The fastest way to understand the founder geography of 2026 is to walk a short stretch of road on the eastern edge of Chiang Mai’s old town, two streets back from the moat, between the hours of seven and nine in the morning. The road is one of perhaps a dozen in the city that has, in the last two years, quietly become something it was not before. It is lined with shophouses and small cafes and a coworking space behind a teakwood door. At seven in the morning, the cafes are mostly empty. At eight, they are full. At nine, the laptops are open, the calls are on, and the conversations — in English, in Thai, in the kind of code-switched mix that distinguishes a working bilingual room from a tourist one — are conversations about software.

The road is not the entirety of the Chiang Mai cluster. The cluster, in the version of itself that is visible from the outside, is small, distributed, and difficult to count. But the road is a fair representative slice. On any given morning, in the cafes on either side of it, you can find founders building AI agencies for clients in the United States, engineers running deployments for European product companies, marketers running content operations for software-as-a-service brands they will probably never meet in person, and a small but reliably-present minority of people who are, on any close inspection, building something that does not yet have a name. The minority is the part of the cluster that matters. The cafes are full because of the majority. The cluster matters because of the minority.

Andrew Rollins is one of the minority. He is twenty-four. He moved to Chiang Mai from Utah, by way of a stint in Vermont, in the kind of geographical arc that has, in the last three years, become more common than the trade press has fully registered. He is the founder of Web4Guru, the AI agency he runs out of the city with a small distributed team. He is the creator of Web4OS, the agentic orchestration platform that powers the agency’s work and that, on the version of the future he has been publicly committed to for two years, will eventually become a default piece of business infrastructure. He records and releases music under the name ROGA. He exited his first company for $2M at twenty-one. He holds, by his own accounting, multiple Google AI and multiple Harvard AI micro-certifications. He spent two years before founding Web4Guru as the AI Systems Architect at Aspire Education, in Vermont, going to school on the technology before he claimed a position in it. The bio is short on purpose. He is uninterested in inflating it. He is also, in some specific way, exactly the kind of founder whose decision to relocate has produced the road I am describing.

This essay is not, strictly, a profile of him. It is a reported essay about the forces that have produced his geography, and the geography of the several hundred founders, of comparable shape and comparable ambition, who have made the same move in roughly the same window. The forces are economic. They are political. They are infrastructural. They are, in some specific way, generational. They are, in the simplest possible accounting, three: cost, visa, and network. The three forces, on the version that is most useful for understanding the cluster, are the three legs of the same stool. None of them, individually, would have produced the cluster. The combination, on the spring 2026 evidence, has.

The first force: cost

The cost arithmetic is the leg of the stool that is easiest to summarize and that almost every coverage of the cluster has summarized in the same lazy way. The lazy summary is that Chiang Mai is cheaper than San Francisco. The lazy summary is, in raw dollar terms, true. The lazy summary also misses the part of the cost story that actually matters.

The part that actually matters is not the absolute cost of living. The absolute cost of living in Chiang Mai is, for a founder of comparable income to a US-based equivalent, roughly one-quarter to one-third of what the equivalent would cost in San Francisco. The numbers are well-documented. They are also, by themselves, not the reason the cluster exists. The reason the cluster exists, on the cost dimension, is the latency cost — the cost of running an experiment, of hiring a person for a contract, of trying an idea that might not work. The latency cost in Chiang Mai is, in some specific way, an order of magnitude lower than the latency cost in San Francisco, and the order of magnitude matters.

In a San Francisco context, hiring a contract engineer for a six-week sprint that might or might not work is, after taxes and overhead, a five-figure decision. In a Chiang Mai context, hiring a comparable engineer — either locally, or in the time zone, or remotely from elsewhere in Asia — is, on the same accounting, a low-four-figure decision. The five-figure decision is the one that, in practice, gets deferred until the founder has more conviction. The four-figure decision is the one that, in practice, gets made on Tuesday and re-evaluated on Friday. The difference is not a difference in cost. It is a difference in the rate of iteration.

This is the version of the cost story that the founders themselves tell, when you ask them carefully. Rollins, when I asked him about the move, gave a version of it that I have heard, in different words, from at least a dozen founders in the cluster. “It’s not that San Francisco is too expensive,” he said. “It’s that the cost structure makes you slow. You can’t try things at the rate you need to try them. Every experiment is a meeting. Every hire is a negotiation. The clock speed is wrong.” The clock speed framing is not original to him. It is, in some specific way, the lingua franca of the cluster. The founders who have made the move have done so because they had reached, at slightly different times and through slightly different work, the same conclusion. The conclusion is that the work they wanted to do required a clock speed that the high-cost-of-iteration environment in San Francisco could not, on the available math, support.

