http://www.youtube.com/watch?v=UKUuUvDSXk4&feature=player_embeddedWasn't sure if this was a GD topic, but since it's all about game engines....
I don't claim to know enough about graphics to completely understand what he's talking about. But I do understand the nature of polygons in graphics, and his claim is essentially this: rather than constructing shapes in an engine out of triangles to create the final form, the basic unit of Euclidean's engine is an atom, a single unit of tiny size. When they pack 500,000 or more of these atoms into a single object, you get a realistic looking shape that doesn't have angular, jagged edges. Their current island that's a kilometer in size has over 8? trillion atoms in it. The thing converts straight from Maya and 3DS so modelers using it basically have no poly count restrictions anymore. (A full engine would mean an entire level could be built with no poly count restriction...a tasty prospect.)
To give you an idea of what this actually means...the ground is composed of several million atoms, so direct when you get close to it is actually a bunch of globes resting on top of each other. They're also doing object scanning and direct rendering, which isn't new. But because they're not restricted by poly count, the item is almost fully reproduced, whereas with polygons the poly counts were so high they couldn't realistically be used.
I have my doubts because someone still has to make this usable from a developer's end, even with their SDK. And there's all sorts of considerations for collision, path finding, physics, ect....But if this can actually run in a game engine at good FPS.....it could revolutionize 3d graphics engines.