So Daniel sent us this one — he's asking whether there are still inhabited places on Earth with zero internet of any kind. No cellular, no Starlink, no dial-up, no satellite phone relay. Places where if you want to send a single bit of data to the outside world, you have to physically travel somewhere else to do it. He's wondering if we've closed the door on the pre-internet era, or if there are swathes of the world that are truly dark. And the harder question tucked inside that — how would you even go about finding them?
This is exactly the kind of question where the official answer and the real answer diverge in ways that are genuinely fascinating. The ITU put out their Facts and Figures report for twenty twenty-five, and here's the headline that gets repeated everywhere: ninety-six percent of the global population is within range of a mobile broadband signal. Sounds comprehensive. Sounds like we've basically solved the coverage problem.
Ninety-six percent. That's the number that makes you think we're done here.
Right. And then the same report says two point six billion people are still offline. Those two numbers sitting next to each other — ninety-six percent coverage, two point six billion unconnected — that's not a contradiction, exactly, but it's telling you something important if you know how to read it.
Enormous amount of work. "Within range" means a cell tower exists somewhere that could theoretically reach your location, under ideal conditions, if you have a device that supports that band, if the tower isn't congested, if there's actually backhaul feeding data to that tower, and if you can afford the service. Any one of those conditions failing and you're effectively dark even though you show up as covered on someone's map.
So the first thing we have to do is define what "dark" actually means operationally. Because Daniel's question isn't about wilderness. It's about inhabited places where a resident cannot send or receive any digital data without traveling to another location.
And that definition matters because it excludes a huge number of places that technically have infrastructure but practically don't. I want to separate two things here. There's "no infrastructure exists" and there's "infrastructure exists but might as well not." The second category is much larger and much harder to measure.
Give me an example of that second category.
Remote communities in the Canadian Arctic and parts of Alaska. These places have satellite backhaul — the infrastructure is physically present. But the per-minute or per-megabyte pricing is so astronomical that for most residents, effective connectivity is zero. You might have a community center with a single terminal that can send email in theory, but in practice, a family living there cannot send or receive digital data as part of daily life. They're covered on a map. They're dark in reality.
So the coverage metric is measuring the wrong thing. It's measuring whether a signal can theoretically reach you, not whether you can actually use it.
And this gets to the core measurement problem. The ITU's two point six billion offline figure — that's a usage metric. It counts people who don't use the internet. But it doesn't distinguish between "chooses not to" and "cannot possibly." A retiree in Switzerland who refuses to get a smartphone and a farmer in the Central African Republic with no tower within a hundred kilometers — they both count as "offline." The statistic flattens completely different realities.
So let's separate those realities. What does deliberately engineered darkness look like?
North Korea is the canonical example. And it's not just that they block the internet — lots of countries filter and censor. North Korea has built an entirely separate domestic network called Kwangmyong. It's an intranet serving somewhere between five and ten million users, and it has no global internet gateway for ordinary citizens. None. The routing simply doesn't exist.
So you can access websites — but only North Korean websites.
Right. It's a walled garden where the walls aren't just policy, they're physical. The fiber connections to the outside world exist — they have to, for government and military use — but they're firewalled at the routing level for everyone else. Foreign diplomats in Pyongyang can get 3G service through Koryolink, the local provider, but ordinary North Koreans can't. And the firewall is so complete that the primary method of getting outside data into the country is still physical media smuggling. USB drives, SD cards, DVDs smuggled across the Chinese border.
That's a fascinating inversion. The most sophisticated data transfer method available to most North Koreans is a person carrying a piece of plastic in their pocket.
It's pre-internet data transfer in a country that technically has fiber optic cables. And that's the thing — if you looked at a connectivity map that just showed infrastructure presence, North Korea would light up. They have cell towers. They have fiber backhaul. They have a domestic data center infrastructure. But the operational reality for twenty-five million people is that the global internet does not exist.
So that's the political category. A government actively chooses to sever the connection. What about the geographic and economic categories?
Let me start with what Starlink has actually achieved, because this is where a lot of the misconceptions live. As of mid-twenty-twenty-six, Starlink has about six thousand operational satellites in orbit at roughly five hundred fifty kilometers altitude. They claim service on all seven continents. And the coverage map does look impressive — it's a lot of blue.
But.
But. Service requires a phased-array terminal that costs five hundred ninety-nine dollars, plus a monthly subscription of a hundred twenty dollars. And that terminal needs a clear view of the sky. The satellites use Ku and Ka band frequencies — these are high-frequency signals that require line of sight. Dense forest canopy blocks them. Deep valleys block them. Heavy precipitation degrades them significantly.
