William Chen is 26 and cofounded Sapient Intelligence, a Singapore-based lab that says it is building artificial general intelligence. In an as-told-to essay [with Business Insider’s Aditi Bharade](<https://www.businessinsider.com/william-chen-sapient-intelligence-studied-china-tsinghua-university-silicon-valley-2026-10>), he explains a college choice that a lot of ambitious kids with his options would not make. He got offers from Georgia Tech and Carnegie Mellon, and from Tsinghua University in Beijing. He went to Tsinghua. Then he worked out a cheap way to get most of what he had turned down anyway.
I think the right way to see this is that a university is a bundle. You pay one price, in tuition or in years of your life, and get a package: a credential, a lab, professors, a scholarship if you’re lucky, and a few thousand classmates who will later be founders, investors and people who answer your emails. Usually you can’t buy those parts one at a time. Chen’s trick was to take the parts Tsinghua priced well and find the rest somewhere else.
Here is what he was weighing. Chen grew up in Shenzhen, went to high school in Michigan, did a lot of robotics there, and met his future cofounder Guan Wang, who was working with him on robotics and autonomous-driving projects as they both got interested in AI. Chinese higher education has a reputation for being obsessed with grades and tests, driven by the gaokao entrance exam. Chen says he cared less about grades because he wanted to start a company, so he worried that a Chinese university would be the wrong place. He went to look before saying yes, knocking on several professors’ doors. What he found, he says, was “very open,” with “great infrastructure there to support student entrepreneurs.” Tsinghua also offered him a scholarship. He enrolled in mechanical engineering, and Wang went too.
So he had the lab and the scholarship. What he didn’t have was Silicon Valley, and he didn’t want to lose touch with the American tech scene. The fix was his high school friends, who had spread out across American universities, the way people from Michigan high schools do.
The 20-day network
In his 2019 winter break, Chen spent 20 days crossing the US and shadowing those friends. He started at MIT, drove to Columbia, then Carnegie Mellon, the University of Pennsylvania and the University of Michigan, then flew west for USC, UCLA, Stanford and Berkeley. That is nine schools in 20 days, a bit over two days each, which matches his account of staying two to three days with each friend. He slept on dorm-room floors when the friend was a guy and in a nearby hotel when she wasn’t. He sat in classes, learned how the clubs and research worked, and met some professors to talk about robotics.
Think about the price. Four years at an American private university costs, roughly, four years at an American private university. Chen paid for a winter of driving and some hotel nights. In return he got a decent sample of the parts of an elite American education that matter most to a founder: what the labs are working on, and who is in them. He described the trip as a chance to “vibe check” and ask, “Hey, is there anything they’re doing differently?” Some of those friends now run startups, and he says they are still close. That’s a network, and he picked it up through the side door.
Sure, it isn’t the same thing. You don’t get the alumni directory, and a couple of days in someone’s seminar is not a degree. But the main thing an outsider loses by not going to a famous school is access, and access mostly runs through people. Chen already had the people. He just had to go and see them.
Then Elon Musk called, sort of
The bet paid off quickly. [Fortune reported last November](<https://fortune.com/2025/11/28/sapient-intelligence-william-chen-guan-wang-turned-down-elon-musk-agi/>) that while Chen and Wang were still students, an open-source model they built called OpenChat got enough attention that Musk emailed them through xAI. According to Chen, he offered a “multi-million dollar pay package.” They said no. “We decided that large-language models have their limitations,” Chen told Fortune. The pair worked in Tsinghua’s brain-inspired intelligence lab, which is the kind of open door Chen says he found on his scouting trip.
They founded Sapient in Singapore in 2024, after graduating. In July 2025 the company [open-sourced its Hierarchical Reasoning Model](<https://www.investing.com/news/company-news/sapient-intelligence-launches-braininspired-ai-model-for-complex-reasoning-93CH-4144308>), which has 27 million parameters. That is tiny by frontier-model standards. Sapient says it was trained on about 1,000 examples without pretraining and handles puzzles such as Sudoku-Extreme, hard mazes and the ARC-AGI benchmark. The [research paper is on arXiv](<https://arxiv.org/html/2506.21734v1>) if you want to check the claims yourself, and they are mostly the company’s own. [Forbes reports](<https://www.forbes.com/profile/wang-guan/>) that Sapient raised a $22 million seed round in December 2024 from investors including Vertex Ventures and Sumitomo, at a valuation of more than $200 million.
These days Chen says Sapient is building “latent-space” models that can solve complex problems much faster and more cheaply than current AI models. Its newest product, Praxist, is an autonomous research tool that companies can use to speed up their research. The [company’s site](<https://sapient.inc/en/about>) lists Wang as chief executive and Chen as cofounder and head of strategy and management. The team is about 50 people, split across Singapore, Beijing and Palo Alto.
That map is the college decision again, now as an org chart. You hire engineers in Beijing, where Tsinghua produces lots of them, and where more elite AI researchers now work than in the US. You keep an office in Palo Alto for the investors, customers and Stanford friends. You put the headquarters in Singapore, which is a reasonable place for a company that would rather not pick between the other two. Chen’s version of the bundle problem has scaled up with him: take what each place does well and don’t pay for the rest.
It also makes the most valuable thing Chen did at 19 look like a road trip. A lot of American universities have spent years arguing that their tuition buys a network you can’t get anywhere else. Chen got his for the cost of 20 days and some floor space.