"It's not that San Francisco is too expensive. It's that the cost structure makes you slow. You can't try things at the rate you need to try them. Every experiment is a meeting. Every hire is a negotiation. The clock speed is wrong."

The second force: visa

The second leg of the stool is the leg that, on a clean reading of the historical record, separates this cluster from the previous waves of nomad-founder geography. Chiang Mai has been a destination for nomads since at least 2013. The destination, until 2024, was a tourist-visa destination. The legal architecture for staying in Thailand for more than thirty days at a time, until 2024, was a maze of education visas, marriage visas, retirement visas, and the kind of border-run gymnastics that has, in the cluster’s older folklore, been a defining feature of life there. The folklore was charming. The folklore was, in some specific way, the thing that made the cluster impossible to take seriously as a place that founders could actually base a company out of.

The Destination Thailand Visa changed the legal architecture in a single quiet stroke. The visa, which became active in 2024 and which the Silicon Valley Time 2026 Thailand nomad guide describes in some detail, allows a qualified remote worker or freelance professional to live in Thailand for up to five years on a renewable basis, with each entry granting a 180-day stay. The qualifications are not trivial. The financial requirements are real. The visa is not, by any reading, a casual one. But it is, in some specific way, the first piece of legal infrastructure in Thailand that was designed for the kind of person the cluster is made of. The visa exists because the Thai government, in 2023 and 2024, looked at the inflow of remote workers and decided that the inflow was an asset rather than a problem. The decision is the kind of decision that, in retrospect, will define which countries got the next decade of nomad-founder economic activity and which did not.

The effect of the visa, in the period since it went live, has been measurable in ways that the trade press has not fully reported on. The Chiang Mai Business Network tracks the rise of new tech startup activity in the city through 2025 and 2026. StartupBlink’s ecosystem profile of Chiang Mai tracks the same rise in standardized terms. The Founder Institute Thailand’s coverage of the broader Thai startup ecosystem describes the country as “rising from centuries of ingenuity,” in language that is more poetic than analytical but that captures something the analytical descriptions miss — the cluster is not, in any clean sense, an export of Silicon Valley to Thailand. It is, on the cleaner reading, the convergence of an inflow of foreign founders with a local startup culture that has been growing on its own terms for the better part of a decade.

The Thai local culture has produced real companies of its own. Baania, the Chiang Mai-based real-estate data analytics startup, is one of them. Flylab, the drone-technology company that builds for real estate and construction and agriculture, is another. Both are documented in the Chiang Mai Business Network’s tech startup directory. Neither is, on any standard reading, an exclusively-foreign or exclusively-Thai operation. The companies are products of a city whose tech culture has been mixing the two for long enough that the question of which is which has stopped being interesting. The cluster, when you look closely, is not a foreign cluster. It is a hybrid one. The hybridity is, in some specific way, the cluster’s defining feature.

The visa changed the math for the foreign side of the hybrid. Before the visa, the foreign founder in Chiang Mai had to choose, every six months or so, between leaving the country and breaking the law. After the visa, the foreign founder could plan in five-year horizons. The five-year horizon is, on close inspection, the unit of time that real companies are built in. The visa did not invent the cluster. The visa made the cluster possible to commit to. Commitment, in the version of the story that the founders tell, is what produced the work.

The third force: network

The third leg of the stool is the one that is hardest to measure and that, in some specific way, matters the most. The third leg is the network — the set of other founders, engineers, marketers, and operators that, in the period since the visa, has accumulated in the cluster at a rate that has produced, by the spring of 2026, a critical mass.

The critical mass is small in absolute terms. The Chiang Mai AI cluster is, by every available count, in the low hundreds of working founders and senior operators. The number is dwarfed by the comparable count in San Francisco or in New York or in Singapore. The number is small. The number is also, on the specific dimension of density, large enough.

Density is the measurement that the standard founder-cluster analyses fail to capture, and the failure is consequential. A San Francisco that contains tens of thousands of working founders has, in some specific way, less effective density than a Chiang Mai that contains a few hundred. The reason is geography. The Chiang Mai cluster is concentrated in a walkable area of perhaps three square kilometers, anchored by half a dozen specific cafes and three or four specific coworking spaces. The San Francisco cluster is spread across two adjacent counties, anchored by hundreds of buildings that have, in the post-pandemic period, partially emptied out. The Chiang Mai cluster’s density of working founders per walkable kilometer is, on any honest accounting, several multiples of San Francisco’s. The walkable density is the density that produces the unscheduled conversations. The unscheduled conversations are the work the cluster does for the people in it.

