I guess for some or generative structures it makes sense, but programming a CAD object sounds like something from the 1970's. Being able to click on a corner and say, "fillet, radius whatever" makes way more sense to me. I must not be "getting it" - could you explain how it is better? I get that it's FOSS, so besides that?
The difference is programmatic cad can have composable, repeatable, and automatable operations.
What if you have a thousand corners to fillet, in some kind of gradient, so each is a little different than the last? You won't be clicking and menu-ing your way out of that.
What if you want to do transforms on models as part of a process, like read a pile of models and do something to each? Programmatic. And then you can pass that process to another transform doing something else.
etc.
In a programmatic cad system, it's just a few lines of code.
Has an effort to add a full GUI to OpenSCAD ever been considered? Would be absolutely amazing to be able to design the way you're describing, and then throw it into a git repo.
