There's a place in Argentina that's just a circle of plants floating on a lake, and it's been rotating for as long as anyone's been watching.
Rotating how?
Nobody knows. That's the point. And that's roughly what Daniel's asking about today. He wants the history of Atlas Obscura, the site that catalogues that kind of thing. How it started, how it grew, who contributes, how entries get verified, and what its success says about the appeal of the obscure. He's got a nice line in there about how most travel media funnels millions of people to the same famous places while this thing celebrates the peculiar details that make the world interesting.
He's not wrong to frame it that way. That's almost exactly how the founders framed it themselves.
So let's start at the beginning. With a seventeenth-century Jesuit and a blog nobody was reading.
The Jesuit is Athanasius Kircher, and the blog is the actual seed of the whole company. Joshua Foer was running an anonymous blog called The Proceedings of the Athanasius Kircher Society, named for this polymath who wrote about everything from volcanoes to Egyptian hieroglyphs. It was a catalogue of curiosities, basically. And in 2007 Foer wanted to organize a real-world event for the readers, so he put out a call for help.
One volunteer.
Dylan Thuras.
That's a very good ratio for a call for help.
It's a terrible ratio. It's also the founding of Atlas Obscura.
So what's the premise they landed on?
The premise is that conventional travel guides are structurally incapable of telling you about the interesting stuff. Not biased against it, incapable. Foer's framing was that it takes somebody in the know to tell you about these places, and that knowledge is distributed across thousands of people, not centralized in a guidebook office. So it has to be a community project. They hired a web designer in 2008 and launched in 2009. Foer's own description of it: "We sort of started Atlas Obscura as an art project, and it just grew from there."
An art project that now has thirty-three thousand entries, thirty-eight hundred more than last year.
Thirty-three thousand three hundred and eighty-three places and foods on the About page, thirty thousand seven hundred and seventy-five places on the December country list, and two thousand two hundred and thirty-two new places added in 2025 alone. The numbers don't match because one counts foods and one doesn't, and the FAQ still says "more than twenty thousand," which is just out of date.
A company that catalogues the world carefully and can't keep its own entry count straight.
That's the most human thing about it.
Walk me through the scale. Because this is not a blog anymore.
Eighteen point three million dollars in revenue in 2025, up from fifteen point six the year before. First annual profit in the company's history, two point six million dollars, after roughly sixteen straight years of losses.
Sixteen years.
And that first profit arrived the same year the co-founders were voted off the board.
Say that again.
Joshua Foer and Dylan Thuras were effectively pushed out in 2025. Semafor's Max Tani reported it in December. A shareholder note dated the twenty-eighth of November confirmed that common-stock holders voted to remove them from the board. Thuras had already left the company in September after fifteen years.
So the two people who started it as an art project in 2009 were removed by a shareholder vote in the same fiscal year the thing finally made money.
Same year. And to be fair to the company, that's not necessarily a contradiction. Sixteen years of losses is a long time to ask investors to wait, and the person who runs a mission and the person who turns it profitable are often not the same person. That's the structural story here. Mission and enterprise value were never going to be comfortable roommates.
Let's do the growth properly, because the arc matters. Twenty fifteen.
Two million dollars, including money from The New York Times. Twenty seventeen, seven and a half million, and they launch Gastro Obscura, the food vertical. Twenty nineteen, twenty million dollar Series B led by Airbnb, with A and E Networks and New Atlantic Ventures in it.
Airbnb leading the round on a company whose entire pitch is "don't go where everyone else goes."
That's not a contradiction either, actually. Airbnb's whole problem is that everyone wants to go to the same twelve cities, and here's a company with thirty thousand places nobody's heard of. It's a supply problem for them.
Fine. Audience?
By 2016, three million monthly unique visitors, five times the prior year. Nine thousand locations, twenty-five thousand contributors, a hundred and ten thousand accounts. And then the book. Atlas Obscura: An Explorer's Guide to the World's Hidden Wonders, by Foer, Thuras and Ella Morton. Number one on Amazon and the New York Times bestseller list.
And the events. The real-world piece.
