Angry Tests

The first volume concentrated on unit tests, because that is where most programmers spend most of their testing effort. This second volume widens the scope. Unit tests, however well written, cannot catch every class of defect, and relying on them exclusively tends to produce a suite that is both expensive to maintain and quietly incomplete. The book examines the other kinds of tests that earn their place in a serious project — integration, end-to-end, property-based, contract, performance, and several less familiar varieties — and discusses when each is worth writing and when it is not.A central theme of this volume is what we call Headless Development: an approach in which the test suite, rather than the running application, becomes the primary feedback loop during development. The user interface, in this view, is something you visit occasionally to confirm what the tests have already told you. We describe how this shift changes the way features are designed, how defects are reproduced, and how confidently code can be changed.The book also gives substantial attention to AI-driven testing. Large language models have begun to change what is practical in this area, both in generating tests and in maintaining them as the code evolves. We describe the techniques that have proven genuinely useful in our work, and are equally candid about the ones that have not.As with the first volume, the focus is on practices that hold up in production over years, not on ideas that sound good in a conference talk. If the first book left you wanting more aggressive tests across more of your system, this one is a continuation of that conversation. Read more