> I love Feynman's idea of minimal computational energy - it suggests that we could actually measure elegance.

It's a lovely idea, but in practice would be hard. "I just updated my interpreter and my code got 20% more efficient!" "How does it compare to that other code?" "I don't know, it's incompatible with the new interpreter."