From London Coffeehouses
to a Global Good Place Methodology Three centuries of history, two contemporary answers, and seven rules for making good places

Starting from London's first coffeehouse in 1652, humanity spent three centuries repeatedly validating the same formula: learning, socializing, and doing business were always one. Entering the modern history of "maker spaces," WeWork and Y Combinator offered two opposite answers—one proved "selling space" doesn't work, the other proved "selling network" is the moat; and in the 2020s, Silicon Valley's Hayes Valley and London's East End are simultaneously filling in a third answer. Finally, seven actionable rules, plus six thinkers, answer a more fundamental question: in an era when machines can do almost anything, why must people still "be together"?

In 1652, in St. Michael's Alley in London's financial district, a Greek named Pasqua Rosée opened the city's first coffeehouse. It soon earned a nickname: "Penny University"—a cup of coffee cost one penny, and paying that penny bought you the right to walk in and debate. Here, you might sit between a nobleman and a cobbler—identity didn't matter; what mattered was only the quality of your argument. The growth was astonishing: by 1663, London had over eighty coffeehouses, and by 1739, more than five hundred and fifty.

Here, your birth doesn't count; only your argument counts.

— 17th-century London coffeehouse, the egalitarian spirit of the "Penny University"

This is precisely what sociology has repeatedly emphasized: a true "good place" is naturally egalitarian and neutral—it pulls people of different identities to the same table, precisely because they are different, not because they are the same. And when a room full of people with diverse backgrounds begins to converse as equals, what grows is far more than idle chatter.

The Jonathan's Coffee House in London's Exchange Alley was where stockbrokers met and set prices—around 1698, the London Stock Exchange took shape over its coffee tables. The Lloyd's Coffee House on Lombard Street, opened by Edward Lloyd, gathered shipowners, merchants, and underwriters to exchange shipping news and underwrite voyages, eventually growing into Lloyd's of London and the entire modern marine insurance industry.

Coffeehouses were even dangerous enough to alarm the Crown. In late 1675, Charles II issued a proclamation attempting to suppress coffeehouses, calling them hotbeds of "spreading false news and sedition"—the resulting public outcry forced the ban to be withdrawn in just ten days or so. This episode恰恰 proved their weight: a place where strangers gather as equals to talk is itself a form of power.

In 1686, the Sicilian Procopio opened Café Procope on the Left Bank of the Seine; it remains in business today as Paris's oldest café. In the eighteenth century, it became the veritable "Enlightenment salon": Voltaire, Rousseau, Diderot, d'Alembert, and Beaumarchais all argued here over hot coffee. It is said that Diderot and d'Alembert's epoch-making Encyclopédie was conceived at this very coffee table.

The philosopher Habermas later gave such places an exalted name—the "public sphere." He said that eighteenth-century coffeehouses and salons were where "private people came together as a public": people read newspapers, argued, and formed public opinion here; a public reason independent of court and church rose from the steam of coffee. Its crux was packing "difference" into the same room: private individuals met as a public precisely because they were different from one another.

Two things of the modern world—markets and reason—
rose together from the steam of the same beverage.

"Good places" are by no means a Western patent. Chengdu's teahouses have been the most important public parlor of grassroots society for centuries: for the price of one cup of tea, you could sit all afternoon, chat, listen to storytelling, do business, or even ask the tea guests to publicly "mediate" disputes. Lao She's play Teahouse simply used a single "Yutai Teahouse" to contain half a century of China—the sign "No politics talk" often hung on the wall恰恰 proves that people couldn't resist talking politics.

In 1989, American sociologist Ray Oldenburg gave all of this a unified name. In The Great Good Place, he proposed the "third place"—neither home (first place) nor workplace (second place), but those places you can go to downstairs, without dressing up, where you can stay all afternoon just to chat: neutral, egalitarian, conversation-centered, accessible, with a crowd of regulars, relaxed atmosphere—a "home away from home." He warned that when these places disappear one by one, society slides toward isolation and fragmentation.

Looking back over these three centuries, we find something we've forgotten: learning, socializing, and doing business were originally one—it was modern division of labor that forcibly separated them into three different boxes: knowledge went into universities, idle talk into bars, transactions into office buildings. When China's AI-origin communities shout about the closed loop of "learning, socializing, and sales," they are actually unconsciously reviving the oldest formula of "good places." But reviving an ancient formula is easier said than done—the last person who wanted to turn "space" into a big business took a hard fall, while another who barely touches space became the biggest winner of this era.

Entering modernity, "making spaces for people who make things" has changed faces several times—and each face sold something different.

