Early on August 12, Lin Junyan announced on the social network X the founding of Pragmatik Labs in Shanghai, focused on developing next-generation agents. He explained the company's name by saying he studied linguistics at a friend's recommendation, then switched to computational linguistics and natural language processing (NLP). According to him, the name reflects a return to the place of events and pragmatism, as they believe artificial intelligence should strive for this.
After leaving the Qwen team at Alibaba, Lin became an entrepreneur. His list of investors includes Gaorong Ventures, HSG, and Tencent, and market reports value the startup at $2 billion USD.
Lin was born in 1993. He received a bachelor's degree from the University of International Relations of China and a master's degree from Peking University. In 2019, he joined the Alibaba DAMO Academy and moved to the core Qwen team, where he led the development of the Qwen series up to globally influential open-source models. On March 4, he announced his departure, saying: 'Goodbye, my beloved Qwen.'
Lin's startup emerged at a significant time: the status of scientists is growing in the Chinese technology industry. More researchers are joining model development teams, leaders of foundational models are close to top decision-making structures, and young researchers leaving large corporations can quickly gain capital due to their technical expertise. The knowledge possessed by these scientists is once again becoming a scarce industrial resource.
These specialists undergo systematic training in computer science or AI, have postdoctoral research experience, or engage in long-term cutting-edge research. In the industry, they act as a link between advanced scientific research and company decisions, defining the technical direction in the absence of ready answers and transforming research into engineering paths. This scarce resource extends to resource allocation: computing power, R&D budgets, and teams follow technical judgments.
China has previously witnessed a similar shift in power. When technologies reach maturity, markets, channels, and scale grow; when the industry ventures into uncharted territory, those who find new answers lead the way. Similar changes are also occurring within companies: Baidu placed Yu Tian, head of fundamental model R&D, directly under Robin Li; Tencent recruited Sunyu Yao, a student from Tsinghua Yao Class and Princeton graduate, to lead large models and AI infrastructure; ByteDance tasked Yu Yunhui, a Peking University alumnus and HKUST graduate, with leading Seed foundational model research.
Among startups, Kimi founder Yan Zhilin holds a bachelor's degree from Tsinghua and a Ph.D. from Carnegie Mellon, while MiniMax founder Yan Junze received a doctorate from the Chinese Academy of Sciences Automation Institute.
This pattern is evident across the entire industry: whether it is large firms or startups, the technical trajectory of the AI era is returning researchers to the center of power. China's main model ecosystem increasingly resembles the era of scientists, where leading researchers capable of anticipating the field's development also control the methods of resource allocation to achieve these goals.
