[chriswarbo]

This problem is solved by choosing a standard language, which requires some way to tell whether one language is "less silly" than another. There's no 'correct' way to judge this, but a reasonable, objective benchmark is how many bits it takes to write a self-interpreter (in a similar spirit to "compiler optimizations should pay for themselves" http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.39.... )

Binary Lambda Calculus is a pretty good for this. I personally prefer Binary Combinatory Logic, but its self-interpreter is a bit longer.

There's probably a way to game this benchmark with a "silly" language, but I imagine anything more complicated than hard-coded literals would end up larger than BLC.