Chris Parsons

This talk introduces "Ralph loops," a method for building AI agents that can iteratively improve their own work by repeatedly attempting a task. The core idea is to leverage the ability of modern AI models to self-correct and refine outputs through repeated execution of a prompt or instruction. This approach moves away from complex, brittle orchestration workflows towards simpler, more robust agentic loops that can be applied to various tasks, including coding and content generation.