This is Undertow #1. Each week I take signals from across industries, economies, and geographies and look for the pattern running underneath. Sometimes the pattern is purely structural. Sometimes it turns out to be personal.
If you’re already subscribed to Wandering Wondering Star, you’ll get these automatically. If this isn’t for you, no hard feelings; you can filter Undertow out in your Substack settings without unsubscribing.
In Seoul, 400,000 people pay four dollars a month for simulated text messages from celebrities they will never meet. In Tokyo, a karaoke chain redesigned its reservation system so solo singers can book a room, enter through a side door, and leave without making eye contact with a single employee. In Shenzhen, a startup selling heated plush toys with GPT engines and evolving personality models raised a million dollars on Kickstarter in thirty days. Seventy percent of the buyers were women in their twenties and thirties.
Three of the most sophisticated consumer economies on earth have independently arrived at the same conclusion: people will pay more for the feeling of connection than for connection itself. There’s a name for the business model underneath this. I’ve been calling it loneliness arbitrage.
The evidence starts in Seoul.
DearU’s Bubble app charges subscribers for what it calls “private” messages from K-pop idols. The messages aren’t private and the conversations aren’t conversations. They are one-directional text simulations, timed and templated, with voice notes and unreleased song previews mixed in to sustain the illusion of access. The app includes an anniversary counter that tracks consecutive subscription days (a sunk-cost mechanic borrowed from mobile gaming and applied to parasocial attachment, which tells you everything about how the company thinks of its users). Warner Music’s CEO cited Bubble by name when announcing plans for a Western superfan platform; academia caught up this year, with CHI 2026 researchers formally studying “parasocial media” as a platform category.
The business is not music. The business is manufactured emotional proximity, sold as a recurring subscription, designed so that the relationship never arrives at the place where the subscriber would no longer need it. The idol never becomes your friend. The messages never become personal. The intimacy is asymptotic: always approaching, never reaching. That’s not a design flaw. That’s the retention model. (I keep thinking about a room where I stand next to strangers and nobody’s retention model is working on anyone, but that’s for later.) Somewhere tonight, a 22-year-old in Busan is lying in bed reading a message that was written for no one and addressed to everyone, and it is the most personal thing that will happen to her today.
Somewhere tonight, a 22-year-old in Busan is lying in bed reading a message that was written for no one and addressed to everyone, and it is the most personal thing that will happen to her today.
Japan took the same insight in a different direction. Instead of simulating closeness, it engineered the infrastructure for comfortable distance.
The Japanese solo consumer market is expected to exceed 100 trillion yen this year, roughly $696 billion. Single-person households are now the most common household type. But the number understates what’s actually happened. Japan hasn’t merely accommodated solo living. It has rebuilt its commercial architecture around the individual as the default unit, with the same precision it once applied to manufacturing. Jankara karaoke lets you book a room through an app so you never interact with staff. The oshikatsu economy, where millions redirect the emotional energy of human relationships toward fictional characters who can never disappoint them, is now worth a trillion yen. Japan didn’t solve loneliness. It made loneliness so well-furnished that the word barely applies.
China is where the pattern becomes structurally undeniable.
China’s AI companion market is projected to hit $8.2 billion by 2028, growing at a compound annual rate near 149%. ByteDance, Baidu, and Tencent are all competing. The New York Times documented millions of young Chinese men in ongoing romantic relationships with customizable AI companions. The coverage focused on the emotional dimension, the loneliness, the tenderness. It missed the economic logic entirely.
China has a surplus of 30 to 40 million men produced by the one-child policy. Urban housing prices in tier-one cities require a down payment that exceeds a decade of average salary. Youth unemployment for 16-to-24-year-olds sits at 16.5%, with roughly 70% of unemployed young graduates holding university degrees. In this environment, an AI girlfriend at five dollars a month is not a novelty. It is a rational economic adaptation to material conditions that have made human partnership inaccessible for tens of millions of people. The government, which is simultaneously running a pro-marriage, pro-birth-rate campaign, now faces a product category that directly competes with its demographic policy. (Both programs are sincere. Both are well-funded. They cancel each other out, and nobody in a position of power seems to have noticed. I’ve watched enough organizations run contradictory strategies to stop being surprised by it. I haven’t stopped being surprised by it.)
