Interesting topic and discussion!

When I started to shift from 2D to 3D work around 1998, my tool of choice was the mainly polygon-based 3ds Max (it has some NURBS tools, but they're limited). I clearly remember how I longed for a modeling method that would be able to combine primitive shapes and seamlessly blend them together. The problem with subdivision modeling for me personally is that you're carefully planning and constructing in stead of sculpting. Constructing doesn't allow for spontaneous creation, with "happy accidents" as the late Bob Ross used to call them.

In those days, the late 1990s, there was a cool plug-in for 3ds Max called Clay Studio Pro. It was a metaball system, but you had many different shapes to work with, including 'Clay Splines', which allowed for a lot of modeling freedom, and everything was animatable as well. Clay Studio Pro was ahead of its time, kind of a very basic 3D Coat avant la lettre. But the problem was that PCs didn't have enough processing power yet to turn Clay Studio Pro into an all-round solution. To achieve smooth meta-surfaces you had to increase the polygons higher than my system could handle.

When I discovered MoI many years later, it was the first program I encountered that combined the accuracy of constructing with the spontaneity of fast and easy creation. And for organic modeling, ZBrush and — for me only recently — 3D Coat turned out to be fabulous tools.

The bottom line is that I really love not having to worry about polygon structure planning, keeping everything quadrangular, and having to keep an eye on edge flow everywhere. So my best friends are MoI, 3D Coat and ZBrush.

Having said this, I do love the polygon-based Blender, but mainly because it fills every gap that's left by MoI, 3D Coat and ZBrush, such as animation and high-level rendering (although I also use Keyshot for rendering).

In the field of sculpting freedom, the only area that's still a bit on the weak side currently is automatic retopology. I dislike processes that require tedious manual work, such as manual retopology, manual UV mapping and manual rigging and weighting. I hope AI / machine learning will improve those areas soon.

So, the bottom line is that I think subdivision is not yet outdated, but it is becoming slightly dated as powerful alternative modeling methods are on the rise.

All the best,