Quasar documentation
An AI-powered systemic risk oracle that reads Hyperliquid's entire on-chain market state to detect and publish the risk of a liquidation cascade before it happens.
Introduction
Hyperliquid is a Layer 1 built for perpetuals trading where, unlike a centralized exchange, everything is public and on-chain: every open position, every leverage level, every liquidation price, and the full depth of the order book. You can read the risk state of the entire market, participant by participant.
Quasar turns that radical transparency into foresight. It continuously reads the on-chain state, uses an AI model to simulate how a price shock would cascade through the market, and publishes a single systemic risk score from 0 to 100 — on-chain, updated every block — that both humans and other protocols can act on.
Quasar is positioned as a public good — infrastructure that makes the whole ecosystem safer. It doesn't trade against users; it measures systemic fragility and makes it readable.
The cascade problem
Hyperliquid's structure makes it vulnerable to liquidation cascades. The network allows high leverage (up to 50x), and liquidity on some markets is thin. When a price shock hits, forced liquidations sell into the order book, which moves the price, which triggers more liquidations, which moves the price again. It's a domino effect.
Because every DeFi protocol built on top — lending markets, stablecoins, vaults on HyperEVM — liquidates against that same order book, when the core of the market shakes, the whole ecosystem shakes with it.
Today these protocols manage that danger with static, blind parameters: a fixed collateral ratio, a raw price feed. They don't see the cascade coming — they absorb it after the fact.
The single biggest systemic risk on Hyperliquid is the feedback loop between forced selling and price — and it's the exact thing static parameters can't measure.
How it works
Quasar runs a three-stage pipeline, every block:
Ingest on-chain state
Open positions, aggregate leverage, stacked liquidation prices, funding rates, open interest, and real order-book depth.
Model the cascade
An AI model runs shock scenarios against the real position map: what liquidates, how much depth absorbs it, the slippage, and the second-order liquidations that follow.
Write the score
The result collapses into one 0–100 score, published to a HyperEVM contract that any protocol can read in a single call.
Why AI earns its place here
Modelling how a shock propagates through thousands of interconnected positions and an order book of varying depth is a non-linear, chaotic system that static parameters can't capture. A model trained on the full on-chain history — available precisely because of Hyperliquid's transparency — learns market regimes and hidden correlations. Because positions are public, it can even learn the behavioural patterns of large actors to anticipate how they'll react.
Systemic risk score
The score condenses the whole simulation into one number from 0 to 100, mapped onto a light spectrum — a nod to quasar redshift — from calm blue to cascade red. It has two audiences: for humans it's a live dashboard reading of how close the market is to breaking; for machines it's a value to read and act on.
The bands above are illustrative. Exact thresholds and the scoring model will be published and documented at launch.
Products
One oracle, surfaced four ways:
Systemic Risk Score
A single number for the whole market, recomputed every block from live leverage, liquidation clustering and book depth.
Cascade Simulator
Stress-test on demand: "if BTC drops X%, what liquidates, how much depth absorbs it, and what breaks next?" — second-order effects included.
On-chain Risk Oracle
The score lives in a HyperEVM contract. Protocols read it with one call and adjust collateral ratios, rates or hedges before the cascade.
Whale Position Intelligence
Tracks the largest positions — size, leverage, distance to liquidation — and flags when whale risk starts to concentrate.
Why Hyperliquid
Quasar is defensible because you couldn't build it anywhere else:
- On a centralized exchange, positions are hidden — you don't have the data.
- On Ethereum, risk is scattered across dozens of isolated protocols with no unified order book — you have no consolidated view.
- On Hyperliquid, everything converges on one transparent order book, and native composability between HyperEVM and the core means the oracle can be consumed on the spot, with no bridge.
Data, risk computation, and corrective action all live in the same environment.
Oracle integration
For builders, Quasar is infrastructure, not a trading bot. Wire the score into your protocol's logic and it defends itself. The oracle exposes a minimal, gas-cheap interface:
// Interface preview — final address & ABI published at launch
interface IQuasarOracle {
/// Systemic risk score, 0 (calm) to 100 (cascade)
function riskScore() external view returns (uint8);
/// Unix timestamp of the last on-chain update
function lastUpdated() external view returns (uint64);
}Reading the score
A consuming protocol reads the score in a plain view call — no keeper, no
off-chain infrastructure. Here a lending market tightens its collateral requirement as
systemic risk rises:
contract RiskAwareLending {
IQuasarOracle public immutable quasar;
constructor(address oracle) {
quasar = IQuasarOracle(oracle);
}
// Minimum collateral ratio scales with systemic risk
function minCollateralRatio() public view returns (uint256) {
uint8 risk = quasar.riskScore();
if (risk > 75) return 180; // brace before the cascade
if (risk > 50) return 150;
return 130; // calm waters
}
}Vaults can scale hedges up ahead of volatility and down after it passes; stablecoins and LPs can throttle exposure to the shared book. The same one-call pattern applies.
Roadmap
- Public dashboard. The Risk Monitor goes live — systemic score, liquidation map and whale intelligence, free for everyone.
- Predictive model. The AI cascade engine trains on historical liquidation events and starts scoring forward-looking scenarios, not just current state.
- On-chain oracle + integrations. The score is published to HyperEVM; first lending markets and vaults wire it into their risk parameters.
- Decentralized feed. Multiple independent operators compute and attest the score, removing Quasar itself as a single point of trust.
FAQ
Is Quasar a trading bot or a signal service?
Neither. Quasar is risk infrastructure — a public good for the Hyperliquid ecosystem. It doesn't trade, front-run or sell alpha; it measures systemic fragility and publishes that measurement for anyone to use.
Can I trust a single score?
The score is a summary, not a black box: the dashboard shows the inputs behind it — liquidation clustering, depth, leverage, whale concentration. The final roadmap phase decentralizes computation across independent operators so no single party, including us, is a point of trust.
Why can't this exist on other chains?
It needs a single, transparent, on-chain order book plus native composability to act on the result in the same environment. Hyperliquid is the only place all of that holds at once.
Disclaimer
Quasar is pre-launch. Nothing in this documentation is financial advice. All figures, scores, interfaces and addresses shown are illustrative and subject to change before release. Nothing here is a guarantee of any outcome, and reading the risk score does not remove risk — it's one input into a protocol's own risk management.