Imagine it is the spring of 2019. A bottom-feeding website, perhaps tied to Russia, surfaces" video of a sex scene starring an 18-year-old Kirsten Gillibrand. It is soon debunked as a fake, the product of a user-friendly video application that employs generative adversarial network technology to convincingly swap out one face for another. Then it is fall. The junior senator from New York State announces her campaign for the presidency. At a diner in New Hampshire, one "low information" voter asks another: "Kirsten What's-her-name? She's running for president? Didn't she have something to do with pornography?" Welcome to the shape of things to come. The technology [is] closer than you might think. And even when fake video isn't perfect, it can convince people who want to be convinced, especially when it reinforces ... stereotypes. It might be impossible to stop the advance of this kind of technology. But the relevant algorithms here aren't only the ones that run on computer hardware. They are also the ones that undergird our too easily hacked media system, where garbage acquires the perfumed scent of legitimacy with all too much ease. It already feels as though we are living in an alternative science-fiction universe where no one agrees on what it true. Just think how much worse it will be when fake news becomes fake video. Democracy assumes that its citizens share the same reality. We're about to find out whether democracy can be preserved when this assumption no longer holds.