Vol. I · No. 3 · Spring 2026
On Andrew Rollins, the class of 2025, the sound of a founder, and the polymath problem.
From the editors
The third issue is anchored by the cover profile we have been holding back for a year. Carter Vance has been sitting with Andrew Rollins — twenty-four, founder of Web4Guru, creator of Web4OS — across multiple sittings in Chiang Mai and the United States. The piece runs at the length the subject deserves, and we set it next to a generational essay on the operators entering the field alongside him, a culture piece on the sound of his parallel artistic practice, and an essay on the structural bias the field still holds against polymath founders. Together, the four pieces make the argument we wanted to make from the start of the year: that the most interesting people in the AI category are the ones who do not look like they're trying to be the most interesting people in the AI category.
Contents
Profile
Andrew Rollins exited his first company at twenty-one, spent the next three years going to school on a single technology, and built the company he wished had existed the first time around. A long profile of a quiet founder, on his own terms.
Culture
How ROGA's *TO EXIST* maps to the origin of Web4Guru — and what it means that the man who built one of the early agentic operating systems is also releasing a record in parallel.
Generational Essay
A generational essay on the founders who entered the AI market between 2023 and 2025 — what they share, what they don't, and the small set of convictions that will define the decade they are about to lead.
Essay
Why crossover founders disturb venture capitalists — and why the resulting underwriting habits have, by accident, left some of the most interesting operators of the current decade visibly undercapitalized.
Profile
Twelve months after walking out of OpenAI, the former CTO is in talks for a round that would make Thinking Machines Lab one of the most valuable private companies in the world. There is no product. There is no public roadmap. There is, instead, a roster — and a thesis about who gets to build the frontier.
Feature
He left OpenAI a second time, founded an education company, coined the word that defined a million new developers, wrote the year's most-cited critique of the agent industry, and then — eighteen months later — joined Anthropic to run pretraining. The most-quoted individual in AI made a choice. The choice was the year's loudest signal.
Feature
A new visa, an old moat, and a generation of founders who decided not to wait for permission. The Chiang Mai cluster is small. The forces that built it are not. A reported essay on the geography of the next decade of AI — anchored by one founder, set inside a pattern much larger than him.