1. Clean the floor!
2. Drop a big cube (for a 1:1 model, about 230m high)
4. Force the blue cube into an angle of the big one
5. Delete the blue cube and the intersected part
6. Turn around, and put the red cube at an opposite angle
7. Remove the red cube and the intersected part
8. Oops, a bit of red remaining inside. Almost done!
9. Better now. See yourself through?
Your CCTV tower is ready! Well, the real one is a bit more stylish, cubes are more distorted, but the idea is the same. If you try to describe it as an addition of things (a L-shaped base, two legs, and an a L-shaped overhang but reversed...), that's difficult. But if you see it as a subtraction, it's pretty simple: remove two basic volumes from a third one, and the building is what remains.
I suppose the stunning elegance of the CCTV tower is partly due to this great use of negative space. What is NOT there gives more strength to what is there. I already noticed that a logo could be greatly enhanced by adding empty areas and that music was more beautiful with silences, but I never really realised this could work in architecture too.