It is an interesting approach, but i do wonder at its ability to be effective. From what they say in the press release/video it would seem to me that it would be possible to still figure out what each piece is doing through multiple attacks. Even if each ID changes randomly its position/what it is calling to is not. So you would need to program your bot to "find" what that ID is interacting with against a known type. in other words to me this falls under "obfuscation" while nice, once it is figured out exactly how it is done would seem to be rather easy to reverse, at least with the limited information released on it. Basically i would think of this as a "start" but not a finished product yet.
then again i could be completely off, programming is not my strong suit. Still as with any security product i wouldnt trust it without getting to know every thing about it.