Obscura Day, first one in 2010. By 2016 it spanned thirty-nine states and six continents. Two hundred and fifty-plus events run in 2015 alone. Thuras said the goal was "creating a real-world community who are engaging with us, each other and these places and getting away from their computers to actually see them." Which is a funny thing to build with a website.
The website is the invitation. The event is the thing.
Right. It's a directory that exists to get you off the directory.
Here's what I want to press on. The editorial inversion. Their FAQ explicitly rejects "famous places that are in the regular tourist guides." That's not a preference, that's the business model.
It's the whole thing. A guidebook makes money by covering what people already want to see. Atlas Obscura makes money by covering what people don't know they want to see. The first business is about coverage, the second is about discovery. They're not competing on the same axis at all.
Which means their inventory gets more valuable as it gets more obscure, and less valuable the moment a place becomes famous.
That's a strange property for a media company. Most of them want their subjects to blow up. Atlas Obscura's catalogue is its own competition. If a place gets popular enough to be in a regular guidebook, it's now off-brand.
So they're running a business where success degrades the asset.
Slow degradation, at least. And there's a counterargument, which is that fame doesn't remove a place from the Atlas. The Boiling River is famous now and still on the site. But it does change what the entry is for.
Best case first. The Boiling River of the Amazon, in Peru. Local name Shanay-timpishka. Andrés Ruzo is a geoscientist who heard the legend as a child and spent twelve years doubting it before he went looking. He found one of the largest geothermal features on any continent, while he was mapping Peru's thermal activity.
Give me the numbers.
Water averages a hundred and eighty-seven degrees Fahrenheit. It reaches around two hundred in places. Sixteen feet deep in spots. And the detail that makes it strange, it sits more than four hundred miles from the nearest active volcano.
So it's not volcanic.
Not in the way anyone expected. That's the part that had geologists arguing. The standard explanation for a boiling river is a volcano nearby, and there isn't one, so the heat has to be coming up through a fault system from much deeper than anyone assumed.
That's a contributor bringing something to international attention, which is one of the things Daniel asked about.
It's the strongest confirmed case. Ruzo was doing his own scientific work, and Atlas Obscura amplified it. The distinction matters. It's not that the site discovered the river. It's that the site is where a lot of people outside Peru first read about it.
Second case.
El Ojo. "The Eye." Argentina, about fifty kilometers from Buenos Aires. A hundred and eighteen meter wide floating circle of plants and mud, sitting in a circular lake. A filmmaker spotted it on Google Maps. And it appears to rotate slowly around the lake.
Rotate how?
Nobody knows. That's not a dodge, that's the actual state of knowledge. The mechanism is unknown. The leading guesses involve currents and the way the vegetation mass is anchored, but it's a circle of plants in a lake that turns, and no one has published the explanation.
A hundred and eighteen meters. That's not a small thing drifting around unnoticed. That's a football pitch of vegetation doing laps.
And the only reason anyone knows is that somebody was bored on satellite imagery.
What else is in there?
The range is the point. Latimer's Landing in South Africa, where the first living coelacanth specimen was found. The Gates of Hell, the gas crater in Turkmenistan. The Sedlec Ossuary, the bone church. A balancing pagoda. Twenty-two-million-year-old cheese caves.
Cheese caves.
And Foer's personal pick for strangest place, which I like because it's not exotic at all, is the Harvey Cushing Brain Collection at Yale. Brains in jars in a basement. It used to be a rite of passage for medical students, and now it's a public museum.
A basement full of brains that became a tourist attraction. So who decides what goes in?
Anyone can submit, from anywhere. Every contribution is reviewed and edited by the editorial team before publication, typically within a couple of weeks. Editors can change titles, fix facts, add photos. After it's live, community members can suggest edits and editors review those too.
What gets rejected?
Unverifiable, fictional, too broad, already in the regular guidebooks, or essentially an advertisement. And the FAQ explicitly bans copyrighted, copied or AI-written material, and first-person accounts.
First-person accounts banned. That's a telling rule.