The most dramatic scene on this timeline is two nearly contemporaneous names offering opposite answers. One treated space as a product and collapsed spectacularly; one barely touches space yet became the most thriving institution of the AI era. They are WeWork and YC.

In 2010, Adam Neumann founded WeWork. Its story sounded incredibly sexy: not a sub-landlord, but a tech company "connecting people through space and changing the world," repeatedly invoking one word—community. SoftBank's Vision Fund kept pumping in capital; by early 2019, the valuation surged to $47 billion, making it one of the most valuable startups in America at the time.

But that summer, when the S-1 filing was laid open, the picture changed completely: 2018 revenue of $1.8 billion, but losses as high as $1.9 billion. Its invented "Community Adjusted EBITDA" stripped out real expenses like design, marketing, and management, dressing up losses as profits. The IPO was forced to cancel; Neumann resigned under pressure—yet walked away with an exit package of up to $1.7 billion, while SoftBank took over at a valuation under $8 billion.

The most biting footnote: Neumann turned around and founded Flow—a real estate company that stuffs "community feel" into apartment rentals, received a16z investment, and was valued at about $2.5 billion in 2025. WeWork's lesson is cold and clear: when a space sells mainly space itself, and community is just the packaging, then once the wind stops, it's nothing. Space has no moat.

In 2005, Paul Graham, Jessica Livingston, and others founded Y Combinator. Its model is counterintuitive to the point of simplicity: it barely gives you space. What it gives you is a small sum of money (now about $500K for ~7% equity), three months of intensive refinement, a Demo Day before top investors, and—most valuable—an entry ticket to the alumni network and a gold-standard brand.

Over twenty years, YC has funded over five thousand companies with a combined valuation of about $600 billion, incubating giants like Airbnb, Stripe, Coinbase, DoorDash, and Instacart, with over a hundred becoming unicorns. In the AI era, it hasn't become obsolete—it's burning even brighter: the 2025 winter batch was about 80% AI companies, with the entire batch growing an average of 10% per week; about a quarter of the teams had 95% of their code written by AI.

WeWork bet on the table; YC bet on the people sitting beside the table, and the entire network behind the door.

— Twenty years later, the verdict is in

This reveals the deepest secret of maker spaces: its moat is never the table itself, but the people beside it and the network behind the door. WeWork reversed cause and effect, so it collapsed; YC got it right, so it thrives more than ever.

The 2020s Third Path: Hacker Houses

Between the two poles of WeWork and YC, the 2020s grew a third path—returning to the oldest formula: co-living, curation, belonging. San Francisco's AGI House gathers AI builders into one building, hackathons by day, dinner with unicorn founders by night; South Park Commons calls itself an "anti-accelerator," specifically accompanying those who "haven't figured out what to build yet"; Vitalik's Zuzalu / Edge City brings hundreds of people to the seaside to live together for a month or two, experimenting as they go. They genuinely grow communities—at the cost of being high-threshold, elite-leaning, and hard to scale.

Zoom into the last two years, and the third path is no longer a few isolated experiments. Every "good place" starts with a small group's spontaneous gathering—hacker houses, shared desks, neighborhood cafés—but once validated as effective, almost without exception, they are rapidly "platformized" by capital or media. This time, China's AI-origin communities (see the other article in this series, "China Chapter") are not an exception, but the newest and fastest-running sample in this global pattern.

Sources: SF Standard · Washington Post (paywall; see also TechCrunch's 2023 coverage of Cerebral Valley AI Summit) · TechCrunch · Northeast Times (full list at end of article)

From hacker house to media brand, from neighborhood nickname to city KPI, the forms differ, but the direction is strikingly consistent: turning "gathering together" from a spontaneous act into a business that can be planned, invested in, and measured. Put the three models side by side, and the real gap of the AI era becomes clear: WeWork proved pure space doesn't work; YC proved network and curation are king—but YC only admits ~1%, takes equity, and forces you to sprint at unicorn speed; it gives you a ticket into the circle, but not a "home" you can return to every day where you won't be kicked out for failing. Hacker houses provide belonging, but are too expensive, too selective, too small.

This global gap is precisely the prescription this article aims to fill—breaking down "how to make a good place" into seven actionable rules. But first, remember a piece of advice from the public space research organization PPS: "To make a great third place, first learn to get out of the way." A good place is not a refined building that's been designed; it's an ecology that's been "cultivated"—the designer's task is to set up the skeleton, plant the hooks, and then yield the starring role to the people who come.

