TXT2TXN: Using AI (LLMs) for Intent-Based Applications

Circle Internet Financial and Blockchain at Berkeley have collaborated to create TXT2TXN, an open-source web application that utilizes Large Language Models (LLMs) to interpret and execute user intents on the blockchain. This innovative tool allows users to express their desired outcomes in natural language without detailing the exact execution steps, which are handled by the application and solvers. The system leverages AI to classify user intentions, such as token swaps or transfers, and translates them into executable blockchain transactions. The project demonstrates the potential for improved user experiences in crypto applications through intent-based architectures, highlighting the role of LLMs in simplifying user interfaces and enhancing transaction accuracy. While the initial tests show promise, the developers acknowledge the need for further research to refine the technology, improve accuracy, and address privacy concerns associated with LLM use.