Sorry, I'm not following. I'm saying the article says less than 30% of repos scanned used Unittest. That was in response to you seeming to say that most projects would choose Unittest.

What've I missed?

Oh, I see, you misunderstood me. I did not claim that “most projects would choose Unittest”. I claimed that for many projects, it would be wise to consider using unittest instead of, for example, pytest, as a way of minimizing usage of third-party modules.

> Every individual third-party library will have to overcome that threshold by being useful enough. And I doubt that, for most people, pytest is that much better to work with than the built-in unittest is.

This sounds more like you're saying that pytest wouldn't overcome that threshold; I was saying that has already happened, at least according to the article's statement.

Maybe. But I would personally guess that most people nowadays are way too eager to use a new library than would be good for their projects in the long run. I very much doubt that most new libraries are that much better than the Python standard library counterparts that it’s worth using them in non-toy projects. Of course, all progress depends on the unreasonable man, etc.