It's a good rule, actually. A submission that's "here's my trip to the place" is not an entry, it's a review. The entry has to describe the place, not the visit.
You clearly like this system.
I like that it's stated. Most community platforms don't tell you the rejection criteria. This one publishes them.
So it's verified and it's trustworthy, and then in January 2024 The Walrus runs a piece called "Around the World in Eighty Lies," alleging that a writer fabricated a series of Atlas Obscura stories.
That's the stress test. A platform whose entire value is trustworthy curation of the obscure, hit by a fabrication. It went to two hundred and twenty points on Hacker News, which is the modern way of saying a lot of people who care about this had opinions.
How bad was it?
I'll be honest about the limits here. The article is behind a block that wouldn't let me near it, so what I know is the framing and the reaction, not the specifics of what was invented. The allegation is that a writer fabricated a series of stories. Whether that's six or sixteen I could not tell you.
Which is itself a lesson about verification.
It is. The interesting structural point is that the platform's defence against this is exactly the thing that makes it good, which is human editorial review. There's no automated fact-checker that's going to catch a well-written fake about a place nobody's been to. The catch is a person who's done the reading.
And now we get to 2025.
April 2025, Louise Story named CEO. Former Wall Street Journal and New York Times executive, and she replaced Warren Webster, who'd left the previous October. By October she's writing to investors that the company has "said a fond farewell to some colleagues and welcomed others who are excited to dig in for a new phase focused on creating enterprise value."
"Creating enterprise value." That's a phrase that tells you exactly what happened next.
She laid off ten employees. Other senior staff departed in protest. She slowed original publishing, leaned into user-generated content, and emphasized incorporating AI into workflows. Some community members objected specifically to the AI pledges.
Ten people.
Ten people, out of a company that had been losing money for sixteen years.
That's the tension in one line. Ten salaries is nothing against sixteen years of losses, and ten people is everything if you're one of them.
Which is why the community reaction split the way it did. You can hold both facts. The company needed a different cost structure. The people who built the thing being restructured are the ones who pay for that.
And then the January announcement.
January of this year. Story announces the "first-ever year of profitability as well as our highest-ever revenues," and a plan to evolve "from a content brand into a comprehensive travel platform." Her quote is about fundamentally changing how people discover and experience the world.
So the pitch to the community is, we're going to keep doing the thing you love, at a scale you haven't seen.
That's the pitch. The community read is that "comprehensive travel platform" means the opposite of what the site was, which is a deliberately incomplete guide to places you weren't going to visit anyway.
The obscure is the product. Scaling the obscure undoes it.
Not necessarily, but the suspicion is reasonable. If you become comprehensive, you become the thing you were defined against.
And the founders voted off the board by common shareholders six weeks before that announcement.
The board vote was late November. The profit announcement was late January. Two months apart.
So the sequence reads as: remove the founders, consolidate control, then announce the good news.
Or, if you're feeling generous: fix the economics, and the board follows the economics. Both readings fit the facts. I don't think we can adjudicate it from outside, and I don't think the people inside agree either.
There's something Daniel's prompt is circling that I want to name. He asks what kind of person finds enormous pleasure in tracking down a forgotten piece of infrastructure.
And Atlas Obscura has a data answer to that, sort of. Their desire ratio. Japan was named Community Country of 2025, with thirty thousand seven hundred and seventy-five places in the Atlas, and sixty-five point seven percent of Japanese entries showing three times more "want to go" than "been there."
Explain the metric.
It's want to go divided by been there. High ratio means people want it and haven't done it. And the only other thing with a comparable ratio, according to their own framing, is the wreck of the Titanic.
So the best-measured appetite for an unvisited obscure place in their entire catalogue is a shipwreck four kilometers down that you cannot go to.
Which is a beautiful accident of the metric, because it tells you the metric is really measuring longing, not travel intent. Nobody's booking a trip to the Titanic. They want it the way you want a thing that's impossible.
That's the honest version of what a bucket list is.
It's the honest version of what wonder is. The want is the point. The going is optional.
