Article: Experimenting with LLMs for Developer Productivity

This article describes an experiment that sought to determine if no-cost LLM-based code generation tools can improve developer productivity. The experiment evaluated several LLMs by generating unit tests for some open-source code and measuring the code coverage as well as the manual rework necessary to make the tests work.

ai article code code coverage code generation cost coverage developer developer productivity development experiment large language models llm llms measuring ml & data engineering productivity tests tools unit-testing work