Connect the seven rules, and you find they all point to the same disposition: builder's humility. WeWork's mistake was wanting too much to "design" community, to monetize every inch of space; while London's coffeehouses, Chengdu's teahouses, and New York's Recurse Center endure precisely because they are willing to yield control—set up the stage, then step aside.

You cannot "manufacture" a community;
you can only build a place where a community
is willing to grow on its own.

These seven rules are, in the end, means. They serve a more fundamental conviction—a question not yet directly answered: in an era when machines can do almost anything, why must people still "be together"? This question belongs neither to urban planning nor to business models; it belongs to philosophy.

The most seductive promise of the "one-person company" is that "one person can still get things done." But hidden in that sentence is a substitution: getting things done and living well were never the same thing. Twenty-three hundred years ago, Aristotle drew this line clearly—humans are "political animals," naturally meant to live in community; a being capable of living in isolation without needing others "is either a beast or a god." The polis exists not merely to enable people to "live," but to enable them to "live well."

What AI does, precisely, is drive the cost of "living" down to the floor—making a living, producing, making things are all getting cheaper. But precisely because of this, it pushes the higher question, starkly, before every "super-individual": you can get everything done now, and then? This question cannot be answered by a desk, by compute, or by subsidies. It must be sought in philosophy.

Arendt's distinction is especially sharp. She divides human activity into three layers: labor (repetitive maintenance of survival), work (the craft of fabricating objects), and action (speech and concerted action among people). AI is rapidly taking over the first two—it can labor and it can fabricate. But "action," by definition, cannot be completed alone: it requires plural humans, a space where they "appear" to one another. Her famous dictum seems written for this era—"Power arises when people act together, and disappears as soon as they disperse." A desk can be remote; compute can be in the cloud; but "action" cannot be outsourced, because its precondition is that others are genuinely present.

Sociologist Putnam supplies the other half. In Bowling Alone, he distinguishes two kinds of "social capital": bonding is superglue, binding similar people ever tighter; bridging is the lubricant, connecting you to different people and worlds. A room where seven out of ten are twenty-something AI enthusiasts is excellent "glue" but a barren "bridge." And genuine creation almost always comes from people who "shouldn't be in the same room" meeting—this is precisely what Habermas meant: the crux of the public sphere is "meeting because of difference."

Here lies a trap we must be vigilant against. Anthropologist Marc Augé proposed "non-places"—airports, chain stores, highways: spaces that are interchangeable, designed only for passing through, not for staying; devoid of identity, relationships, or history. If an OPC space has only standard desks and transaction KPIs, it will degenerate into a "non-place" for passing through, not a "good place" for staying.

The Chinese philosophical tradition gives the repeatedly invoked "useless whitespace" an older name. Zhuangzi spoke of the "usefulness of the useless": the gnarled tree that wasn't cut down because it was useless for timber—precisely because it was "useless," it could flourish and provide shade for everyone. A place that converts every encounter into a transaction and fills every inch of space with KPIs may look efficient, but is actually sawing off its own trunk with its own hands.

And the sharpest cut comes from philosopher Byung-Chul Han. He says we have gone from disciplined "subjected individuals" to self-driven "achievement subjects"—no longer with an external boss, because each person has become their own boss, employee, and exploiter. The "one-person company" is precisely its completed form: you finally own the means of production (a computer plus a swarm of agents), but therefore also bear all the market's risks and all its loneliness alone. This is exactly what lets Marx's "alienation" return in new forms—the stronger the tools, the more people may be estranged from others and from themselves. And Marx's direction is precisely the endpoint this series has been building toward: the remedy is not sharper tools, but an "association of free individuals"—a place that can reconnect the scattered.

We can outsource production to agents,
but we cannot outsource "becoming human."

Thus, the philosophical status of public space in the AI era is fundamentally elevated. It is no longer an accessory to production, but infrastructure for resisting alienation and enabling "action" and "meaning" to occur. Outsourcing "being together" to machines is this era's greatest illusion—because it is precisely "being together" itself, not its products, that is the part of being human that cannot be automated.

A cup of coffee three centuries ago, a gaiwan tea in Chengdu, the bar counter in Wudaokou that changes face between day and night (described in the other article in this series)—they continue the ancient impulse of humans spontaneously creating "good places." This formula hasn't changed in three centuries—what has changed is that this time, when machines have taken over nearly all the "doing," we must consciously, with our own hands, rebuild it. This doesn't require policy or a district to happen: it can be a break room in a company where KPIs aren't discussed, a bookstore in a neighborhood willing to let people sit all afternoon, or a few friends liberated by AI into "one-person" operations, agreeing to gather every week, each writing their own code.