Unit testing is the cornerstone of ensuring code quality and stability in software development. However, the traditional approach to unit testing can be cumbersome and time-consuming, often diverting developer focus from creating new features(ie. Nobody wants to do it if focused on development). In his presentation „LLM testing, or how I learned to stop worrying and, let an LLM to worry about my unit tests“, Henry introduces a way to harness the power of Language Model (LLM) technology and manage the control flow for creating tests in a controlled way. There is a proof of concept demo in GitHub (click to check it) which he is planning on expanding to demo at the conference
By building control logic on top of LLM, Henry will show how to automate the generation of unit tests, significantly reducing manual effort and enabling developers to concentrate on innovation. Moreover, he will demonstrate the integration of human verification into the automated testing pipeline, ensuring the reliability of generated tests(so we are not yet solving the oracle problem 😉