So even within the blue area on the map, there are micro-dark zones where the signal simply can't reach the ground.
And the economic barrier is even more fundamental. The World Bank estimates roughly seven hundred million people live on less than two dollars and fifteen cents a day — that's the extreme poverty line. A Starlink terminal costs five hundred ninety-nine dollars up front. That's nearly a year's income at the poverty line, for the hardware alone, before the monthly fee. The economics simply do not work for a significant fraction of the global population, and no amount of satellite launches changes that.
So Starlink has solved the coverage problem in the sense that a signal exists overhead, but it hasn't solved the access problem, and it arguably can't solve the access problem without a fundamentally different economic model.
Right. And this is where we get to the truly dark inhabited places. Not places where the signal is weak or expensive — places where there is no signal of any kind, from any provider, at any price.
How would you even find those places? You can't query a database of locations with no internet. ISPs don't publish non-service maps.
This is the search problem, and it's difficult. The absence of connectivity is by definition not recorded by connected systems. You can't ping a place that has no receiver. So researchers and humanitarian organizations use proxies. The best ones are: first, ITU universal service fund data — these are programs designed to subsidize connectivity in underserved areas, and you can look at maps showing where no funded projects exist at all. Second, WorldPop population density data, which uses satellite imagery and census data to estimate where people actually live. Third, cellular license holder maps — if no company has even bothered to acquire a license to operate in a region, that's a strong signal that no towers exist.
Cross-reference those three and the gaps start to appear.
And they're not small gaps. Let me give you two concrete candidates. The Amazon basin — Brazil alone has more than two hundred indigenous communities with no grid electricity. No grid electricity means no router, no modem, no device charging infrastructure. Some of these communities have a satellite phone reserved for emergencies, but routine connectivity? Zero. You cannot send an email from these villages. You cannot receive a text message. If you need to communicate digitally, you travel by boat to the nearest town with a connection.
That's inhabited, habitable, and dark by Daniel's definition.
Completely. Second candidate: the rural northwest of the Central African Republic. National internet penetration in the CAR was estimated at fourteen percent in twenty twenty-five by the ITU. And that's the national average, which includes the capital Bangui. In the rural northwest, there are areas with no cellular towers within a hundred kilometers. No satellite ground infrastructure for terminals. The combination of low population density, extreme poverty, and ongoing instability means no telecom company has built anything there, and no satellite provider has established a distribution channel for terminals.
A hundred kilometers from the nearest tower. That's not "poor signal." That's radio silence.
And the CAR is not unique. Parts of South Sudan, eastern Chad, the highlands of Papua New Guinea — these are places with permanent settlements, people living their entire lives there, and no pathway to send a single bit of data to the outside world without physically moving.
So the answer to Daniel's question is yes. Truly dark inhabited places exist. But they're shrinking, and the nature of the remaining darkness has shifted.
That's the key insight. The pre-internet era hasn't closed — it's moved. In the nineteen nineties, most of the world was dark because the infrastructure simply didn't exist. Fiber hadn't been laid, cell towers hadn't been built, satellites were sparse and expensive. Today, the remaining dark spots are dark for different reasons. Some are political — North Korea's firewall is an active choice to sever the connection. Some are geographic — the Amazon canopy blocks satellite signals, deep valleys create radio shadows, permafrost makes fiber burial prohibitively expensive. But most are economic — the business case for connecting a village of two hundred people with an average income of a dollar a day simply doesn't pencil out for any commercial provider.
And that's not a technological problem anymore. It's a market failure.
The technology exists to connect almost any point on Earth to the global internet. We have satellites in low Earth orbit. We have solar-powered cell towers that can operate off-grid. We have mesh networking protocols that can extend a connection across kilometers. The question isn't "can we connect this village." It's "who pays for it, and who maintains it, and who trains people to use it, and who provides the devices, and who replaces them when they break."
That chain of dependencies is the real infrastructure. The satellite is just the first link.
And every link in that chain costs money and requires institutional capacity. This is why I get frustrated with coverage maps that paint everything blue and declare victory. A Starlink terminal in a village with no electricity is not connectivity. It's a paperweight. You need solar panels or a generator, you need someone who can troubleshoot the terminal when it fails, you need a payment mechanism for the subscription — if the village operates on a cash economy, how do you pay a recurring digital bill? These are solvable problems individually, but they stack up, and in the poorest and most remote places, nobody has solved the stack.
So when we talk about closing the digital divide, we're using the wrong metric. Coverage is the easy part. Affordable, usable, sustainable connectivity is the hard part.
And it's harder to measure, which is why coverage dominates the conversation. It's a clean number. Ninety-six percent. You can put it on a slide. "Affordable, usable, sustainable connectivity" is a mess of qualifiers that requires actual fieldwork to assess.
