Of course it can. But look at the existing examples - usually they explicitly or implicitly operationalizing predictions as predictions about certain human activities. For example, predictions about the Millennium problems aren't about the truth or not, but whether the prize is awarded.
That's close to accurate but note that some have both a truth claim and an operational claim. See e.g. http://predictionbook.com/predictions/2838 which predicts that P != NP and predicts that it will be proven by a certain time.
In general your point about operational predictions is a good one, since predictions should be actually falsifiable.
I was actually thinking of your predictions when I said 'implicit'. How would that prediction be judged? Obviously by whether the community of mathematicians/complexity theorists like Scott Aaronson or the Clay Institute says that it was proven or not proven.
There are a variety of cosmology and science related predictions about things that have not been discovered yet. There are for example predictions about when we will discover extraterrestrial life as well as other predictions about observed events. See e.g. http://predictionbook.com/predictions/3186 http://predictionbook.com/predictions/2803 .
I like to make predictions about things like quantum mechanics and decision theory and cosmology and fun stuff like that. Can PredictionBook be used for this? If so, what should the guidelines be? Figured I'd make this a Discussion post so as to encourage the habit of betting on logical uncertainty.