A long-standing idea in the Western philosophical tradition is that our knowledge of the past and present is more secure than our knowledge of the future. Given standard assumptions about the relationship between knowledge and assertion, this would suggest that it is, in general, easier to make assertions about the past and the present than it is to make assertions about the future. I present some cases that seem to suggest just the opposite. In these cases, one is in a position to assert, at time T1, that something will happen, but one is not in a position to assert, at later time T2, that the event in question didhappen. I consider two possible explanations of this phenomenon. The first is a semantic account that treats future operators as epistemic modals, while the second is a pragmatic/epistemic account that implies that one can lose knowledge simply by moving through time. I argue for the latter, and then consider what it might tell us about the nature of knowledge.