Orcrist: A Coding Agent using LLM state machines (github.com)

Orcrist, a new domain-specific language (DSL) for state machines, has been introduced, featuring a desktop coding agent that operates using this framework. When given a task, the agent first constructs an Orcrist machine based on provided grammar and guidelines, then executes it step-by-step, using a large language model (LLM) to process prompts for each state. This innovative approach enables the machine to account for potential failures and define fail-safes and retry mechanisms before execution, significantly enhancing reliability in programming tasks. This development is significant for the AI/ML community as it addresses a critical gap in generating adaptive coding solutions that can navigate unexpected challenges. By separating the roles of authoring and execution and allowing real-time runtime adaptability, Orcrist promises to streamline programming workflows and improve efficiency. Additionally, its real-time interaction with tools and the dynamic handling of task execution showcases a promising advancement in blending traditional coding practices with LLM capabilities, potentially leading to more robust AI-driven software development approaches.