I find that I don't agree with this comment, though perhaps if I thought about it more I would..

I often categorize people as 10-points-smarter-than-me, 20-points-smarter-than-me, etc, just naturally as I go about my day, and I'm (currently) fairly confident of my evaluations.

Sometimes I can get a pretty good estimate by speaking with someone for 5 minutes -- but I'm aware this is heavily weighted towards verbal acuity, which is just one dimension. A high verbal acuity for me is a marker of high IQ, though average verbal acuity is not strong evidence either way. (I also understand someone can have very high verbal skills while missing some others, so there is an upper bound to what I can predict.) I'm recently studying signs of high social acuity, and I think I'm getting more perceptive at noticing and distinguishing levels that are one or two levels above mine.

Someone that is relatively deficient in one of these parameters can recognize higher levels by evaluating in hind-site how effective and original a particular solution, choice of words or behavior was. It's much easier to judge a behavior than come up with the best behavior yourself. For example, during a meeting I'll realize that someone is manipulating me or others very well. It's much easier to recognize that manipulation that to recreate it. Though it does take some compensatory experience -- since recognizing these manipulations can be as subtle as realizing that a awkward situation has been avoided or people feel better about something that you would have predicted, and determining the intention of the manipulation, and whether it was deliberate or accidental, makes the computation more noisy.

The main way that I judge an IQ higher than mine is if they are faster or more clear on something I've already thought about (this translates to slight increases in IQ) or can succeed at things that I cannot do well or cannot do at all (this is where someone would be at least a level ahead). For example, I'm fairly good at solving problems, and working within a given frame, but I am not very good at choosing problems, because I'm not very strong in selecting frames -- this is a higher order cognitive skill I cannot do well, and no amount of time will increase my success very much. Thus I seek out mentors that are "10 or 20 points above me" to help me with that bit... I consider myself to be 'borrowing' their IQ points for a very short amount of time, so that I can then go and work on a problem within the way they've chosen to frame it. (Again, while I'm not as good at picking a frame, I feel like I am competent at evaluating whether my problem can be well solved within one.)

What I find curious is how people successfully model people of higher IQ in fiction. For example, the doctor in House seems very witty. Is he modeled by someone at least that witty? Likewise with Sherlock Holmes. I realize that the scenarios are contrived, so that the intelligence is mostly illusory, but how intelligent must one be to write a story that convinces someone, say, more intelligent than themselves watching a film that the film is about someone vastly more intelligent than them?