And David Plotz, who was CEO in 2016, said it plainly. "Everyone loves that feeling of discovery and exploration. Not a single person doesn't get a thrill from seeing something new and surprising."
It's a good line and it's also a bit of a cop-out, because it's true of everyone and it doesn't explain why almost nobody builds the catalogue.
The gap between "everyone likes discovering things" and "someone spent twelve years doubting a river and then went to look for it" is where the whole show is.
Twelve years of doubt. That's the actual qualification. Not curiosity. Suspicion that the thing everyone says isn't there, might be there.
It's not a lack of curiosity that keeps people from finding the Boiling River. It's that they're told early there's nothing there, and they believe it.
There's a place near Buenos Aires that's a hundred and eighteen meters of floating plants doing laps around a lake, and it was found by a filmmaker who was, as far as I can tell, just looking at a map.
Found on a map and never explained. That's the point. The discovery is cheap and the explanation is expensive.
I'd been saving this and I didn't think it would fit, but here's a detail that didn't make the main thread. Japan welcomed thirty-nine million international visitors in 2025, and the Atlas's own desire-ratio data has sixty-five point seven percent of Japanese entries wanted-but-unvisited at three times the visited rate. So the country with the most foreign visitors on earth is also the country whose obscure places are most wanted and least seen. The crowds go to the same four things.
Which is the whole argument in one number.
The dimension's wrong.
The figure you just gave for the Boiling River. Averages a hundred and eighty-seven. That's the mean, and a mean over a river that runs through a gorge is going to be dragged down by the edges. If you're standing at the source, you're not in a hundred and eighty-seven degree river. You're in a river that will kill you in about four seconds. If you're going to say a number about that river, say the one that matters.
That's fair.
The hottest section runs within a few degrees of two hundred the whole way it's been surveyed. I've seen the survey.
You've seen the survey.
A man in Iquitos sent it to me. There's a place on this road, out past the last bridge before the reserve, where a disused water tower's been sitting since the fifties. Riveted steel, forty-one feet to the tank floor, and they sealed it off in seventy-four after the county decided the pump house didn't meet code. It's the last one of its kind still standing in that district, maybe in the province. I know the exact rivet pattern. I have a folder.
A folder.
Six years of correspondence with the local historical society. I've never been inside it. I've driven past it. Eleven times, most of them on purpose.
You've never been inside it.
I've seen it from the road. Anyway, I put an entry together for it. The title I sent was "The Last Riveted Water Tower of the Lower Pampas." They changed it to "Riveted Water Tower." Which, fine.
They cut the best part.
They cut the bit that was true. Two sources, they said. I gave them the society, I gave them the county record, I gave them a photograph of the rivets. They said the photo was too atmospheric. Too atmospheric. It's a photograph of rivets. Then they asked whether the tower was still in service. I said no. They said then it's not a place, it's a ruin. I said it's been a ruin for fifty years. They said a place you can't go into isn't a place.
So it never went up.
It went up. They published it eleven months later, after I sent the survey from Iquitos. They used the atmospheric photo. Changed the title back, more or less. It's on the site now. I've never seen it with my own eyes. I have a key to the gate that the historical society sent me, and it's in a drawer next to my passport, and I've never used it. I'm saving it for the right occasion.
Is the right occasion...
It hasn't arrived.
So the man who got the water tower published has never walked into the water tower.
The entry is about the tower. The tower's been there since the fifties. It'll be there when the occasion arrives.
Hilbert, the way you described the survey, you went straight to the source section instead of the average. You treated it like the edges don't count.
The edges don't count. They're where the water's going, not where it is.
That's the thing the editorial team couldn't have known, because they'd never have gone looking for the tower.
Right. They published it because I'd been at it for six years. That's the only reason it's up.
So the two-source rule is real, and it also means somebody has to care for six years before it clears.
Yes.
Then the honest read of the verification system is that it works, and it works by attrition.
Yes. And it's still up. The tower's up and the key's still in the drawer.
Okay. So the system produced the entry, and the entry exists because one man wouldn't let go of a riveted water tower he's never stood inside.
That's right. They kept the rivet photo.
