The protocol for hidden leverage

Abstract

Every leveraged position on a public chain shows its size and the exact price at which it will be liquidated. Trackers publish that data in real time, and the market has learned to use it: traders coordinate to push price into a known liquidation level, copy-traders mirror large wallets, and liquidation hunters wait at the line.

Mutexa is a privacy layer for leverage. It works on top of the perps and lending venues traders already use, including Hyperliquid, Solana perps, EVM perps, Aave and Morpho. Mutexa splits each position into slices held by accounts that cannot be linked to the trader or to each other. Each slice carries different margin or loan-to-value, so the single liquidation price becomes a band. Each slice stays below common whale-alert sizes, so trackers see ordinary users instead of a whale. The trader sees the whole book in one view, computed on their own device. The market only ever sees the slices.

Mutexa is non-custodial. Keys are generated and kept on the user's device, and Mutexa never holds funds.

1. Introduction

Transparency is one of the strengths of on-chain finance. Anyone can verify a venue's solvency, audit its open interest and check that a liquidation happened at the price the rules say it should. That same transparency also exposes every individual trader.

On a centralized exchange, a trader's liquidation price is known to the trader and the exchange. On a public perps venue it is known to everyone. On Hyperliquid, explorers and trackers such as HypurrScan and CoinGlass list any wallet's positions, leverage and liquidation price. On Solana perps, on GMX, on Aave and on Morpho, the same information sits in public accounts and is reposted by analysts as it happens.

For small positions this rarely matters. For large ones it changes the game. A position whose liquidation price is public is a target, and the larger the position, the more people are watching it.

Mutexa exists to give leveraged traders and borrowers the same basic privacy they would expect anywhere else: the market should not know exactly where you break.

2. The problem: a public liquidation price is a target

2.1 Liquidation hunting

When a liquidation price is public, it becomes a coordination point. In March 2025, a trader publicly recruited others on X to push Bitcoin into the liquidation level of a Hyperliquid whale's 40x short, worth about $368M, with liquidation at $85,592. 10x Research warned that "this transparency opens the door for coordinated efforts."

The pattern has continued. In May 2025, James Wynn's $1.1B, 40x Hyperliquid position was followed in real time. In May 2026, a whale sold about $36.8M of HYPE to defend a $103M short as price approached the liquidation level everyone could see. In June 2026, a closely watched ETH long on Hyperliquid was liquidated seven times in ten hours while trackers broadcast each new liquidation zone. On October 10, 2026, trackers showed a 391 BTC short at 13x with its liquidation price, about $99,801, for anyone to aim at.

2.2 Lending is just as exposed

A loan has a liquidation price too. Analysts regularly publish the exact price at which large Aave positions would be liquidated. One flagged a 58,000 ETH leveraged position that would be liquidated if ETH fell to $1,555. In March 2026, an incorrect wstETH price liquidated about $27M of Aave positions at once, a reminder of how much leveraged capital sits at known, clustered levels.

2.3 Copy-trading and front-running

The same public feeds that liquidation hunters read are read by copy-traders and front-runners. A large wallet's entries, exits and stops become signals for everyone else. A trader with an edge loses it the moment their wallet becomes worth watching.

2.4 Existing privacy tools don't solve this

Privacy in crypto has so far focused on three things, none of which hides an open position:

• Identity privacy (NEAR Confidential Intents, Ballast) hides who is trading and where the capital came from. The position and its liquidation price still appear on the venue.
• Transfer privacy (Railgun, Hinkal, Privacy Pools, Solana shielding pools) hides deposits, withdrawals and balances, not open positions.
• Private venues (Aster Hidden Orders, Omni by Variational) offer privacy by asking traders to move to a different venue, with different liquidity.
• Encrypted DeFi infrastructure (Fhenix, iExec) supports new protocols built on its own stack. It does not change what existing venues publish.

As of October 2026, no live product hides position size and liquidation price on existing public perps or lending venues, on any chain.

3. Design principles

Mutexa is built on five principles.

1. Stay on the venues with the liquidity. Traders should not have to give up the deepest order books and pools to get privacy. Mutexa adds privacy on top of public venues instead of replacing them.
2. No single line to hunt. A position should not have one public liquidation price. It should have a band, spread across accounts nobody can connect.
3. Look ordinary. Each visible piece of a position should look like a normal user, below the sizes that trigger whale alerts.
4. Keep the keys and the full picture on the device. The only place the whole position exists is the trader's own device.
5. Be honest about the limits. Privacy has costs and boundaries. Mutexa shows them before every trade.

4. How Mutexa works

4.1 Slices and unlinkable accounts

When a trader opens a perp position or takes a loan through Mutexa, the position is split into slices. Each slice is held by its own account on the venue.

