He's most likely referring to having uber sampling on.
But this is an issue related to what the OP talks about, how critical is actually achieving those settings in terms of enjoying the game?
What would be interesting is a blind test is done with a sample of gamers (perhaps with different backgrounds) and seeing what the actual threshold is before individuals can actually reliably determine which is providing a better experience.
Compare a GTX 980 and GTX 960 for instance using GFE recommended settings at 1080p across a sample of games (or a r9 290x vs r9 270x). How many people can actually reliably tell the difference by playing them one after another? Not in terms of side by side and comparing details or combing through static screenshot comparisons but actually playing regularly? No seeing the settings menu or using any sort of fps counters either, just impression purely from regular playing and not specifically looking for differences.
It would be interesting to do this type of test with each of the following conditions -
1) Trying to keep performance (frame rate, frame times, min fps) constant
2) Trying to keep graphics settings constant
3) Adjusting both aspects
However I'm not sure how much interest a hardware site would have in actually investigating this due to the complexity of doing so and it conflicts with their own interests as well, since they do need to interest readers in reading about new hardware.
Sure there are likely some outlier individuals with extreme sensitivity and may notice even minute changes (say the difference between a 290x and 290) but I have a feeling that even the majority of self proclaimed "enthusiasts" will not. It's basically like the whole "golden ear" audiophile situation, most people really are not as special as they think they are.
Granted at the moment there is somewhat transitional period currently with VRAM requirements which does make the current gap between high end and mid range cards more noticeable.
Um yes so am I. When I say maxed out besides AA I mean it.