HYPE dropped roughly 20%. Binance listed JELLY the same day, triggering a 560% price spike that made the situation significantly worse.

Hyperliquid's validators responded by voting off-chain to delist JELLY and forcibly settle all remaining positions at $0.0095 - while the token was trading around $0.50 on spot DEXs. That settlement price was not a market price. It was chosen to protect HLP solvency.

The decision worked. It also raised serious questions about what decentralized governance actually means on Hyperliquid.

How the attack actually worked

Three accounts, one coordinated play

Arkham Research identified three wallets at the center of the incident: addresses beginning 0xde9, 0x20e, and 0x67f. The mechanics, documented by Arkham and covered in detail by CoinDesk and Decrypt, followed a clear pattern.

The primary account (0xde9) opened a short position in JELLY perpetuals worth approximately $4.1 million notional. The two supporting accounts - 0x20e and 0x67f - simultaneously opened long positions of $2.15 million and $1.9 million respectively on the same token.

This is not a standard trade. It is a coordinated setup designed to exploit how Hyperliquid handles liquidations on illiquid tokens.

Dumping the position onto HLP

The short position was intentionally liquidated. When a position on Hyperliquid is too large to close through normal market flow - especially in a low-liquidity token - the liquidation engine passes the remaining position to HLP, the protocol's backstop liquidity vault.

HLP is funded by HYPE stakers and protocol participants. It functions as the market maker and insurer of last resort. By engineering a liquidation large enough to overwhelm available liquidity, the attacker effectively transferred a toxic short position directly onto HLP's balance sheet.

The long positions held by the supporting accounts were not liquidated. They were positioned to profit from any price spike that followed - which is exactly what happened when Binance listed JELLY hours later.

Halborn's post-incident security analysis described the attack as a deliberate exploitation of the interaction between low-liquidity perpetuals and Hyperliquid's liquidation-to-HLP fallback mechanism. This was not a bug in the traditional sense. It was a logical consequence of how the system was designed to handle insolvent positions.

How HLP absorbed the damage

HLP's unrealized loss peaked at negative $13.5 million as it held the inherited short while JELLY's price spiked. HYPE dropped approximately 20% during the episode as traders priced in the risk of a broader HLP insolvency event.

The Binance listing of JELLY was not coordinated with the attack - there is no evidence of that - but the timing was catastrophic for HLP. A 560% price spike on a token where HLP held a large short position compressed the window for any orderly resolution.

OAK Research noted that the attack exposed a structural vulnerability: HLP, while well-capitalized under normal conditions, was not built to absorb coordinated manipulation of low-liquidity tokens at this scale. The insurance fund existed, but the speed and magnitude of the price move left little room for automated risk controls to respond.

The governance response

Hyperliquid's validators voted to delist JELLY perpetuals and forcibly settle all open positions at $0.0095. The decision was made off-chain, through validator coordination - not through any on-chain governance mechanism or community vote.

The settlement price of $0.0095 was well below the spot price at the time. Traders holding long JELLY perpetual positions - including those with no connection to the attack - had their positions closed at a price that bore no relationship to what the token was trading for on spot markets.

What the validator vote actually revealed

Hyperliquid presents itself as a decentralized protocol. The JELLY response demonstrated that a small set of validators can, in practice, make unilateral decisions about settlement prices and market delistings - with no on-chain vote and no advance notice to traders.

This is not unique to Hyperliquid. Most protocols that call themselves decentralized have some form of privileged actor - a multisig, a foundation, a validator set - that can intervene in emergencies. What JELLY did was make that reality visible and concrete.

CoinDesk's reporting on the incident highlighted the gap between Hyperliquid's decentralization narrative and the actual decision-making process. The validators acted quickly, and arguably correctly from a solvency standpoint. But the mechanism they used was opaque, off-chain, and not subject to any formal governance process that traders could have reviewed in advance.

