I would have called bullshit on this if there wasn't that little demonstration in there, which I admit shocked me a bit. It works, though I still have three problems with it:
I can't help but wonder how much eyestrain this would cause on the long run since staring at one point, while obviously cuts down on the wasted time of eye-movement, would probably tire out one's eyes faster than skimming through the pages.
Secondly, I can pretty much already read about 500 words per minute in the "old fashioned way" if I strain myself a little, but at that point I already notice myself slipping and my reading comprehension getting worse and worse with every sentence. Our human brain, and our short-term memory in particular, is just not made for this kind of high-speed retention, and while I think slowly cranking up the speed with a program like this could help with conditioning to achieve better retention rates, I wouldn't hold my breath for the results.
Finally, and maybe most importantly, it think this technology would be the least useful at the most time-consuming texts, such as scientific/legal/technical literature and novels. In the case of the former, actually understanding the text is a lot more important than being able to power-read through it, while in the case of the latter being able to read on your own pace and taking your time to imagine the scenes and act them out in your head is pretty much the point of the entire reading experience, something you cannot do on a "250-1000 words per minute" rate.
That said, I do think this tech has merits and many people will probably find good use for it, but I would take some of its claims about reading comprehension and that "1000 words per minute" claim with a pinch of salt for now.