The intent of the MingwPy developers is to expand this website and the MingwPy project (code to build and distribute the compiler toolchain), and start using it in the first half of 2016 as the default method to distribute Windows wheel binaries for Numpy, Scipy and other scientific Python projects.
For more details, please see this proposal (which was accepted by both NumFOCUS and the PSF in early Jan 2016):
Progress on milestones in the proposal
Phase 1 of the proposal has been finished. The compiler toolchain has now been re-built by several people based on the public build scripts and documentation.
Phase 2 is about halfway to being finished:
One item from phase 3 is almost finished already: MingwPy itself can now be installed as wheels for Python 2.7, 3.3 and 3.4 from an Anaconda channel . Once those wheels have proven themselves in daily use for a while they will be uploaded on PyPI as well.
Other developments and observations
There is also interest in MingwPy from outside the Scientific Python community. Tony Kelman, who is maintainer of the Windows builds for Julia, has been active in providing feedback and new ideas. Julia uses MinGW-w64 for all its Windows builds, and it looks like there’s quite some synergy here (also on OpenBLAS builds and improvements for example). Other examples reported on the mailing list or GitHub were use of MingwPy for building Kivy , for a Blender plugin written with C99 variadic macros (which MSVC doesn’t support), and for building NuPIC (machine learning platform from Numenta).
Some topics that people have shown interest in and are being discussed on the MingwPy issue tracker or mailing list, but are outside the scope of the currently funded proposal are:
The received funding from the PSF and NumFOCUS and the progress that MingwPy has shown, seems to have also encouraged related activities. A good example is PEP 513  and Manylinux , which has enabled Linux wheels on PyPI - this effort was driven by people from the scientific Python community and several MingwPy devs made major contributions here. Another important step forwards is that Windows wheels for Numpy  are now on PyPI. Those are a stopgap solution based on MSVC until Numpy/Scipy wheels built with MingwPy will be pushed to PyPI, but have already proven useful to users.
|||(1, 2) https://pypi.anaconda.org/carlkl/simple/mingwpy/|