Let's talk about what the fieldwork actually reveals. You mentioned the Amazon basin. What do we know about connectivity patterns there from people who've actually gone and looked?
There's been some excellent academic work on this, particularly from Brazilian and Peruvian researchers mapping connectivity in the Amazon. What they find consistently is that the official coverage maps — the ones that show ninety-six percent population coverage — are generated by propagation models that assume flat terrain. They take a tower location, apply a radius based on the technology generation, and declare everything inside that circle "covered."
And the Amazon is not flat terrain.
The Amazon is hills and dense canopy and winding rivers that create radio shadows in every direction. A tower that a propagation model says should cover a village thirty kilometers away might be completely blocked by a ridge with hundred-foot trees on top of it. The only way to know is to go there with a signal meter. And when researchers do that, they consistently find that actual coverage is significantly less than modeled coverage — sometimes by thirty or forty percent in heavily forested areas.
So the ninety-six percent number is an upper bound at best, and probably inflated.
Almost certainly inflated. But nobody has the incentive to correct it downward. Governments want to show progress. The ITU aggregates what governments report. Telecom companies want to show their coverage is extensive. The entire institutional ecosystem is oriented toward overstating connectivity, not understating it.
Which means the truly dark places are even harder to find — they're not just absent from maps, they're actively obscured by maps that claim they don't exist.
The maps are lying, and the lies are systematic. And this creates a weird paradox. If you're a researcher trying to find unconnected communities, you can't trust any official source. You have to triangulate from negative space — look for the places where no subsidy program operates, no license has been issued, no tower construction has been documented, and population density data suggests people live there.
That's detective work, not data analysis.
It really is. And it's detective work that most funding agencies aren't interested in paying for. They want to fund connectivity projects, not connectivity gap analysis. So the dark places stay dark and also stay invisible to the systems that might eventually connect them.
Let me push on something. You said the remaining darkness is primarily economic and geographic. But isn't there also a temporal dimension? In some places, connectivity exists but only intermittently — the generator runs for a few hours a day, the satellite link is available when the NGO staff are in town, the cell tower works until the diesel runs out.
That's a crucial point. Intermittent connectivity is its own category, and it's probably more common than truly zero connectivity. But from the perspective of someone trying to live a digitally connected life, intermittent might as well be zero for most purposes. You can't run a business that depends on timely communication. You can't participate in online education. You can't receive emergency alerts reliably.
So we might want to add a third category to our typology. There's no infrastructure. There's infrastructure that's too expensive to use. And there's infrastructure that's too unreliable to depend on.
And that third category is the one that's growing fastest as cheap equipment gets deployed without adequate maintenance budgets. A solar-powered cell tower installed by an NGO in twenty twenty-two that stopped working in twenty twenty-four because the batteries failed and nobody replaced them — that community shows up as "covered" on the map forever, but they've been dark for two years.
The map has a memory, but the infrastructure doesn't.
Beautifully put. And this is the gap between the ITU's snapshot methodology and lived reality. The ITU surveys governments periodically. A tower that existed during the last survey cycle exists in the database permanently, or at least until someone actively reports it as decommissioned. Nobody's job is to report decommissioned towers in rural Chad.
So if Daniel wanted to find a truly dark inhabited place tomorrow — not read about one, but identify one on a map and verify it — what's the actual method?
Here's how I'd do it. Step one: pull the ITU's universal service fund data and look for administrative regions with zero funded connectivity projects of any kind — no fiber rollout, no tower subsidy, no satellite terminal program. Step two: overlay WorldPop population density data at the one-kilometer resolution to confirm people actually live in those regions. Step three: cross-reference with the GSMA's mobile coverage maps, which track where cellular license holders have actually deployed infrastructure, not just where they've modeled coverage. Step four: look for regions where no cellular license has even been issued — if no company has bothered to acquire spectrum rights, there's almost certainly no infrastructure. The overlap of all four filters is your candidate list.
And then you have to go there and check.
You have to go there. There's no remote sensing method that can tell you definitively whether a location has zero connectivity options. You can rule out the obvious — no towers visible in satellite imagery, no fiber routes on infrastructure maps — but you can't rule out that someone has a satellite phone, or that a humanitarian organization installed a Wi-Fi hotspot that isn't in any database. The absence of evidence isn't evidence of absence.
Which brings us to the paradox Daniel was circling. The hardest part of finding a place with no internet is that it has no way to tell you it exists.
That's the line that should stick with people. A connected place announces itself constantly — every device pings a tower, every transaction leaves a digital trace, every social media post is geotagged. A dark place leaves no trace in any connected system. It's not just invisible to the internet. It's invisible to the entire apparatus we use to measure the internet.