Changes after the criticism

Following significant criticism from the DeFi community, Hyperliquid made changes to its validator set composition and governance processes. The team acknowledged the centralization concerns and committed to expanding the validator set and improving transparency around emergency governance procedures. Those changes are ongoing. Whether they are sufficient is a separate question - and one active traders should track rather than assume resolved.

The "code is law" debate

The JELLY incident dropped the DeFi community into a debate that has never been fully settled: when a protocol can intervene to prevent catastrophic loss, should it?

The argument for intervention

HLP insolvency would have been a genuine systemic problem. If HLP had collapsed, every trader with open positions on Hyperliquid would have faced uncertainty about settlement. The insurance fund might not have covered the gap. Forced settlement at a chosen price, while harmful to some position holders, prevented a worse outcome for the broader protocol.

The argument is straightforward: protocols are not neutral infrastructure. They have stakeholders. Allowing a coordinated attack to drain the insurance fund and destabilize the platform is not a principled defense of decentralization - it is a governance failure.

The argument against

The counterargument is equally specific. Hyperliquid's validators chose a settlement price that protected HLP at the direct expense of traders who held legitimate long positions. Those traders did not attack the protocol. They were caught in a forced settlement at a price far below market.

That is selective enforcement. The protocol decided who bore the loss. In doing so, it demonstrated that the rules of the system are not fixed - they can be rewritten by a small group of validators when the stakes are high enough. That is a material risk that was not disclosed in any trading interface before the incident.

There is also a moral hazard argument. If validators can intervene to prevent losses, the threshold for future intervention is now ambiguous. The precedent is set. The next case may be less clear-cut.

The DAO hack parallel

The closest historical parallel is the Ethereum DAO hack of 2016. A vulnerability in The DAO smart contract allowed an attacker to drain approximately $60 million in ETH. The Ethereum community faced the same choice: let the code execute as written, or intervene.

The community chose intervention. A hard fork reversed the hack. A minority rejected it on principle and continued the original chain as Ethereum Classic. Whether that decision was right has never been fully resolved.

The JELLY incident is not identical - Hyperliquid's intervention was governance-level, not a chain fork - but the underlying tension is the same. When a decentralized system faces a choice between following its own rules and preventing a catastrophic outcome, what it chooses tells you what it actually is.

Hyperliquid chose to intervene. That choice has consequences for how you model governance risk on the platform going forward.

What changed at the protocol level

Hyperliquid made several concrete changes following the JELLY incident:

• Listing criteria tightened. New perpetual markets now face stricter requirements around liquidity depth and market capitalization before approval.
• Open interest caps for low-liquidity tokens. Hyperliquid introduced or tightened OI caps on tokens below certain liquidity thresholds, limiting how large a position can grow relative to available market depth.
• Insurance fund mechanics reviewed. The team reviewed how HLP absorbs liquidated positions from illiquid markets and adjusted the fallback logic to reduce exposure to coordinated manipulation of low-cap tokens.
• Validator set expansion. Steps were taken to broaden the validator set, reducing the concentration of governance power the off-chain vote exposed.

These are meaningful changes. They reduce the specific attack vector JELLY exploited. They do not eliminate governance risk. A different attack on a different token with different liquidity dynamics could still produce a similar situation - and the governance response might look similar.

What it means for active traders

The JELLY incident clarifies something many traders treat as abstract: non-custodial does not mean governance-proof.

You can hold your own keys, sign every order yourself, and never deposit funds with a centralized exchange - and still be subject to a forced settlement at a price chosen by a validator committee. Custody risk and governance risk are different categories. Both are real. Both belong in your risk model.

This is not an argument against trading on Hyperliquid. Hyperliquid has the deepest liquidity, fastest execution, and largest open interest of any perpetual DEX. None of that changed because of JELLY. But the incident is a clear reminder that the protocol layer carries risks that exist independently of whether you control your own wallet.