Accounts are derived on the device from a signature the trader makes, and the keys never leave the device. Each trade or loan cycle can use fresh accounts that have never been used before, so there is no history connecting them.

4.2 Private funding rails

An account is only unlinkable if its funding is unlinkable. Each slice account is funded through a privacy rail, so no deposit connects it to the trader's main wallet or to the other slices.

Rails are chosen per chain. Candidates include shielded pools such as Railgun on Ethereum, Arbitrum, Polygon and BNB Chain, NEAR Confidential Intents for cross-chain funding, and Solana shielding pools such as Umbra. Cross-chain funding lets a trader fund slices on one chain from assets held on another without leaving a public bridge trail.

4.3 The liquidation ladder

Each slice is given different margin or loan-to-value. Because liquidation price depends on margin, the slices liquidate at different prices. Instead of one public line, the position's risk is spread across a band.

This changes the economics of hunting. Pushing price to the top of the band liquidates one small slice, not the whole position, and nobody watching can tell the slices belong together.

4.4 Alert-safe sizing

Whale trackers alert on size. Mutexa sizes each slice to stay below common alert thresholds on each venue, so no single slice triggers an alert, attracts copy-traders or draws a crowd.

4.5 Venue adapters

A venue adapter speaks each venue's own interface: its order types, margin rules and account model. Mutexa starts with the largest public venues. Candidates include Hyperliquid, Jupiter Perps and other Solana perps, EVM perps on Arbitrum and Base, and lending markets such as Aave, Morpho and Kamino. Launch venues are announced at launch.

A venue adapter SDK lets builders add new venues, so coverage grows beyond what Mutexa ships itself.

4.6 Jittered execution

A split position is only private if the slices don't arrive at the same moment with the same size. Mutexa places orders and loans slice by slice with randomized timing and size. A private TWAP can work a large entry or exit across slices and over time. Maker slices can rest post-only orders where the venue has an order book.

4.7 The local risk engine

The trader's device runs a risk engine that watches every slice on every chain. When price nears the closest liquidation level, it can add margin or repay through the rail, or trim that slice.

It never moves funds directly between slices. A direct transfer would link them on-chain. All rebalancing goes through the rail.

4.8 Hidden stops

A resting stop order on a public venue is one more line to hunt. Mutexa can keep take-profit and stop-loss levels on the device and send them only when they trigger, so there is no visible stop waiting on the book.

4.9 Private exit

On exit, slices close with jitter and funds return through the rail to a fresh address or a shielded pool.

4.10 One book across chains

The trader sees the complete picture in one app: total size, blended entry, PnL, health and liquidation band for every position on every chain. It is computed on the device from the trader's own accounts. Nothing on-chain links the pieces together.

5. Privacy model

5.1 What Mutexa hides

Mutexa hides a position's total size and its liquidation price from public trackers, analysts, copy-traders and liquidation hunters. It hides the link between slices and the link between the slices and the trader's main wallet.

5.2 What Mutexa does not hide

Mutexa is designed to be clear about where its protection ends:

• Venues still see their own accounts. Mutexa hides positions from public trackers, not from the venues themselves.
• Clustering by timing is still possible. A skilled analyst can try to group slices by when they were opened. Jittered execution and fresh accounts make this harder, not impossible.
• Very large positions still move markets. Splitting hides who holds a position, not the market impact of trading it.
• Privacy has a cost. Splitting adds venue fees, and privacy rails add their own fees and delays.
• Integrator fees mark fills. Where Mutexa collects fees through a venue's integrator or builder code, an observer can tell a fill came through Mutexa. In that case a trader's cover is the set of all Mutexa users.

5.3 Measuring privacy before trading

Mutexa gives traders the tools to judge their own privacy:

• Tracker view shows a trader's slices the way public trackers and analysts see them, before and after a trade.
• Linkability score gives a plain rating of how easily the slices could be clustered, with suggested fixes.
• Cost preview shows the extra fees, bridge costs and time that privacy adds, before the trader confirms.

6. Security

Non-custodial by design. Keys are generated and kept on the user's device. Mutexa's servers never hold funds or keys. Funds sit in accounts only the user controls.

Open source. The client, the venue adapters and the slicing logic are planned to be open source, published on the official Mutexa GitHub at github.com/mutexa-labs, so anyone can verify what the software does with their keys and orders.

Network privacy. API traffic can be routed through Tor so that a shared IP address doesn't tie slices together.

Audit. An independent security audit is completed before launch.

7. Product

Mutexa brings these capabilities together in one app:

Paper mode lets anyone rehearse a split with live prices before using real funds. Desk mode lets several traders share a risk view without any on-chain link between them. Agent access lets AI trading agents trade through Mutexa so their strategy isn't readable from one public wallet. Proof of PnL, planned for a later phase, will let a trader prove their total result to a partner without revealing the accounts behind it.

