Okay, here's the low-down on modern stealth aircraft development (I've got a Masters in it). The Russian you refer to, Ufemstef (forgive the spelling...not near my reference material), didn't invent stealth, or even the "math" of stealth, not directly. What he came up with is a simple mathematical forumula for predicting the electromagnetic scattering signature of a triangular plate. Up to that time, the only shapes we could accurately and efficiently predict the radar signatures of were a sphere, a square plate (though with some limitations), and a cylinder of infinite length. It's very difficult to produce an aircraft that will fly using only these few, simple primatives. So, while we understood the principles of radar scattering, the computers of the day were not up to doing iterative predictive necessary to design a flying combat aircraft. This is all a gross simplification of the extremely complex issues surrounding the problems. So, even though some aircraft had somewhat stealthy features, this was more often a case of getting lucky. The SR-71 Blackbird is one example where the unique aerodynamics required actually helped, rather than hurt, its stealthiness...in most cases, it does just the opposite.
Now, along comes a young engineer at Lockheed's famed Skunkworks. He is purusing foreign recently translated scientific journals when he comes across a fairly inoccuous artical by an equally inoccuous Russian mathematician in an unclassified Soviet journal. In an epiphany that must have sent Stalin rolling over in his grave, he realizes he has the answer to a question that has been plaguing aerospace engineers since radar was discovered. You see, you can build a plane out of triangular facets, and still make it fly. Because of the simple genious of Ufemstef's equations, it is now possible for the computers of the day to design an incredibly stealthy airframe in a virtual environment. Now, instead of taking months for a supercomputer to arrive at a dubious signature prediction -- for every design concept -- it is possible to test hundreds of different plan-forms in the course of weeks, and at multiple radar frequency bands.
Combining the new radar cross-section (RCS) prediction algorithms with American know-how and ingenuity, they quickly develop the Pave Blue prototype. This forms the basic blueprint of the F-117 Nighthawk.
The B-2 is also owes its design to Ufemstef's equation. Both the 117 and the B-2 were predicted using mathematical models made up of triangular facets. The difference is, when the F-117 was designed, computers limited the number of facets that could be reasonabley used in a radar cross-section prediction to around 200. That's why the F-117 is so blocky looking. However, by the time the B-2 was under full-scale develpment, computers and computer programing had pushed the number of triangles/model up into the millions. Makes the triangles small enough, and it begins to approximate a smooth surface. So, the B-2 was made with smooth curves because it could be "predicted" with enough tiny triangles to adequately approximate that smooth surface.
In an asside, I actually met Dr. Ufemstef. He was brough in by the university to lecture us. He was a charming and humble man, with a great sense of humor.