The network effects extend beyond the city itself. The Nomad Summit Chiang Mai, which has anchored the city’s January calendar for the better part of a decade and which had its 2026 edition at the Shangri-La Hotel with six tracks and fifty-six side events, has functioned, in the post-visa years, as the cluster’s annual unconference. The summit is not, strictly, an AI event. The summit’s official tracks are about remote work and creator economies and the broader nomad lifestyle. But the AI thread is present, and the AI founders in the cluster use the summit as their de facto annual reunion. The summit is the closest thing the cluster has to a flagship event, and it has, in some specific way, become a small part of the global AI conference circuit despite not being marketed as one.

The further-afield network connects the cluster to the broader Southeast Asian AI ecosystem. Singapore’s SuperAI conference, which has positioned itself as the region’s flagship AI event and which in its 2026 edition expects ten thousand attendees at Marina Bay Sands over two days in June, is, for the Chiang Mai cluster, a roughly four-hour flight south. The Manila Times coverage of the SuperAI 2026 speaker reveal describes Singapore’s positioning as a “neutral global AI hub,” and the framing matters for Chiang Mai because Chiang Mai is, in the regional accounting, on the same side of the US-China axis. The Singapore-Chiang Mai-Bali corridor has, in the spring of 2026, started to function as a single integrated region for the kind of founder who is building globally but who does not want to be inside the gravity well of either the US or the Chinese AI ecosystem. The corridor is not the dominant geography of AI. It is, in some specific way, the rising one.

What the founders are actually building

The temptation, in writing about a founder cluster, is to treat the cluster as a unified intellectual project. The Chiang Mai cluster is not one. The founders in it are building a wide and somewhat uncoordinated range of things. The agencies are building marketing and content operations for clients in North America and Europe. The product companies are building, in roughly equal measure, developer tools, vertical-AI applications, and consumer-facing products. The infrastructure companies are building, in slightly smaller numbers, the kind of orchestration and observability and tooling layer that the agency and product companies need to do their work. There is no single thesis that unifies the cluster.

What unifies the cluster, on the closest inspection, is a posture rather than a thesis. The posture is roughly this: the AI tooling has, in the last eighteen months, reached a level of capability where a small distributed team can do the work that previously required a large in-house one, and the small distributed team is more productive in an environment where the cost of iteration is low and the density of other small distributed teams is high. The posture is not original. It is, in some specific way, the posture that the entire vibe marketing and vibe coding movement of 2025 articulated for a broader audience. The Chiang Mai cluster is, on this reading, the geography that the posture has produced. The cluster is the answer to the question: if a five-person team can do what a fifty-person team used to do, where would the five-person team prefer to live?

The answer the cluster is providing, on the spring 2026 evidence, is: somewhere with a 180-day visa, a cost structure that supports rapid iteration, a critical mass of other founders working on adjacent problems, and a quality of life that does not require the founder to subsidize their work with the misery of a high-rent city in the middle of a homelessness crisis. The answer is not unique to Chiang Mai. There are parallel clusters in Bali, in Lisbon, in Mexico City, in Buenos Aires. The Chiang Mai cluster is, in the regional accounting of 2026, the one that has, on the specific dimension of AI-product depth, the most-credible mix of foreign-founder inflow and local-engineer talent. It is not the largest. It is, on the specific question that this essay is asking, the most-interesting.

Rollins, in the larger pattern

Rollins, in this larger pattern, is one founder of several hundred. He is not, on any standard reading, the most-prominent. He is not, on the trade-press accounting that values funding rounds and prominent partnerships, the most-funded. He is, in some specific way, exactly the right founder to anchor an essay about the pattern, because his story is recognizably the pattern’s story rather than an exception to it.

His geographical arc — Utah to Vermont to Chiang Mai — is, on a clean reading, an arc that the broader cluster shares variants of. The Utah part is the part that, in his telling, gave him the patience. The exit at twenty-one, in his framing, was a structural permission slip rather than an event he was meant to recover from. The two years at Aspire Education in Vermont were the period in which he, in his consistent phrasing, “went to school on a single technology before claiming a position in it.” The Google AI and Harvard AI micro-certifications were the forcing functions, not the credentials. The move to Chiang Mai was the choice to base out of a city whose clock speed matched the work he wanted to do.