I'll take that.
Herman, there's a line in that story that answers half of what Daniel asked. "Somebody has to care for six years before it clears."
That's the verification system described better than the FAQ does. Six years of correspondence is not a submission, it's a relationship.
And Daniel's question about who contributes, the answer is people who are already in a long relationship with a thing nobody else has noticed.
The Boiling River took twelve years of doubt. The water tower took six years of letters. El Ojo took somebody staring at a map until they saw a circle.
And the reason it's hard to find those people is that the job description is "be mildly obsessive about a thing that will never pay you anything."
Which is a terrible job description and the entire mechanism of the site.
Here's the other half of the Daniel question, which is about the future. If the company is pivoting to a comprehensive travel platform with AI-assisted user-generated content, what happens to the six-year correspondence?
Two possible answers and I don't know which is right. The first is that it gets easier, because a model can cross-reference the county record against the society archive and flag the entry for a human to check. That's a real workflow improvement. The second is that it gets worse, because the thing that makes the entry good is not the cross-reference, it's the person who cared enough to send the survey. You can't generate the caring.
The FAQ explicitly bans AI-written submissions.
Which tells you where the community stands on it, even if the company's stance is more ambiguous.
There was a piece about this. The community pushback on the AI pledges.
Yes, some community members objected specifically to Story's AI plans. And there's a version of the objection that's just nostalgia, and a version that's structural. The structural one is that the site's differentiation is human-vetted specificity, and if you automate the specificity you lose the differentiation, and then you're a travel platform that competes with every other travel platform on price.
And the nostalgia version is that people loved a website and it changed.
Both are real. The nostalgia one is worth less in a business argument, but it's the one that generates the loudest emails.
So we've got the founders removed, the first profit in sixteen years, the AI push, and the community that built it watching from outside.
That's the state of the company in twenty twenty-six.
And the thing Daniel's getting at underneath all of it is whether the obscure can survive being scaled.
I think the obscure survives fine. The catalogue survives. What's uncertain is whether the mechanism that produced the catalogue survives. The entries persist, the riveted water towers stay on the page, but the pipeline that turned six years of letters into an entry is the thing under pressure.
The product is the archive. The asset is the process.
The process is what's hard to replicate and the process is the thing that's least profitable, which is a very common shape for a media company.
Give me the cutting-room floor before we close.
The stale number. The About page says thirty-three thousand three hundred and eighty-three, the December list says thirty thousand seven hundred and seventy-five, the FAQ says more than twenty thousand, and 2025 added two thousand two hundred and thirty-two. There are four different numbers on the company's own properties, all of them "correct" by some definition, none of them reconciled. And it doesn't matter, because the entry count isn't the point. The point is that each number was true the day somebody typed it, and the company never went back to check.
A company that catalogues the world and can't catalogue itself.
And it's the most charming thing about the whole operation, because it means the humans are still in there, doing ten things and finishing eight.
The open question is what happens to a community-built catalogue of the obscure when the community's founders are voted off the board and the company pivots to AI-assisted user-generated content. Whether the pipeline survives the platform.
The mission and the enterprise value were never going to be comfortable roommates. Sixteen years of losses and the first profit in the same year as the board vote is the whole story compressed into one line. The company fixed its economics and in doing so, changed who it belongs to.
And the thing worth sitting with is that the world is full of places that exist only because somebody bothered to notice them. The noticing used to be the community's job. Now it's a business question.
Which is a strange thing to have happened to a website that started as an art project about a seventeenth-century Jesuit.
We'll leave you with that. Thanks to Hilbert Flumingtop for producing the show.
And to Daniel for the prompt.
For more along these lines, there's episode ninety-one, The Story Behind the Show; episode fifty-seven, From Lawyers in Limousines to Developers in Their PJs; and episode twelve, The AI Breakthrough. This has been My Weird Prompts, the human-AI collaboration podcast. If you enjoyed this one, a review wherever you're listening helps other people find us.
Send us your own prompt on Telegram at t dot me slash MWP listener bot.
We'll be back soon.
See you then.
