We replaced the research desk with a swarm of agents. They generate, stress-test, and trade quantitative strategies — 24/7, across every major market.
Four things kept everyone else out of quant: data, compute, talent, infrastructure.
All four just collapsed.
| Strategy | Class | Inception | AUM | Sharpe | Max DD | YTD | 30d | Status |
|---|---|---|---|---|---|---|---|---|
Momentum Alpha BTC · ETH · SOL — 4H | Trend-following | Jan 2025 | $34.2M | 2.87 | −3.2% | +34.7% | Live | |
StatArb Engine Crypto / Equity — 5m | Statistical arb | Mar 2025 | $28.9M | 3.41 | −1.8% | +22.1% | Live | |
Mean-Rev BTC Single-asset — 15m | Mean reversion | Apr 2025 | $18.4M | 1.92 | −4.7% | +14.6% | Live | |
Vol-Surface Arb Options — BTC/ETH | Volatility | Jun 2025 | $22.7M | 2.44 | −2.9% | +18.3% | Live | |
Cross-Exchange Basis Perp / Spot — 1m | Carry | Jul 2025 | $11.3M | 4.12 | −0.6% | +9.8% | Live | |
Vol-Scalp v2 ETH options — 1m | Volatility | Pending | — | 2.21* | −2.1%* | — | Validation | |
Regime-Switch SOL SOL — 1H | Trend-following | Pending | — | 1.74* | −5.4%* | — | Validation | |
Pairs-Trade Mining Multi-asset — exploratory | Statistical arb | — | — | — | — | — | Generation |
Each agent proposes a strategy — entry, exit, risk, leverage, asset universe. The system can spawn a thousand candidates per regime in under a minute.
Atlas runs every candidate through five elimination gates. The survivors aren't optimised — they're the ones that didn't break under stress.
Monte Carlo. Walk-forward. Adversarial regimes. Live-paper for six weeks before a single dollar moves. If the seal isn't green, the strategy stays in the lab.
Order-book imbalance, funding rates, on-chain flows, liquidations. The same telemetry institutional desks consume — piped into every agent's prompt.
For four decades, ~100 firms have captured 60–70% of quant profits. The other $10T+ in capital had no way in.arquant.ai is that way in.