It's the digital equivalent of the dark matter problem in physics. We infer its existence from the gaps in what we can see.
And just like dark matter, most of the mass is in the gaps. Two point six billion people offline. That's not a rounding error. That's a third of humanity.
But as you said, that number includes people who have coverage and choose not to use it, or can't afford to use it. How many of those two point six billion are truly dark by Daniel's definition — no option at any price?
Nobody knows the exact number, and that's the point. The best estimates I've seen from researchers who've tried to disaggregate the ITU data suggest it's somewhere in the low hundreds of millions — maybe two hundred to four hundred million people who live in places where no connectivity infrastructure exists at any price point. The rest of the two point six billion are covered but not connected, for reasons of cost, literacy, device availability, or relevance.
Relevance is an under-discussed factor. If you live in a community where nobody you know is online, where no services you need are online, where no content exists in your language — the internet isn't a thing you're excluded from, it's a thing that doesn't pertain to you.
And that's a harder problem than running fiber. Content and services in local languages, digital literacy training, a reason to connect beyond "the UN says you should" — these are the unglamorous parts of closing the digital divide that don't make it into the satellite launch press releases.
So where does this leave us on the question of whether the pre-internet era is closed? My read is: it's not closed, but it's been redefined. The pre-internet era used to be a technological condition — the wires hadn't been run. Now it's an economic and political condition — the wires exist or could exist, but the market and the state have decided, actively or passively, not to extend them to you.
And that's a more durable form of darkness. A technological gap can be closed by invention. An economic gap can only be closed by subsidy or by rising incomes, and the places that are still dark are dark precisely because neither of those things is happening there.
The satellite constellations are supposed to change this calculus. Starlink and its competitors — OneWeb, Amazon's Project Kuiper — they're building systems that can theoretically serve any point on Earth without ground infrastructure beyond the terminal. If terminal costs come down by an order of magnitude, if someone solves the power problem, if distribution channels reach the last mile...
That's a lot of ifs. And even if all of them are solved, you still have the subscription cost. A hundred and twenty dollars a month is more than the monthly income of someone at the extreme poverty line. You'd need the price to drop by a factor of twenty or thirty to make it viable for the poorest seven hundred million people. That's not happening with the current generation of technology.
So there's a floor. A "dark floor" below which the economics will never work without sustained subsidy.
I think that's right. And we should be honest about that rather than pretending that launching more satellites will somehow solve poverty. Connectivity is downstream of economic development, not upstream of it. Giving someone internet access doesn't create an economy — it amplifies an economy that already exists. If there's no economy to amplify, the terminal gathers dust.
That's a sobering conclusion. But it's also a useful corrective to the techno-optimism that dominates this conversation. The places that are still dark are dark for reasons that satellites alone can't fix.
And they'll stay dark until someone addresses those reasons directly. Which means the pre-internet era isn't a closed chapter. It's an ongoing reality for hundreds of millions of people, and it's sustained by structural forces that show no sign of disappearing.
So if you wanted to find a truly dark inhabited place tomorrow, your best bet is the intersection of extreme poverty, low population density, difficult terrain, and institutional neglect. The Amazon basin, the Central African Republic's rural northwest, parts of South Sudan and Papua New Guinea. These aren't uninhabited wilderness. They're places where people live, and where the global internet simply doesn't reach.
And the number of such places is shrinking, but the rate of shrinkage is slowing. We've picked the low-hanging fruit — the places where it was relatively easy and relatively profitable to build infrastructure. What's left is the hardest cases, and they're hard in ways that don't yield to purely technological solutions.
That's the real answer to Daniel's question. Yes, truly dark inhabited places exist. No, we haven't closed the door on the pre-internet era. And the interesting part isn't where they are — it's why they're still dark, and what that tells us about the limits of infrastructure as a solution to exclusion.
The hardest part of finding a place with no internet is that it has no way to tell you it exists. And that's not just a search problem. It's a description of what it means to be dark in a connected world. You're not just offline. You're off the map.
And now: Hilbert's daily fun fact.
Hilbert: In the Inca Empire, quipu record-keepers used knotted cords to track tax obligations, and one particularly detailed quipu from the Chancay culture encodes a llama-to-potato conversion rate of roughly one adult llama equaling one hundred forty-seven chuño freeze-dried potatoes, which a modern economist might describe as the earliest known futures contract on tubers.
I have so many questions about the negotiation that produced that exchange rate.
One hundred forty-seven feels very specific. Someone counted.
We'll be back soon. This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop. If you want to send us a question like Daniel did, email the show at show at my weird prompts dot com.
