Mark Chen, a senior researcher at OpenAI, revealed on a podcast aired on December 2nd that Meta's recruitment team would bring hot soup to candidates' doors in the competition for top AI talents. In response, OpenAI offered its own "homemade soup" - this seemingly absurd "soup war" has become the latest method of talent acquisition in Silicon Valley.
Chen said that Meta's "soup offensive" mainly targeted senior researchers, "they stood at the door with hot soup, which was really hard to refuse." However, according to his statistics, most of the core employees who were poached eventually chose to stay at OpenAI, "people have more confidence in the company's mission and technology roadmap, and soup is just an extra bonus."
Currently, OpenAI's research department has about 500 people, and they are working on over 300 cutting-edge projects simultaneously. Chen emphasized that exploratory research, rather than "following and replicating," is the company's competitive advantage. "We are willing to invest computing power in uncertain new paradigms instead of just focusing on others' papers."
Facing high salaries from competitors like Meta, Chen said that internally they use "clear communication of priorities" to allocate resources. "Leadership lies in the courage to cut secondary projects." He revealed that several new models with performance comparable to Gemini 3 are already under internal testing and will be released soon.
Industry observers point out that as the technological gap narrows, "emotional value" has become a new variable in recruitment. A bowl of hot soup precisely addresses the loneliness of late-night coding. As the AI talent gap widens, similar "warm-hearted recruitment" is likely to escalate further.
