So, smart includes some kind of experience-derived common sense that isn't measured well by IQ tests...? Perhaps someone should tell the people who want to use IQ tests to measure the quality of programmers...

Paul Graham uses the word "smart," essentially, to mean someone who "works smart" rather than simply "working hard." In other words, it's someone who won't build a 12-foot fence if a 6-foot fence is sufficient. And it's someone who's capable of realizing as much in the first place.

This isn't "common sense," per se. It's more like uncommon sense. Most people follow an effort in -> result out paradigm, believing the two are perfectly correlated. A "smart" person, in PG's view, attempts to min effort and max result. (I've also heard this sort of hacker described as "lazy," in a positive and non-derogatory sense. A "lazy" person finds the most efficient and effective ways to do things, minimizing man-hours and resources.)

The relationship between "smart"/"lazy" and IQ hasn't been studied all that well, but it seems plausible that there is some sort of relationship. Perhaps it's a strong one, and perhaps it's not. Either way, I don't believe an IQ test as a single-factor qualification gate will filter effectively for "smart"/"lazy."

I use the word efficient instead of lazy. If I can do x and 1 other thing if I do x efficiently, why not strive to find as many things I can do efficiently as possible?

In this case, the word "lazy" is being used as a sort of tongue-in-cheek reference for "efficient." I don't think it's meant to be interpreted as being literally lazy.