Developers use GPT-4 and Claude @AnthropicAI to create a self-evolving quantitative system, with the first strategy achieving an annualized return of 23%. A developer worked through two nights to successfully build a self-evolving quantitative trading system. During the development process, approximately $500 was spent on GPT-4, and another $500 was spent on Anthropic's Claude model. The first dynamic strategy backtest generated by the system's iterations shows an annualized return of 23%, with a maximum drawdown of 17%.