Hamel Husain argues that testing LLMs like traditional software is a mistake. Start with data analysis to understand what the model actually does before you write a single test, because LLMs behave too unpredictably for straightforward test cases.
It's really important that we don't think of evals as just tests. It's a common trap that a lot of people fall into because they jump straight to the test, like, let me write some tests. And usually that's not what you want to do. You should start with some kind of data analysis to ground what you should even test. And that's a little bit different than software engineering where you have a lot more expectations of how the system is going to work. With LLMs, it's a lot more surface area. It's very stochastic.
