When I first started programming and the code did not behave as I expected, I struggled to understand why, leading to a lot of frustration. It was only after I discovered unit testing that things began to click. Testing not only helped me validate my work but also taught me how to write clearer, more structured code.

However, as I gained more experience, I encountered projects where testing became a source of frustration itself. Overly complex unit and integration tests made it difficult to understand what was actually being tested. When these tests failed, it was sometimes hard to tell whether the problem was in the code, the test, or the setup around it. The confidence that testing had once given me started to disappear.

In this talk, I will share what these very different experiences taught me about making tests genuinely useful during development. Using Java, JUnit and Mockito, I will explore testable code, mocking, parameterized tests and what to question when a test setup becomes harder to understand than the code it is supposed to test. I will also share some of my more recent experiences using artificial intelligence at work, including what happens when generated code comes with passing tests but still misunderstands the requirement.

Testing can do much more than tell you whether your code works. I want to show how it can help you understand the code you are writing, shape its structure and give you confidence while changing it. The interesting part is what happens when tests stop doing that.

Learning about unit testing was one of the most valuable things that happened to me while I was learning to program. It gave me a way to understand what my code was doing, helped me structure it better, and made programming feel much less frustrating.

Later, I also experienced the opposite side of testing, where complex setups made tests difficult to understand and trust. In this talk, I share how those different experiences changed the way I think about testing and why I now see unit tests as much more than a way to check whether code works. They can also be a tool for understanding, designing, and changing code with more confidence.