About 250 million years from now, living on the coast could feel like being stuck inside a hot, wet plastic bag. And that bag would actually be the best home on the planet. Inland areas would be hotter than summer in the Gobi Desert, and up to four times as dry. This is life on Pangea Ultima, the supercontinent that an international group of scientists has predicted will form on Earth in a quarter of a billion years.

“It wouldn’t be a fun place to live,” Alexander Farnsworth, a climatologist at the University of Bristol, told me. Farnsworth is the lead author on a new paper published today in Nature Geoscience detailing how a supercomputer model predicted what Earth would be like in the far-distant future. According to his team’s calculations, 250 million years from now, the continents will reunite and Earth will become unbearably hot, rendering much of the land uninhabitable and leading to mass land-mammal extinction. If the team is right, everything would be, as Farnsworth put it, “very bleak.”

The possibility of a future supercontinent isn’t the shocking part of the new study. Continents drift around the planet at about 0.6 inches a year, much slower than your fingernails grow, but on a long enough timescale, their subtle migration can dramatically alter the Earth’s appearance. “We know we’ve had several supercontinents in the past, so it makes perfect sense to say it’s not going to stop now,” Damian Nance, a geologist and supercontinent-formation expert at Ohio University who was not involved in the new research, told me. Pangea, the most recent one, has the widest name recognition, but geologists believe that several others have formed throughout Earth’s history. Roughly 1 billion years ago, the Amazon and the Baltics were neighbors on the supercontinent Rodinia. Several hundred million years before that, another tectonic hodgepodge called Nuna dominated the planet.

[From the archives: The growth of continents]

But geologists have long debated what the next supercontinent could actually look like. One theory, known as “Amasia,” is pretty much what it sounds like: The Americas will drift westward across the Pacific, smash into Asia, and take up residence near the North Pole. Another school of thought predicts that the Americas, Africa, and Eurasia would instead squeeze out the Atlantic Ocean and reunite along the equator. Pangea Ultima—first described in 2003 by the paleogeographer Christopher Scotese, another author on the new paper—would be the outcome of such a fusion.

In the new paper, Scotese, Farnsworth, and their colleagues attempt to describe life on Pangea Ultima. The supercontinent, they write, would be a victim of its own size: With the temperature-regulating benefits of oceans restricted to the shores, land temperatures would increase by a whopping 14 degrees Celsius. (To put this in perspective, the Paris Agreement aims to keep global temperatures from rising 1.5 degrees Celsius above preindustrial levels.) The continent’s interior would bake, becoming a desert shrubland dappled with long, barren stretches. Volcanoes and other geological mayhem would pump carbon dioxide—more than doubling our planet’s current levels—into the atmosphere. This could lead to short-term cooling, but ultimately, the authors write, it could warm the planet about 11 degrees Celsius above preindustrial levels. The sun would also be an issue: Using previous forecasts, the team predicted that it’d be 2.5 percent brighter in 250 million years, sending more heat down to an already sweltering Earth.

The model doesn’t account for every possible variable that would influence Pangea Ultima’s climate. Crucially, it ignores any additional warming that human beings might cause by emitting greenhouse gases. Elena Shevliakova, a climate modeler in NOAA’s Geophysical Fluid Dynamics Laboratory who was not involved in the research, pointed out that it also fails to factor in possible cooling factors such as ice sheets, lakes, and straits. “This is, in some ways, the worst-case scenario possible,” she told me.

But between the sun, the volcanoes, and the deserts, the model suggests that parts of present-day South America could reach upwards of 140 degrees Fahrenheit in the summer and cool to only 113 degrees in the winter. Such temperatures, sustained over millions of years, might threaten all life on Earth, the authors argue. They predict that as little as 8 percent of the planet’s land could remain habitable for mammals, if they survive that long.

[Read: A climate catastrophe paved the wave for the dinosaurs’ reign]

That, other researchers caution, is a huge if. Daniel Schrag, a geologist at Harvard, says that if Pangea Ultima were to form (which is far from certain, in his mind), assuming that mammals would still be around is a leap. After all, we mammals have been around for only about 175 million years so far. Besides, life—mammals included—has demonstrated its ability to evolve and adapt to new environments.

Making a claim about the state of the world this far into the future “seems reckless and speculative at best,” Schrag wrote in an email. But other experts told me that the paper might have some utility. Shevliakova said that long-term projections act as a kind of stress test for climate-projection tools; in this case, the team applied a UK Meteorological Office model, often used for near-term climate-change projections, to a very different time period and question. The fact that the model behaved as expected this far in the future “reflects the robustness of the methods and science being used to deal with present-day climate change,” Shevliakova said.

Nance, the Ohio University geologist, said that long-term predictions can also help fine-tune our forecasts for the next 50 to 100 years. “You can sort of step outside the box a bit and look at other processes besides fossil-fuel burning that might increase or decrease carbon dioxide in the atmosphere, and over what time frame these processes happen,” he said

Those uses hold true whether the far-future world turns out to be more or less hellish than predicted. They might, in fact, be the most important lesson to be taken from this paper, because we can’t know whether Farnsworth and his team got it right. As Shevliakova put it, in 250 million years, it’s not like you and I are going to be around to check.

Amazon is in a race among big tech firms to exploit the potential of artificial intelligence.