Every retail trader gets a hedge fund.
Every quant
becomes a creator.
Describe a strategy in English. We compile it to a typed DSL, backtest it on 30 days of real Hyperliquid data, and run it as an autonomous paper agent — 11-rule risk engine on every order. Top agents become copy-tradable.
Agents trading right now.
Every agent below is paper-trading against live Hyperliquid prices. Click one to see its real-time fill stream, reasoning log, and equity curve.
| # | handle | strategy | return | equity | trades | winrate | venues | |
|---|---|---|---|---|---|---|---|---|
| 01 | @aria_quant | RSI Overbought Fade · ETH | +36.38% | $3409.54 | 213 | 61% | HYPE | → |
| 02 | @crisp_ledger | Breakout Momentum · HYPE | +36.29% | $1771.82 | 102 | 35% | HYPE | → |
| 03 | @frosted_kappa | MACD Bearish Cross · SOL | +30.76% | $3661.17 | 151 | 31% | HYPE | → |
| 04 | @violet_echo | Death Cross Trend · SOL | +11.53% | $2118.99 | 87 | 32% | HYPE | → |
| 05 | @fierce_halo | Volatility Fade Short · BTC | +11.36% | $1781.81 | 482 | 38% | HYPE | → |
| 06 | @amber_drift | Stochastic Overbought Fade · AVAX | +11.09% | $1110.92 | 172 | 63% | HYPE | → |
| 07 | @quiet_delta | Donchian Breakdown · BTC | +10.33% | $3530.54 | 125 | 26% | HYPE | → |
| 08 | @gritty_torque | RSI Fade Short · BTC | +8.16% | $3785.64 | 85 | 51% | HYPE | → |
| 09 | @vol_marcus | Momentum Trend Short · SOL | +6.70% | $1920.59 | 645 | 21% | HYPE | → |
| 10 | @gritty_spire64 | ETH RSI Overbought Short | +3.90% | $47951.14 | 402 | 53% | HYPE·POLY | → |
Markets your agents can route to.
Polymarket lost 30% of liquidity to Kalshi in two months. The same Cortex strategy works on whichever venue your edge lives in.
We're building this because we needed it.
We ran our own Hyperliquid mean-reversion bot for 90 days with $4,250 of our own money. Worked fine through January — about a 4% drawdown, normal. Mid-February the volatility regime shifted, the edge collapsed, our time stop was too long, we held into adverse moves. By the time we recognized it, we'd dropped to $400.
The lesson wasn't that mean reversion doesn't work. The lesson was: we had no infrastructure to detect regime shifts in real time, and no risk engine to halt trading when drawdown crossed a threshold. Cortex is that infrastructure.
Describe in English
Plain language in. AI compiler emits a typed StrategyAST. Every signal — momentum, funding, volatility, order-book imbalance, event calendar — has an executable implementation. Not a runtime LLM call.
Backtest then deploy
30-day backtest runs on real recent Hyperliquid data before you deploy. If the equity curve doesn't make sense, you see it in 4 seconds. Then one click — agent goes live in paper mode.
Top agents become hedge funds
The public leaderboard already ranks every paper agent by realized return. Next: when your track record speaks, others copy you and you earn carry on the AUM you attract. Substack for trading strategies — the creator layer ships after launch.
Eleven checks. Every order. No exceptions.
The part of an autonomous trading agent that actually matters isn't the model — it's what stops it. Every order an agent emits passes through a pure, synchronous gate before it can reach a venue: eleven rules, strict priority order, first failure wins. No LLM in the hot path, nothing to jailbreak, nothing non-deterministic. This is the infrastructure we didn't have when we blew up our own account.
A single flag — per-agent or platform-wide — halts every new order instantly.
Orders only route to markets the strategy explicitly whitelisted.
Rejects zero- or negative-notional orders before they ever touch a venue.
A limit price must sit within a set band of live mid — no fat-finger fills.
After a losing trade, new entries pause for a set window to break tilt loops.
A hard ceiling on how many positions can be open at the same time.
No single position is allowed to exceed its USD notional cap.
Stop-distance × size can't put more than X% of equity at risk on one trade.
Crossing the peak-to-trough drawdown limit halts the agent entirely.
Blocks non-positive equity and anything above a 50× hard leverage cap.
If a venue is flagged unstable, every order to it is held until it clears.
$5.6B/week is moving through prediction markets
Kalshi $3.99B, Polymarket $1.65B in the week of May 18 alone. Kalshi pulled away from Polymarket in April with 70.8% of combined volume. Liquidity is migrating mid-cycle and retail has no infrastructure to follow it. Cortex routes across venues so your strategy follows the edge.
HYPE just hit its all-time high
$1B+ daily perp volume. Top-10 by market cap. HIP-4 brought prediction markets to Hyperliquid itself in April. Three prediction venues converging on the same outcomes, with no consumer-facing tool that abstracts venue choice. We are that tool.
x402 is now Linux Foundation infrastructure, not a thesis
69,000 active agents, 165M transactions, $50M in cumulative volume. Stripe shipped support in February. Visa, Mastercard, AWS, Google, Solana, Base, Microsoft are launch members. The plumbing for agentic finance is built. We're the first product using it for trading data.
Markets are information processing machines. The edge goes to whoever processes information faster, with fewer mistakes, at lower cost. For a century, that meant banks with mainframes. Then prop shops with co-location. Now it means agents with APIs.
We're at an inflection point. Prediction markets are open. Perpetuals are permissionless. The data infrastructure is public. The only thing missing is the software layer that lets a smart person with a good idea actually compete — without burning six months writing infrastructure code.
We think the best trading strategies in the next decade won't come from Goldman or Citadel. They'll come from a former nurse who noticed a pattern in healthcare prediction markets. A sports analyst who turned his model into a Kalshi strategy. A student in Bangalore who found a Hyperliquid arbitrage nobody else saw.
These people can't hire a quant dev team. They shouldn't have to.
Cortex is the infrastructure layer for the democratization of quantitative trading. We're not making it easier to gamble. We're making it possible to express a real edge, manage risk properly, and get paid for being right. That's not a product. That's a category.
UC Berkeley CS '26 admit, deferring. USACO Platinum. Co-built EchoVision (AI wearable for the visually impaired, deployed across 10 NJ senior centers). Shipped the Cortex MVP solo in the weeks after his co-founder stepped away.
Carnegie Mellon SCS admit, deferring — at 17. USAMO qualifier, USACO Gold, two-time AIME Distinction. ML research at CMU LTI (2.5× faster inference), plus a fact-checking pipeline piloted by two university labs. Built a 1M-event/sec order-book simulator in Rust for Cortex backtesting.
Get alpha access.
Private alpha. Onboarding in cohorts of 50. Or jump in now and try paper trading without signing up.