The careers of Z.ai's Tang Jie and Moonshot AI's Yang Zhilin, once teacher and pupil at Tsinghua University, show that China's AI leap is no sudden development
University lab nurtured the computer scientists who are using ingenuity and imitation to chase down Anthropic and OpenAI; ‘they know perfectly how to monetize their work’
Context & Ripple Effects
The WSJ profile closes a loop that our related coverage has been assembling piece by piece: the debate over whether Yang Zhilin 'left the US' framed him as an emigre, but the deeper story is that he and his former teacher Tang Jie were both formed inside Tsinghua's own system — the same institution behind DeepSeek and nearly 5,000 AI patents over two decades.
What changed is that both pupils of that system are now at the frontier: Yang's Moonshot delivered a 'DeepSeek moment' with Kimi K3 after early doubts about revenue, while Tang's Z.ai pursues an explicitly open strategy, arguing in a recent memo that frontier capabilities should stay as widely accessible as possible. The teacher-pupil framing matters because it shows China's labs aren't copying Silicon Valley careers — they're running a parallel apprenticeship model descended from Andrew Yao's Tsinghua course.
First-order effects
- Anthropic and OpenAI now face two frontier challengers founded by people trained in the same lab lineage, meaning their competition is coordinated by shared training culture rather than isolated startups.
- Moonshot and Z.ai are pursuing visibly different monetization paths — Kimi K3's breakthrough versus Z.ai's open-access positioning — so the Tsinghua pipeline is producing strategic diversity, not clones.
Second-order effects
- US labs' recruiting pitch of 'come to America' weakens when the counterexample is a domestic pipeline that produced China's highest-valued unicorn without it, forcing Western labs and governments to rethink talent-flow assumptions.
- If Z.ai's open-access stance gains traction among Chinese labs, closed-model pricing power at Anthropic and OpenAI faces pressure from free-to-inspect alternatives aimed at exactly the developers who benchmark frontier capability.
Third-order effects
- Frontier-lab formation is becoming institutionalized through universities rather than venture ecosystems — a structure where a single academic course can seed multiple competing labs, which has no clean equivalent in the US model.
- If the pattern holds, the relevant question for policymakers shifts from 'can China import talent?' to 'how fast can one university system keep minting frontier founders?' — a rate question regulators and export-control designers have barely modeled.
The trend: China's frontier-AI labs are increasingly homegrown products of an indigenous university pipeline — Tsinghua above all — rather than ventures built on returned US talent.