Ok, a minimal OpenGL GUI to start, for devices that don't have an LCL, but do have a GPU, sounds low-entry. Very doable.
But, we could go much farther in that. Because of native controls requiring both a language interface for the events (hard problem in many cases), and support of the widget sets (lots of work). And the platform independent interfaces being pixel-based.
OpenGL is a standard. Yes, the implementation can be lacking, but with the shader model, we don't really need much more than surfaces (triangles and rectangles), textures and shaders. Most everything else is irrelevant.
Ok, a fast geometry shader would help immensely, for things like drawing controls and (especially) fonts. Or calculating the initial signed distance font texture. But those are all things that can be calculated up front.
As we are all programmers and not artists, making things slick looking is hard work. Nice to have and to look at, but that can be done at the end, to make it better.
If we start with the minimal requirements to make it work, it is very doable.
But, yest, that will be most of the work. I could do a distributed computing library in a few weeks time by myself, comparatively.