The thing that stood out for me, and made me think "nice", was Brandon's idea of creating a vector field type. One can then apply operators directly to the field, without having to think in terms of lower level constructs. In that sense, I think the '9-lines' claim is fair.
Any comments from anyone (Brandon?), as to whether configuration and boundary conditions can be simply applied to the vector field type?
The technique here which I guess is effectively finite element was far more difficult to implement at the time and reserved for professional tools like Fluent (my company today uses Fluent for this stuff today though not me directly and I'm sure we pay quite a bit for it.)
I browsed through the code and it was far better than anything I ever wrote at university. I think its deceptive to say 9 lines of code given the volume of support libraries but I would have killed to get my hands on this 17 years ago. Boundary conditions and configuring the actual volume was actually the hard part and took a majority of my time back then and I'm guessing that probably hasn't changed. Anyway cool stuff if only a few years late for me to make use of.