I suspect the community feels the same. There's always something beneficial that comes out of these for me. Whether it's seeing how *not to do something or picking up a little piece of code that can be repurposed somewhere else.
No-one is seriously suggesting we write CSS-only games. You're rather missing the point.
It's one thing to write something that uses 30 lines of JS code only, than to create this and say "Look Ma, no JS!!"
Nevertheless, demos like this are pretty cool.
We're hackers. We're supposed to like things such as obfuscated c competitions and 5k demos!
People going against perceived wisdom and doing the wrong things with the wrong tools is a great way to learn and innovate, I know CSS very well and I am fairly sure I wouldnt have been able to come up with this, it may not be the next big game engine but there definetely scope for very practical new techniques and insights from abusing tools in this way
If you lose that sense of playfulness with tech you might as well go work in some other field, because tech is about using things in ways they were never designed to be used... because it's fun first, and later because your under a weird constraint and you have this corpus of seemingly useless (at the time of acquisition) knowledge to draw on.
Things like this only promote a deeper understanding of what can be done in pure CSS. Yes he took it too far, but that's the point.
Now that I understand the author's intent, it's furthermore impossible for me to go along with a narrative that every experiment has equal or potential to lead to some mind-rattling insight. If anything, the posts about how someday shaders might build entire games etc just demonstrate how, behind the wall of basic comprehension of how GLSL's inner workings and things like Chomsky grammar concepts create fundamental limitations, the barriers one will encounter are both insurmountable and undetectable. See http://stackoverflow.com/questions/1732348/regex-match-open-...
CSS itself has annoying limits that can be summed up with the existence of tools like LESS and Stylus that add expressions to generated CSS, giving us an infinitely more capable styling system, albeit still suffering from annoying limitations that keep me sour to many CSS uses. Quite simply, everything in CSS transitions has an API so far from how animation is dealt with programmatically that I find only the most basic utilizations to -ever- be appropriate. This doesn't even mention the fact of shims, fallbacks, and the JS that deals with them. The idea that we're "pushing limits", much less making progress with examples like this is not one I'm buying into.
Hacking things to make them do something they were not designed to do is the life and blood of all things computers. It's just so damn fun.