01. Young people are pushed to the front of the stage
Not long ago, leaders of Guangdong Province walked into Tencent’s headquarters and had a discussion with Tencent’s management. In front of a long table, Tencent founder Ma Huateng, who is wearing a jacket and still has thick hair, looks particularly young.
But the focus of people’s attention soon fell on the 28-year-old star scientist Yao Shunyu sitting side by side with him. In a photo that was later widely circulated by netizens, Yao Shunyu was dressed plainly, with a focused expression and a strong sense of “nerd”.
In the discussion that day, Ma Huateng and Yao Shunyu spoke respectively. While observing, Yao Shunyu stood behind Ma Huateng and accompanied him. This rare scene in the same frame quickly aroused discussion among people on the Internet.
Before this, for such a grand and high-profile occasion, the only people the big bosses could bring out were presidents and CEOs, and a young AI scientist and the founder would not be arranged to attend with great fanfare.
When Yao Shunyu sat next to Ma Huateng, Alibaba’s team of post-90s senior executives was also quietly taking shape. However, unlike Tencent’s poaching of star researchers, Alibaba prefers to take the “young and strong route” and hand over the leadership of core training positions to internally selected people.
Under the call of Alibaba Group CEO Wu Yongming to support young people, Liu Dayiheng, who was born in 1993, became the “number one” person in charge of Qwen’s large model. At the 2026 Yunqi Conference, after Wu Yongming finished talking about Alibaba’s AI strategy, Liu Dayiheng took the stage to talk about models.
▲Liu Dayiheng
In the past, Liu Dayiheng’s presence in the industry was not high. In a photo from the Yunqi Conference, the other four big guys are all wearing dark suits, but Liu Dayiheng, who is in the middle, only wears a yellow pullover sweatshirt, which is quite obvious.
Different from Yao Shunyu’s narrative of overseas elites, Liu Dayiheng, a young man who completed his bachelor’s, master’s and doctoral degrees at Sichuan University, has neither an Ivy background nor work experience in a major overseas company. Before joining Alibaba, he was a young man who had been busy in papers and laboratories for a long time.
However, Liu Dayiheng’s technical foundation is not weak to stand out in a place where stars shine like Damo Academy. He has a PhD background in computer science. He once interned at Microsoft Research Asia, was selected into Huawei’s “Genius Youth” program, and participated in Qwen’s pre-training from its early days.
Before the former head Lin Junyang resigned, Liu Dayiheng was originally responsible for pre-training within Qwen. In 2026, his responsibilities were expanded from pre-training to post-training and coding, and his title became Qwen LLM project leader, directly led by Wu Yongming.
In addition to technical guys like Liu Dayiheng, Chen Yusen, who was born in 1992, also stood on Ali’s power stage. During the special performance of the Yunqi Conference, Chen Yusen, who has curly hair, took out a microphone with three small green octopuses wrapped around his back, which made him feel relaxed at the beginning.
Chen Yusen first made a joke about the tension at work, saying that he hoped to have time to relax at work. When giving a live speech, he hardly looked at the teleprompter and went off-script throughout. However, compared with this relaxation, the burden on Chen Yusen’s shoulders is not small.
In June 2026, Chen Yusen took over as CEO of DingTalk and became the youngest business unit CEO in Alibaba’s history. At the Yunqi Conference in September, he had another title: Vice President of Alibaba Group.
02. The technical rights obtained are not the same.
Before the arrival of Yao Shunyu, Tencent had long been aware of the importance of AI, but there were many problems: excessive pursuit of benchmarks; models that “know how to take exams” but poor real-life scenarios; and the quality of data annotation has only been 60% to 70% for a long time.
In order to pave the way for the next business, Tencent directly gave Yao Shunyu the right to report directly to the headquarters. Yao Shunyu is also very good at getting rid of old assessment burdens. When he first took office, he made a request to Tencent President Liu Chiping, “not to look at the big model ranking list for a year.”
After that, Yao Shunyu pushed forward with node adjustments almost every three months. The organizational structure adjustment became commonplace for Tencent’s AI-related business. Some redundant platforms and content divisions were not ambiguous at all. Those that should be merged and those that should be abolished were not at all ambiguous.
Yao Shunyu’s boundaries of authority and responsibility have also been expanding. When he landed at Tencent in December 2025, he was the chief AI scientist and the head of the AI Infra Department and the Large Language Model Department. In July this year, his management authority was expanded and he took over the responsibility of the head of the newly established basic model department. He held multiple positions and coordinated the research and development of full-modal mixed element basic models.
After more than half a year of organizational rewriting, Tencent’s Hunyuan model has become stronger in the industry, and the morale and style of its AI business have taken on a new look. In the eyes of many, Tencent has finally officially sounded the clarion call for counterattack in the Agent field.
Unlike Yao Shunyu who implemented reforms from the top down, some other star researchers did not have such good luck.
After joining Xiaomi, Luo Fuli, the legendary DeepSeek “genius girl” who was poached by Lei Jun with an annual salary of tens of millions, began to exert her efforts. Not only did she make high-profile live broadcasts, but she also appeared frequently in interviews.
Unlike Yao Shunyu, who is cautious in her words and deeds, Luo Fuli, who is outspoken, often falls over in the court of public opinion. Her self-introduction in front of the camera seemed a bit reserved. During the recruitment process, she proposed that only sophomore and junior undergraduates be recruited. She was also accused of “creating anxiety.”
