Think figurines (Blender) vs gears (CAD).
Constraints, among many other important features, just aren't as well represented in Blender.
An analogy is C vs JavaScript. Can you do "memory management" in JavaScript? Sure, but you're fighting the tool. Ditto for building a complex frontend in C.
The desire to "just learn one thing" is naturally strong. But the "design 3d things" problem space is as large (if not larger) than "programming computers". Hence the proliferation of tools with very different approaches (the underlying representation in CAD is generally brep [1], which is much different than vertices / edges / faces at the core of Blender)
The good news is the underlying thinking is somewhat transferrable, especially for core concepts.