Part of that is improvements in HPC. You can now do computational analysis of these designs and evolve a good solution way faster than was done by hand in wind

Part of that is improvements in HPC. You can now do computational analysis of these designs and evolve a good solution way faster than was done by hand in wind tunnels for the Concorde and other supersonic aircraft. I don't think this would be possible without today's computational technologies, which allow orders of magnitude more "experimentation" (even if virtual) and therefore iteration in design. Aside: personally I'm waiting for someone to take up the Avrocar, aka the original flying saucer UFO. It's basically the jet version of the helicopter or quadricopter drone from an aero engineering tradeoff standpoint. At one point the Army thought that they'd stop buying jeeps and start buying flying saucers instead, but the Avrocar project fell apart for dynamical stability reasons. It was just too difficult to keep it in level flight during flying regime transitions, and they weren't able to design a control interface a human could use to compensate. But that was with 50's tech, and I'm pretty sure with modern ML techniques you could design a fly-by-wire feedback response control system that would have no trouble stabilizing Avrocar or an improved version of it, just like Stanford got a helicopter to fly upside down (a very analogous stability regime actually). It would fill the niche for medium-distance drone transport or perhaps short-range personnel transport (like choppers).