8. Business model

Mutexa earns from volume, not from holding money. The protocol charges a small fee on routed notional. Where a venue has an integrator program, such as Hyperliquid builder codes or Jupiter referral fees, the fee is collected through it. Where that would weaken privacy, a separate fee path is used. Rates are published before launch.

9. The $MUTEX token

$MUTEX is the access token for Mutexa. It is issued on Robinhood Chain (chain id 4663), an Ethereum-compatible layer 2 built on Arbitrum. Robinhood Chain is the token's home; Mutexa itself keeps working across every venue it supports, on any chain. The contract address is announced at launch.

Holding $MUTEX unlocks more of the protocol: more slices per position, more venues, faster rails and lower routing fees. Tier thresholds are set at launch.

The token is designed around a simple, fair model: a fair launch with a fixed supply, no presale, no team allocation, no staking and no emissions. If Mutexa ever runs a buyback, it is funded only from real fee revenue.

10. Roadmap

Being first matters, so the order of work is deliberate:

1. Identity. Domain and official channels.
2. First venue. The position splitter and liquidation ladder on the first venue, with tracker view, linkability score and paper mode.
3. Token. The $MUTEX launch and access tiers.
4. Coverage. More perps venues, then lending markets and cross-chain funding.
5. Builders and teams. Strategy API, agent access, desk mode and the venue adapter SDK.

Mutexa respects each venue's terms and geographic restrictions.

11. Conclusion

On-chain leverage has a structural flaw: the most sensitive number in a position, the price at which it breaks, is published for everyone. As perps and lending volumes grow, that flaw becomes more expensive for every trader worth watching. Hyperliquid alone carried about $11.1B in open interest in October 2026, and Solana perps reached $255.6B in volume in the first half of the year.

Mutexa fixes the flaw without asking anyone to leave the venues they trust. Positions are split across unlinkable accounts, liquidation prices become bands, and the full picture lives only on the trader's device.

Leverage, unseen.

References

Market data

• Hyperliquid public API, October 18, 2026: about $11.1B open interest and about $2.6B 24-hour notional volume across 234 markets.
• Solana perps reach a record $147B in Q2 2026 and $255.6B in the first half, up 57.1% year on year. CryptoBriefing, July 2, 2026.

Liquidation hunting and exposure

• Traders targeting Hyperliquid whales' positions trigger liquidations, 10x Research. Cointelegraph, March 18, 2025.
• James Wynn's $1.1B Hyperliquid position tracked in real time. Cointelegraph Magazine, November 13, 2025.
• Public whale liquidations as a trading signal on Hyperliquid. CryptoSlate, June 2026.
• Whale sells about $36.8M of HYPE to defend a $103M short. Bitcoin.com, May 2026.
• 391.48 BTC short on Hyperliquid with liquidation near $99,801. ChainCatcher, October 10, 2026.
• About $27M of Aave positions liquidated after a wstETH price glitch. CoinDesk, March 10, 2026.
• 58,000 ETH leveraged Aave position flagged with liquidation at $1,555. Phemex.

Mutexa

• Website: mutexa.io. Docs: docs.mutexa.io. App: dapp.mutexa.io. GitHub: github.com/mutexa-labs.

Token chain

• Robinhood Chain, chain id 4663, Arbitrum-based layer 2 with ETH as gas. Explorer: robinhoodchain.blockscout.com.

Related projects

• NEAR private perps on Hyperliquid. CryptoBriefing, September 17, 2026.
• Ballast. GitHub.
• Aster Hidden Orders. The Defiant.
• Omni by Variational. Blockworks.
• Fhenix privacy landscape. Fhenix. iExec privacy roadmap. iExec.
• Privacy infrastructure for DeFi. Hinkal.
• OffMarket, a privacy layer on Polymarket's order books. The Block, October 14, 2026.

Third-party contracts referenced (checked on-chain October 18, 2026; no Mutexa contract exists yet)

• Hyperliquid Bridge2 on Arbitrum: 0x2Df1c51E09aECF9cacB7bc98cB1742757f163dF7
• USDC on Arbitrum: 0xaf88d065e77c8cC2239327C5EDb3A432268e5831
• Railgun proxy on Arbitrum: 0xFA7093CDD9EE6932B4eb2c9e1cde7CE00B1FA4b9
• Aave V3 Pool on Ethereum: 0x87870Bca3F3fD6335C3F4ce8392D69350B4fA4E2
• Jupiter Perps program on Solana: PERPHjGBqRHArX4DySjwM6UJHiR3sWAatqfdBS2qQJu

This document describes the design of Mutexa. It is not financial advice and not an offer of any token. Leverage carries risk of loss.