This post details a Bash script that runs a TDD loop where an LLM is iteratively used to write Python application code to make a fixed set of tests pass. Similar things have been done before1; this script merely demonstrates that such things can be done simply by stitching together open-source LLM tools. Tooling The script uses two commands from Simon Willison: llm — a command line interface for interacting with Large Language Models (LLMs).