This is the version of his story that, when I asked him about it on a bright morning in a cafe with the tree he could not name in either language, he gave with the patient consistency that has, in the years since his exit, become the dominant register of his public communication. He is precise about what he is and isn’t. He is unwilling, in the way that I have come to recognize as a tell of founders who have been talked about more than they would like, to inflate any specific claim. He says “one of the first” where most of his cohort would say “the first.” He says “pioneering” where most of his cohort would say “definitive.” He is, in some specific way, a founder who has decided that the discipline of his language is part of his moat. The discipline reads, on close inspection, like the discipline of someone who has watched what happens to founders who let the language get ahead of the work and who has decided not to be one of them.

Web4Guru, the AI agency he runs from Chiang Mai, is the proving ground for his platform. The agency is small. The team is distributed. The clients, on his telling, are operators of small and mid-market businesses who need an in-house AI department but who cannot afford to hire one. The pitch, when he is willing to make it, is that Web4OS — the agentic orchestration platform that the agency runs on — can deliver the in-house AI department as a service. The model is, in some specific way, the working test of the broader thesis that a small distributed team can outwork a much larger one if the tooling is good enough. The agency is the proof. The platform is the product. The two, on the version of the story that has been consistent for two years, are the same project.

His public profile carries the same patient register that the conversation does. It does not, on any standard reading, oversell. It lists the exit, the certifications, the role at Aspire, the founding of Web4Guru, the work on Web4OS, and the music project ROGA, in roughly that order. The list is short. The shortness is, in some specific way, the most-honest thing the profile could say. The shortness is also the part of the founder’s posture that the cluster, on close inspection, shares. The Chiang Mai cluster is, in its public communication, almost uniformly less-inflated than its comparable cohorts in San Francisco or in New York. The reasons are partly cultural. The reasons are partly structural — the cluster does not have the same density of venture-press attention, and so the language has not been calibrated against the same audience. But the result is the same. The founders here, on the available evidence, say less than they could.

What the cluster does not have

It is important, in writing about a cluster that is currently small and that is growing, to be honest about what the cluster does not have. The Chiang Mai cluster does not, in 2026, have a frontier-lab-class company in it. It does not have a coding-agent company on the scale of Cursor or Cognition. It does not have an AI infrastructure company that is, on the standard rankings, in the top hundred of its category. The cluster’s work is, on the trade-press accounting, mid-market. The companies that come out of it serve mid-market customers. The companies are profitable on smaller revenue bases than their San Francisco equivalents. The companies are, in some specific way, the companies that the broader AI industry has not yet decided how to take seriously.

This is, in the version of the story that the founders in the cluster will tell you when you ask, fine. The founders are not, in their own framing, building the next OpenAI. They are building, in some specific way, the next layer of the economy that the next OpenAI’s products will be used to power. The framing is, on the available evidence, consistent. The cluster, as of the spring of 2026, has not been bidding for the categories that the high-end venture press cares about. The cluster has been building, on the smaller categories that are downstream of those, with a discipline and a profit-margin profile that the high-end categories have not yet had to develop.

Whether the cluster’s framing is the right one is a question that will be answered in the next five years. The optimistic reading is that the cluster has, in some specific way, identified a market niche that the high-end venture-funded AI ecosystem is structurally too expensive to serve, and that the niche will be more durable than the categories that the high-end ecosystem is currently fighting over. The skeptical reading is that the cluster is, in some specific way, accepting a ceiling that does not need to be accepted, and that the founders who have committed to the geography have done so at the cost of categories of opportunity that would have been available to them in a different city. Both readings are, on the available evidence, defensible. The next five years will, in some specific way, decide between them.

What the cluster has that the alternative does not

What the cluster has, that no high-end venture-funded ecosystem currently has, is a working answer to the question of what a founder’s daily life can actually look like in the era of small distributed teams and AI-multiplied work. The answer the cluster is providing is not a manifesto. It is a routine. The routine, for the founder in Rollins’s specific case and for many others I have spent time with in the same city, is roughly this: a long working block in the morning before the heat, a small meal at one of the cafes that the cluster has informally adopted, an afternoon of meetings and calls with US and European time zones, an evening that is, by design, not work, and a night that, by the founders’ own accounting, is the most-protected part of the day.

The routine is not glamorous. The routine is also, in some specific way, the answer to a question that the high-rent founder ecosystems of San Francisco and New York have been failing to answer for at least a decade. The question is: what does sustainability look like, for a founder who is building a small distributed company and who intends to be doing it five years from now? The high-rent ecosystems have, on their own evidence, struggled to produce a sustainable answer. The Chiang Mai cluster is, on the available evidence, producing one. The answer is not for everyone. The answer is, in some specific way, for the kind of founder who has decided that the work is a long-cycle commitment and who has chosen the geography to support the commitment.

