I agree that you shouldn't spend a lot of effort going beyond the requirements, but sometimes you will reap huge benefits because you happened to go somewhat beyond the requirements in a given instance. It's like the property of brittleness - you don't want to spend too much time making everything adjustable and bendable, but neither do you want to make everything as hard and precise and brittle as possible.
Someone linked an article elsewhere on HN about the true meaning of mediocrity that I enjoyed and really made me think about this:
A real life example - client discovers that users cannot actually get into the system. Contractor says: "Ah, you wanted a login screen? This wasn't in the specs! You need to pay me more to implement a login screen."
>13. Design is based on requirements. There's no justification for designing something one bit "better" than the requirements dictate.
This is YAGNI stated a different way.