Anyways, I don't critique their experiment but their intention. If you experiment to make it 'prettier' then it's a fail. Experiment to make it more functional.
Your point "functional is pretty" - is very true. Nothing better than the thing that works.
Judging intention is probably the least useful form of argument.
Why should it be a fail to experiment with making it prettier? What exactly is wrong with that?
You are falling victim to your own critique. You are judging based on your subjective opinion about some ones intentions. Which is fine. Just don't claim it to be somehow a more rational argument.
This is easier with some kind of sites than with others. If it's a company site trying to sell a product or a service, then obviously the intention is to... sell you the product or service. If the "design" doesn't push that intention or gets in the way of you learning about the product and why you want to buy it, then they've failed.
With personal sites and blogs it's harder to judge, and I'd agree with you. But I think in those places design (at least in the sense that the author is talking about) is less important. Someone's personal site might in part be a sandbox or playground to try out new HTML5 or CSS techniques.
Everyone's opinion is subjective. Claiming that that somehow invalidates the argument is disingenuous. Like everything, it's a continuum -- some things are easier to "objectively" judge, others are not and are more a matter of taste and experimentation.
Such a simplistic point of view.
Designers take inspiration from other works. A beautiful piece of work doesn't have to be functional if it's largely just used to express ones creative outputs or to inspire others.
Concept cars and fashion shows often do the same thing: help inspire others and influence upcoming styles.
Saying that their experimental efforts are failures because it's not completely functional (a concept car that can't function, a wardrobe that few souls would wear, a website that has out of the ordinary behavior) is completely missing the point.
Pretty might not work if put in actual use, because it is no longer an experiment.