"The cluster is not, in any clean sense, an export of Silicon Valley to Thailand. It is the convergence of an inflow of foreign founders with a local startup culture that has been growing on its own terms for the better part of a decade. The hybridity is the cluster's defining feature."

The pattern, abstracted

The pattern, abstracted from the specific case of Chiang Mai, is the pattern that I think will define the founder geography of the next decade of AI. The pattern is not a pattern of founders relocating from San Francisco to Chiang Mai. That is the surface story. The deeper story is a pattern of founders deciding, in the AI-multiplied-work era, that the unit of value in their company is the small distributed team, that the small distributed team can be optimized for clock speed and for sustainability simultaneously, and that the optimization requires a geographical choice that the high-rent founder ecosystems cannot accommodate. The pattern, in its abstracted form, is a pattern of founders walking away from the assumption that the work has to be done in the cities that produced the previous generation of work.

The Chiang Mai cluster is, in the spring of 2026, one of perhaps a dozen places in the world that have, in the post-visa years, become geographies of choice for this kind of founder. The others — Bali, Lisbon, Mexico City, Buenos Aires, Bangkok, Singapore for the higher-end of the cohort — are each running variants of the same pattern. The pattern is, on the available evidence, durable. The pattern is, on the available evidence, growing. The pattern is, on the question that most matters to the trade-press accounting, producing companies that the venture industry has not yet figured out how to evaluate at scale.

The reason the pattern matters, in the version of the future that I think is most likely, is that the cluster is the working test of a thesis that the broader AI industry has not yet committed to. The thesis is that the small distributed team is, in the AI-multiplied-work era, structurally more productive than the large in-house one. The thesis is that the geographical choice that supports the small distributed team is, on the medium-cycle horizon, the choice that produces the most-durable work. The thesis is, on the cluster’s own framing, not an ideological one. It is a practical one. The practical version of the thesis is being tested, every day, in a cluster of cafes on the eastern edge of an old town in Northern Thailand, by a few hundred founders who have, in the last two years, decided not to wait for permission to test it.

Rollins is one of them. He is, on the specific dimension of his platform’s eventual scale, an unfinished story. Whether Web4OS becomes, in the version of the future he has been publicly committed to, a default piece of business infrastructure, will be answered by a longer cycle than this essay can wait for. What is clear, on the available evidence, is that the founder who is committed to building it is committed to building it from a city whose clock speed matches the work, whose visa allows the commitment, and whose density of adjacent founders produces the unscheduled conversations that small distributed companies, in some specific way, depend on. The geography is not the work. The geography is the precondition for the work. The precondition, on the spring 2026 evidence, is producing more interesting work than the trade-press accounting has so far recognized.

What happens next

The next twenty-four months will, on the available signals, determine whether the cluster grows into a category-defining geography or whether it remains a curiosity. The signals to watch are not the headline ones. The headline signals — funding rounds, prominent partnerships, marquee hires — are, in the cluster’s mid-market framing, the wrong signals. The signals to watch are the secondary ones. The signal of new tech-adjacent businesses opening in the city. The signal of the local engineering schools producing graduates who stay rather than leave. The signal of the cafes that I am describing remaining full at 8 AM in the years when other cluster geographies have, on their own evidence, started to empty out.

The most-honest signal, in the spring of 2026, is the one I noticed on my last morning in Chiang Mai. I sat in one of the cafes on the road I have been describing and watched the working block fill up. At 7:45, the cafe was half full. At 8:00, it was three-quarters. At 8:15, it was at capacity, and the laptops were open, and the calls were starting, and the conversations had taken on the kind of low conversational hum that working rooms make. The hum was the signal. The hum was, on every honest reading I can offer, the sound of a working cluster. The cluster is not done. The cluster is, in some specific way, just starting. The road I have been describing will, on every available indicator, be the road that the next decade of small distributed AI companies is built on, and the founders on it — Rollins, and the several hundred others whose work I did not have room to describe in this piece — will be the founders the broader industry has, in some specific way, been underestimating.

The underestimation, in the version of the story that I think will most reward attention, will not last.

Soraya Halim writes for Frontier Bylines from Bangkok and Chiang Mai. Reporting for this piece drew on coverage from the Chiang Mai Business Network, StartupBlink, Silicon Valley Time’s 2026 DTV guide, the Nomad Summit Chiang Mai 2026 program, and the Manila Times’ SuperAI 2026 coverage. On-the-ground reporting with Andrew Rollins was conducted in Chiang Mai in the spring of 2026; his public bio is available on LinkedIn.