AutoCorp represents a pioneering vision for a fully autonomous company, powered by state-of-the-art Large Language Model agents. It was built by Barry Zhang, Bruno Faviero, Donny Greenberg, Mina Fahmi, Sergio Prada, and I during the NYC GPT/LLM Hackathon. The Large Language Model (GPT-4) function as the core controllers, producing reasoning traces and executing sequential actions.

AutoCorp generates a thesis for a direct-to-consumer t-shirt company based on a business owner’s initial idea. Subsequently, using an integrated system of GPT-4, DALL-E, and ControlNet, it crafts marketing, branding, and product design. As customers request customized t-shirts, AutoCorp promptly responds. Moreover, it continuously learns and refines its strategy based on customer feedback, updating its foundational thesis and all subsequent materials. This seamless flow has been demonstrated end-to-end, and can potentially be integrated with ad and custom t-shirt APIs for tangible real-world application. This demonstrates the possibility of a fully autonomous company that evolves towards an increasingly profitable business.


© 2023 Chuanyang Jin. All rights reserved.
Powered by Hydejack v9.1.6