>> In reality, both dimensions are orthogonal, and vary in a continuum. Hence, you'll find super-coder assholes, nice incompetent people, and everything in between.

That is definitely true, but you don't get a uniform statistical sample responding to job postings.

The super-coders who work well with people don't come on the market as often.

  >> the author oversimplified reality to come to her conclusion.

The author's conclusion is that you should ignore one axis (I don't agree).

I don't think the point was that you HAVE to choose between coding skills and social skills. The point was when the two come into conflict you should choose coding skills. They are making the point that coding skills are more important and should weigh more in the equation. The other option is to just keep looking but there is a cost associated with that as well. It's a generalization of a very complex HR calculus. At least that's how I took it.