What you can control is your own risk posture:

• Setting hard limits on position size in low-liquidity tokens, where liquidation mechanics are most likely to produce adverse outcomes.
• Using configurable leverage caps so a single position cannot exceed your defined risk tolerance, regardless of what the protocol allows.
• Setting notional position limits that cap your total exposure before a trade is submitted.
• Using slippage controls that show you the worst-case fill price before you sign anything.

This is exactly the kind of tooling that Stacked Markets builds into its terminal. Configurable leverage caps, notional position limits, halt switches, and circuit breakers for rapid order bursts are front-end controls that sit between you and the protocol. They do not change what Hyperliquid's validators can do in a governance emergency. But they mean you are not relying on protocol defaults to manage your own exposure.

Every order on Stacked Markets routes through IOC limit orders with slippage bounds. The worst-case fill price is shown before the wallet signing prompt appears. Stacked Markets holds no funds and no keys - matching, margin, funding, and settlement all happen on Hyperliquid's on-chain order book.

The JELLY incident is a useful frame for thinking about this. Hyperliquid's protocol-level risk controls did not prevent the attack. The governance response worked but created new risks for traders with open positions. The lesson is not to avoid Hyperliquid. It is to add your own risk controls on top of whatever the protocol provides - because protocol-level controls can be changed, overridden, or simply inadequate for the specific situation you are in.

---

Configurable risk controls on top of Hyperliquid's liquidity. Leverage caps, notional limits, circuit breakers, and IOC slippage bounds - sitting between you and the protocol.

Start at Stacked Markets ->

FAQs

What was the Hyperliquid JELLY incident?

A coordinated manipulation attempt in which three wallets opened offsetting positions in JELLY perpetuals on Hyperliquid, then deliberately engineered a liquidation that forced the protocol's HLP vault to absorb a large short position. HLP's unrealized loss peaked at negative $13.5 million. Hyperliquid's validators responded by delisting JELLY and forcibly settling all positions at $0.0095 - well below the spot price at the time.

Who was behind the JELLY attack?

Arkham Research identified three wallets - 0xde9, 0x20e, and 0x67f - as the primary accounts involved. The short position of approximately $4.1 million was held by 0xde9. The two long positions totaling approximately $4.05 million were held by the other two addresses. The identities of the individuals behind these wallets have not been publicly confirmed.

What is HLP and why did it absorb the position?

HLP is Hyperliquid's liquidity provider vault, funded by HYPE stakers and protocol participants. It acts as the backstop for positions that cannot be liquidated through normal market flow. When the JELLY short position was too large to close against available liquidity, the liquidation engine transferred it to HLP. This is by design - HLP is the market maker and insurer of last resort on Hyperliquid.

How did Hyperliquid's validators decide the settlement price?

The decision was made off-chain through validator coordination, not through an on-chain governance vote. Validators agreed to delist JELLY and settle all open positions at $0.0095. That price was chosen to protect HLP solvency, not to reflect what JELLY was trading for at the time of settlement. Traders with legitimate long positions were settled at this price regardless of their involvement in the attack.

What is the "code is law" debate and how does JELLY relate to it?

"Code is law" is the principle that smart contracts and protocol rules should execute as written, without human intervention. The JELLY response was a direct rejection of that principle - validators intervened to override market prices and force a settlement. The debate mirrors the 2016 Ethereum DAO hack, where the community chose to fork the chain to reverse an exploit. JELLY shows that Hyperliquid will intervene when solvency is at risk.

What protocol changes did Hyperliquid make after JELLY?

Hyperliquid tightened listing criteria for new perpetual markets, introduced or tightened open interest caps for low-liquidity tokens, reviewed HLP's liquidation fallback mechanics, and took steps to expand the validator set. These changes reduce the specific attack vector JELLY exploited but do not eliminate governance risk entirely.

Does non-custodial trading protect you from governance risk?

No. Holding your own keys and signing every order yourself means you are not exposed to exchange insolvency or withdrawal freezes. It does not protect you from protocol-level governance decisions - forced settlements, delistings, or changes to liquidation mechanics - that a validator set can make unilaterally. Custody risk and governance risk are separate categories. Both require active management.