I've always felt that writing code was easy, and I still do. It's the other stuff -- figuring out what code to write -- that's hard.

This hits the nail squarely on the head and drives it a good way into the wood.

In a perfect world I'd be able to spend my time designing and coding based on a perfect understanding of a perfectly defined, immutable, project spec. But if the world were perfect it would be no fun.

Maybe you will get lucky and have reasonable requirements and a spec. But they won't be complete. You're still going to have to gather/clarify requirements with the person who produced the docs.

Maybe the big, `strategic' design has been taken out of your hands; you're still going to have to deal with the tactical design issues of functional composition, naming of parts, algorithm choice and all the other details that help you build a good program instead of a ball of mud.

Of course, you may want to be micro managed, but if that's happening wouldn't it be much quicker for your puppetmaster to just fire you and do the job himself?

You don't have to excel at all these things. You don't have to do them all all the time. But if you intend to get better at your craft you're going to end up either knowing how to do them, or you'll have some pretty strong ideas on the subject that you could put into practice if you had to. It's all part of the process of active learning that (I claim) is one of the fundamental skills/goals of any half way decent programmer.