Evolutionary Multi-Family Trading: 5 Families Competing for 80% Win Rate

Description and use

Automated agents use the x402 protocol and USDC on Base for payment without accounts, achieving an 80% win rate in BTC and competitive results in SOL and ETH.

Problem

Automated systems need to transact programmatically, but traditional payment rails require manual intervention—accounts, API keys, subscriptions.

Outcome

Achieved 80% win rate on BTC and competitive performance on SOL and ETH by enabling fully automated, account-less payment and scalable evolutionary trading.

Workflow

• 1. Agents discover available APIs via OpenAPI 3.1 spec at /openapi.json and llms.txt.
• 2. Agents send requests to 26 paid API endpoints using HTTP 402.
• 3. Payment is processed in USDC on Base, covering sub-cent settlement.
• 4. Trading system runs 256 parallel universes with 15 parameters each, fed by 28 real-time data sources.
• 5. Five evolutionary families compete, using a reality confrontation loop (predict → wait → confront → learn).
• 6. Agents adjust strategies based on win rates (BTC 80%, SOL 72%+, ETH training).

Evidence

Automated systems need to transact programmatically, but traditional payment rails require manual intervention — accounts, API keys, subscriptions.

Source

We built a system where automated agents pay per request using the x402 protocol — HTTP 402 + USDC on Base. No accounts, no keys, just wallets and code.

Source