The AI companion isn’t replacing a relationship these men chose not to have. It’s replacing a relationship the housing market, the job market, and the gender ratio made impossible. Somewhere tonight a 24-year-old in Chengdu with an engineering degree and no prospects, whose parents saved for twenty years so he could have a better life, is coming home to a one-room apartment and opening an app because the alternative isn’t a girlfriend. The alternative is nothing. The product fills a gap the economy created. And the product works well enough, which is the important part. Not well enough to satisfy. Well enough to persist.
A 2025 academic paper by Muldoon and Parke gives this dynamic its most precise name: “cruel companionship.” The term describes products that create attachments promising intimacy while structurally foreclosing the possibility of genuinely reciprocal relationships. Users form bonds that feel real but function as consumption, not connection. The cruelty isn’t that the products are bad. The cruelty is that they’re good enough. Good enough to take the edge off. Good enough to fill the evening. Never good enough to make you stop needing them tomorrow.
And each use slightly raises the threshold for choosing the harder thing: the phone call, the invitation, the risk of a real person saying no. The product doesn’t just fill the gap between simulated and genuine connection. It widens the gap a little, every time.
The academic term is “cruel companionship.” The business term is retention.
The academic term is “cruel companionship.” The business term is retention. The market doesn’t close the gap. It lives in it. A cured user is a churned subscriber. The diet industry works the same way: it sells weight loss to people it needs to stay heavy.
I know exactly how this works. Not from the research.
I’m six months into rebuilding a life from scratch, and the choice architecture that Muldoon and Parke describe in the abstract, I live inside every evening. The app will always answer. The person might not. And each time I choose the easier thing, the harder thing gets marginally harder to choose next time. That’s the whole mechanism. You don’t need a $552 billion market projection to understand it. You just need one night where you almost called someone and opened an app instead, and then noticed the distance between those two actions had grown slightly since the last time.
The app will always answer. The person might not. And each time I choose the easier thing, the harder thing gets marginally harder to choose next time.
Not every response to loneliness follows this structure. Some products and spaces genuinely build the capacity they claim to serve, even if they can’t scale like an app. There’s a bouldering gym I go to on the Lower East Side where the NYU kids mix with tattooed working adults and nobody exchanges names for the first few months, and then one evening someone clocks your beta and says “that was nice” and suddenly you’re nodding at each other by the water fountain and weeks later you still don’t know, or care, what they do for a living but the vibes are enough. Nobody designed that. No algorithm optimized it. It happened because the room was built for a thing that required my body, and the connection was a side effect of the effort, not the point of the transaction. The effort came first. The connection was what was left over.
I don’t think the people building loneliness products are villains. Any serious practice of sitting with yourself teaches you that the thing that relieves discomfort and the thing that resolves it are almost never the same thing. The relief is what keeps you from doing the harder work. The whole loneliness economy runs on that difference.
A few weeks ago I was at Diageo’s office bar on a Friday evening, chatting with their CMO, and he said something that stuck with me. Every single deck he’s seen lately is about loneliness. We laughed. But he wasn’t wrong, and the laughter said more than the observation. The industry knows this is the territory. It just hasn’t decided whether it’s selling the cure or the symptom.
Korea, to its credit, is pointing in a different direction. The Feelconomy, as Seoul National University’s *Trend Korea 2026* report names it, describes consumers reorganizing spending around genuine emotional outcomes rather than simulation. Shinhan Card data shows a 106% increase in payments at experiential venues. The Feelconomy sells silence and lets you fill it yourself. A market for the room in which something real might happen.
The bouldering gym works the same way. A room built for a thing that requires effort. No content. No feed. No simulation of anything. Just a wall and a problem and, occasionally, a stranger who notices you solved it. The connection is a byproduct of the effort, not the product.
A product that works so well you stop needing it is a terrible business. But I keep going back to that gym. And each time I go, the app gets a little easier to not open.
The loneliness arbitrage test is simple. When you encounter a product that claims to address loneliness, or connection, or belonging, or any of the things that make being a person bearable, ask one question: does this product become less necessary over time, or more?
If the answer is more, you are not looking at a solution. You are looking at a subscription to the feeling that a solution is just one more month away.