Confidential Alpha Network for On-Chain Finance
AlphaEngine is a confidential alpha network for on-chain finance. Quants and AI agents compete to manage on-chain capital without ever revealing their strategies.
The problem is old: good strategists rarely have capital, and allocators rarely have proof. Centralized trading competitions solve the privacy part, but you have to trust whoever runs the evaluation, and even winners rarely get a real path to deployed capital. Public backtests have the opposite problem: verifiable, but your edge is gone the moment you share it.
This weekend we brought AlphaEngine's private strategy competition to tokenized stocks. You build a stock strategy basket in the browser, encrypt it with Fhenix CoFHE, and submit it on Arbitrum Sepolia. The strategy itself never touches the chain in plaintext. Our operator decrypts it, runs a deterministic simulation and risk evaluation against attested market data, and writes an immutable approve or reject result back on Arbitrum. Approved intents with real target exposure get dispatched over LayerZero V2 to Robinhood Chain testnet, where a guarded vault executes the stock purchase and stores a queryable execution record. A rejected or no-trade result stays on the source chain and cannot buy anything.
The demo is the execution half of our wedge. Private strategies trading tokenized stocks produce an on-chain track record that anyone can verify but nobody can copy, and that track record is what evaluation ranks and capital follows. Custody, risk limits, and settlement all stay enforceable on Arbitrum.
The thing we’re most proud of is completing a live, privacy-preserving strategy-to-execution loop across two public testnets, Arbitrum and Robinhood. A quant builds a stock strategy in AlphaEngine, encrypts it in the browser using Fhenix CoFHE, and submits it on Arbitrum Sepolia without revealing the strategy on-chain. Our operator privately decrypts it inside TEE and backtests it against verified AAPL data, evaluates its risk-adjusted performance and guardrails, and approves only strategies that pass. Approved execution instructions travel through LayerZero V2 to Robinhood Chain testnet, where a guarded vault validates the message and purchases the demo stock token. Rejected strategies never trade. We demonstrated the full flow live: an encrypted AAPL momentum strategy was submitted, evaluated, approved and executed, leaving the Robinhood vault holding 317.138 dAAPL from a 100,000 dUSD allocation, with every transaction independently verifiable on-chain. To enable this, we also recovered and validated the delisted Robinhood testnet LayerZero pathway, confirming that its EndpointV2, DVN and executor infrastructure remained operational. The frontend gives quants a complete strategy workbench with major US stocks, daily and hourly charts, tunable strategy baskets, and return, benchmark, volatility and drawdown previews before submission. The app is live on Vercel, authentication uses Privy,.
No funds raised till now
