It's certainly very cool, but is it really that useful? Or I could just be missing the bigger picture. I actually hope that's the case so I can learn something new today. But it just seems to me like making a png/gif image is much easier and more practical. I do love to see technology pushed to its limits though!

I guess one of the cooler applications of this shape making method would be the ability to transform one shape into another very smoothly with a relatively small filesize (as opposed to an animated gif).

There are more benefits:

Costs nothing to change colors. This makes it great for iterating.

No bandwidth costs as CSS is rendered on the client.

Infinitely scalable in size. No rasterization.

Availability of other CSS functionality, like shadows, animation etc

The last one is an issue, actually. The space invader one is the extreme case: it's already a shadow. And the border tricks for triangles are also shadow-resistant.