There is some merit to that approach. Yes TML handles rules and logic. The current TML does not handle knowledge at this time but a knowledge representation language would allow for it. The rules of how a Rubik's cube operates, and the logic, can be encoded into TML, and TML can solve it according to it's rules.

But as of right now I don't know if TML is at that stage where it can do something like that, right now the language is very rough.