Andrew Gallagher critiques the use of large language models (LLMs) in generating unit tests for software. He argues that this approach leads to poor-quality tests that often verify what the code does rather than what it should do. Gallagher emphasizes the common workflow among developers, where they rely on AI to produce code, tests, and documentation in a streamlined manner. He points out that while this process can be efficient, it often results in a set of brittle unit tests that don't provide meaningful validation. Gallagher illustrates his point with a React button component example, showing how an LLM might generate a series of tests. The tests may check for default properties and ensure the button renders correctly but miss essential checks for edge cases or logical errors. He highlights two main issues with the generated tests: the excessive number of tests and their focus on confirming the existing behavior instead of ensuring the correct functionality. This misalignment can lead to a false sense of security regarding the quality of the code being tested. The article underscores the importance of thoughtful test design. Gallagher advocates for developers to take a more hands-on approach rather than relying on AI-generated tests. He suggests that meaningful unit tests should not only verify that the code works as intended but also ensure it behaves correctly under various conditions. By emphasizing the need for developers to engage more critically with their testing practices, Gallagher aims to steer the conversation towards better quality assurance in software development.