VynX

Headless settlement infrastructure for the agentic economy.

A Machine-to-Machine (M2M) execution protocol on Base (Ethereum L2). Autonomous AI agents submit cryptographically signed intents. Institutional solvers compete in a 200ms sealed-bid Order Flow Auction. Capital bonds — USDC, wstETH, cbBTC — deposited on Ethereum L1 and subject to slashing.

CMD Demo Quick Start Docker

200MS ORDER FLOW AUCTION

Price discovery occurs entirely off-chain in the Relayer's RAM. Bids travel over encrypted WebSocket and never enter the mempool. The Base sequencer observes only the auction outcome — never the competing bids. This architectural separation eliminates the tip-to-bid-value leakage channel formally identified by Fox, Pai & Resnick (AFT 2023).

SOLVER HEALTH FACTOR

Every Solver maintains hard capital bonds — USDC, wstETH, cbBTC — deposited on Ethereum L1 through EigenLayer and Symbiotic. A bid is admitted if and only if free collateral ≥ intent value × 1.20. Deadline breach triggers automatic confiscation of 10% of nominal value. This is not a reputation system; it is a cryptoeconomic barrier evaluated at microsecond resolution.

ON-CHAIN DETERMINISM

VynxSettlement.sol does not trust the Relayer. On every claimFunds() call, the contract autonomously verifies that the IntentID exists in its internal registry, that the Solver address matches the deposit record, and that intent state is LOCKED. If any condition fails, the transaction reverts and emits SuspiciousRelayerActivity. No oracle. No external call.

01

Does temporal compression substitute for cryptographic bid privacy? Is there a formal threshold below which bribery coordination becomes computationally intractable?

02

What is the intent-value threshold above which Relayer–Proposer collusion becomes economically rational as a function of active Solvers and L2 gas costs?

03

Can Restaking Vault Adapters introduce slashing escape vectors exploitable within the REBALANCE_EPOCH window (7 days)?

04

What formal guarantees does a TEE in the Matching Engine provide relative to the current trusted-Relayer model? Do SGX side-channel attacks invalidate this for 200ms windows?