However, Xiaomi still gave her a rocket-like promotion speed. After only 10 months of joining the company, Luo Fuli was promoted to level 22 in September 2026. This is the highest level in Xiaomi’s ranking system, equivalent to the vice president level, setting a new record for Xiaomi’s fastest promotion speed.
In contrast, Wang Teng, who was in charge of Redmi’s business during the same period, took eight years to reach level 21. He was fired by Xiaomi last year for leaking secrets, and later went into his own sleep business. But it only took Luo Fuli 10 months to achieve a career advancement.
Of course, promotion comes with real delivery. Not long ago, MiMo-V2.6-Pro led by Luo Fuli scored 46 points in the Artificial Analysis Intelligence Index, ranking first among open source weight models, surpassing Kimi K3 and Qwen3.8 Max.
Lovely is pleased with the results of the new model. In her own words, “launched a silent surprise attack.” However, although it has won a reputation externally, it is not easy to implement internally. It even required extreme pressure to implement it.
She said she was shocked on the first day she used the new model. She tried to persuade the team to use it, but failed, so she gave a hard order: Anyone in the MiMo team who has less than 100 conversations tomorrow can resign.
The management style of “Genius Girl” is indeed domineering. Afterwards, Luo Fuli explained to the public that she did not really evaluate the team based on the number of conversations, but her purpose was just to force everyone to use it first. However, the internal team’s unwillingness to cooperate is already evident.
Whether it is Yao Shunyu’s top-down reform or Luo Fuli’s bottom-up implementation. The pressure faced by young technical leaders is essentially the same. Big bosses spend a lot of money. In the final analysis, what matters most is the AI achievements they bring.
03. Being popular and falling out of favor can happen overnight.
The support given by the bosses also exposed their respective situations.
While Ma Huateng was saying that the ship was leaking, he was also urging the ship to speed up; Wu Yongming personally led Token Foundry, which was equivalent to canceling the buffer of “reporting to the top”; Lei Jun also made a high-profile announcement of his determination to invest in AI and decided to invest 60 billion yuan.
The positions of the big guys are different, but their anxieties about the AI competition are the same. The bosses also tacitly changed the vocabulary they talked about: from “strategic patience” in the past to “personally”, from “continuous investment” to “60 billion yuan in three years”, etc.
However, what the young leader received was not a stable territory. When technology goes smoothly, they are the center of the company’s AI narrative and the company’s stars in the AI era. When the wind changes, rights are quickly taken back.
Although star researchers can bring direction, talent and morale, they cannot eliminate the inertia of large companies on their own. After all, what they really want to reshape is often not a model, but the organization, resource allocation and business logic behind the model.
Yao Shunyu’s test question is not just about “making the model stronger”, but also about reuniting the AI business of a giant company. He not only needs to correct the model training and evaluation system, but also connect research, computing power, and infrastructure with Tencent’s complex business scenarios.
In terms of business scenarios, Tencent has many products available, but each product has its own rhythm, interests and technical requirements. What Yao Shunyu has to do is to make these businesses willing to use Hunyuan, rather than looking for external models or building AI capabilities independently.
Liu Dayiheng faced another exam paper. Qwen’s technical assets are already very rich: open source models, code models, multi-modal models, and developer ecosystem are Alibaba’s most influential AI assets. But the thicker the assets, the harder it is to turn around.
After Liu Dayiheng took over, he must continue to catch up with the world’s strongest models and maintain the open source community. At the same time, he must serve Alibaba Cloud and convert the models into API, computing power and enterprise service revenue. At the same time, he must also form synergy with Alibaba’s internal e-commerce, DingTalk, Quark, and international businesses.
As for Chen Yusen, he must graft the speed of his previous entrepreneurial projects onto the complex organization of a mature platform like Qianwen Office, while also dealing with the platform’s own culture, products and commercialization issues.
However, once the route fails, the product is not paid for, or the business results are delayed, the resources and permissions originally opened by the big guys may be quickly withdrawn. Among them, Alibaba, which just staged the transfer of power in the first half of the year, has learned a lesson from reality.
In March this year, Lin Junyang, the former technical director of Qwen, who is only 32 years old and known as the youngest P10 in Alibaba, single-handedly created Qwen’s open source spirit.
But in the end, as the team structure changed from “vertical integration” to “horizontal division of labor,” Lin Junyang’s management scope was narrowed, and the reporting relationship and voice changed. In the end, he chose to leave his job and start a business.
The boss’s anxiety will not go away, it will only spread. The person who is pushed to the front and is in the same frame as the founder today may also be standing at the other end of the reporting chain tomorrow. What big factories leave to young people is never an iron rice bowl, but a limited-time pass.
Related Reading
- Nvidia live updates: Jensen Huang ousted, CFO Colette Kress to serve as interim CEO2026-11-13
- Are mangoes not afraid of red fruits?2026-11-12
- Original: After Huang Renxun, who is most likely to take over Nvidia?2026-11-12
- Samsung’s roll-screen mobile phone is exposed, creating a new form ahead of Huawei and Apple, and folding screens are expected to be eliminated2026-11-12
- Musk’s favorite car is going to be discontinued.2026-11-12
