Ask HN: Isn't NumPy/SciPy used for serious academic/enterprise applications?

I was surprised to read someone qualified to speak on the topic suggests otherwise: In Data Analysis with Open Source Tools, P. K. Janert surveys several languages/tools, including R, Octave, and Python with NumPy/Scipy. Regarding the last, he writes: "[...] NumPy/SciPy has its own share of problems. The project has a tendency to emphasize quantity over quality: the number of features is very large, but the design appears overly complicated and is often awkward to use. Edge and error cases are not always handled properly. On the scientific level, NumPy/SciPy feels amateurish. The choice of algorithms appears to reflect some well-known textbooks more than deep, practical knowledge arising from real experience. "What worries me most is that the project does not seem to be managed very well: although it has been around for nearly 10 years and has a large and active user base, it has apparently not been able to achieve and maintain a consistent level of reliability and maturity throughout. Features seem to be added haphazardly, without any long-term vision or discernible direction. Despite occasional efforts in this regard, the documentation remains patchy. "[...] [B]ecause of the overall quality issues, I would not want to rely on it for “serious” production work at this point." The book reviews I've read have generally been positive so I'm not sure if the SciPy community has any thoughts on this. Regarding the implementation of mostly "textbook" (standard) algorithms, I can see that this would be the case since there are large number of practitioners using and extending SciPy (in contrast, for instance in R there is a strong precedent for statistician to implement their newest algorithm in that language).