Well, I as a child (say: 6-8 old) was really rather thinking deeply how a suitable way (what programmers would call "data structures" and "programming abstractions", "design of a programming languages" - but of course on a child's level) might look like so that the computer can "understand" and precisely execute my game ideas; my games as executable files would rather be the central side product of this.
I also (I am really not lying or boasting!) was thinking as a child (just to be clear: this was my child's imagination; from my present knowledge I know of no suitable way to make this actually work) how if the abstractions are there, one could easily "combine" existing games to make new games - perhaps even partly automatically by a computer:
Imagine this: using some photo editor, you can use the magic wand or lasso tool to select some part of a photo, and copy-paste this part into another photo. Why isn't it possible, if you, say, wrote both a space shooter game and an economic simulation game to select and copy-paste some part of the latter into a former to turn the space shooter into a space shooter that also contains economic simulation aspects?
The "best actually existing" (really bad) approximation of this that I have seen in life is how the GURPS tabletop role-playing game system (https://en.wikipedia.org/wiki/GURPS) enables (in principle) to "copy-paste" elements from one role-playing setting into a whole different "incompatible" one (but it is not that you would want to do this for your typical role-playing session :-) ).
Yes, this is how a 6 to 8 years old child (the former me) thinks about game development. :-)
