These discussions always surface something: The problem with the terminology. If you look at the arguments (including my own posts in this thread), people are

These discussions always surface something: The problem with the terminology. If you look at the arguments (including my own posts in this thread), people are basically talking about different kind of tests, with different costs and value propositions, and dumping them all under the same umbrella. The term "unit test" has been badly overloaded to mean a whole lot of stuff. Without first categorizing type of tests, and defining what we mean by unit test, we can't even have the discussion about if it's dead, useful, whatever. Even after defining that, there's a couple of sub groups that strictly define the value of testing based on specific criterias and ignoring others (eg: "the goal of tests is to know if you're breaking something when you refactor"). Tests have a LOT of benefits, and different type of tests get a different subset of these benefits, at varying cost. Without those definitions and associated tradeoffs, the discussion is basically a waste of time. You can see it in this thread: people are going "TDD is XYZ and the purpose is ABC". And you have about 6 variations of XYZ and ABC. Everyone is sure their version is the right one. Myself included.