$3.3 billion: The "AI version of WeChat" secures a sky-high seed round, with Jensen Huang making an investment.

On January 21st, TechSugar reported that today, the newly - established US AI startup Humans&, which was founded just three months ago, announced that it had secured a seed - round financing of $480 million (approximately RMB 3.341 billion) and achieved a valuation of $4.48 billion (approximately RMB 31.186 billion), directly entering the unicorn club.

This round of financing was led by SV Angel and Georges Harik, the seventh employee of Google and a co - founder of Humans&. The institutions and individuals participating in the investment include Nvidia, Jeff Bezos, the founder of Amazon, Google Ventures (GV), and Emerson Collective (a company owned by Laurene Powell Jobs, the widow of Steve Jobs), etc.

Like many AI startups, there is a strong presence of Chinese people at Humans&. Six out of the 18 founding members are Chinese, accounting for up to one - third.

The founding team of Humans& (Source: Humans& official website)

The core concept of Humans& is: AI should empower humans, not replace them. The founders said that their goal is to develop software that promotes human - to - human collaboration, which can be understood as “An AI version of an instant messaging app”, and it can also handle tasks more suitable for machines, such as searching the Internet.

The company plans to use existing AI technologies to train AI in new ways. Its goal is to train a more interactive AI system, for example, one that can actively ask users for information, store this information for future use, and has “curiosity” and “memory”, allowing it to seamlessly integrate into group chats with colleagues, friends, or family members.

However, to secure such a large - scale financing in the seed round, Humans& doesn't just rely on its concepts. The co - founder team of the company consists of five people with backgrounds spanning academia and industry, and the founding members are from xAI, Anthropic, Google DeepMind, OpenAI, Meta, Reflection, AI2, Stanford University, and the Massachusetts Institute of Technology and other well - known companies and institutions.

Among the five co - founders, Andi Peng was a researcher at Anthropic, participating in the reinforcement learning and post - training of Claude 3.5 to 4.5. Georges Harik, the seventh employee of Google, was involved in building Google's earliest advertising system.

Eric Zelikman and Yuchen He are both former researchers at xAI. Among them, Eric Zelikman was deeply involved in the development of Grok2, Grok3, and Grok4 and pioneered the STaR algorithm, the first algorithm to train language models for reasoning through self - generated reasoning chains. Currently, Eric Zelikman serves as the CEO of Humans&.

Another core member of the founder team is Noah Goodman, who is a professor of psychology and computer science at Stanford University. He also teaches linguistics and works at the Stanford Institute for Human - Centered AI founded by Fei - Fei Li.

The co - founder team of Humans&, from left to right: Noah Goodman, Andi Peng, Georges Harik, Yuchen He, Eric Zelikman (Source: Humans&)

When Humans& announced its financing, The New York Times also published an earlier interview with the Humans& team.

After leaving xAI in September 2025, Eric Zelikman, the CEO of Humans&, assembled the founding team. He believes that current chatbots are designed to answer questions but are not good at asking questions because the technology doesn't make enough effort to understand what people really want.

Eric Zelikman added: “AI has great potential to enable better human collaboration, but the current ‘question - answer’ paradigm won't take us to that future.”

Andi Peng also left a top AI lab. At Anthropic, Andi Peng was mainly responsible for work related to large - model alignment to prevent AI from lying or harming users' mental health.

But later she gradually realized that there was a bigger problem: like many other AI companies, Anthropic is trying to develop a technology that will systematically replace human labor.

Andi Peng said: “Anthropic is training models to work autonomously. They really like to emphasize that the models can operate independently for eight hours, twenty - four hours, or even fifty hours to complete a task. But that has never been my motivation. I think machines and humans are complementary.”

Georges Harik, another co - founder of Humans&, said that many of the company's investors are real people who care about the future of humanity.

Conclusion: The mass exodus of talent from top AI labs gives rise to sky - high seed - round financing

The seed - round financing of Humans& is astonishing, but it's not an isolated case in the current AI financing environment. Currently, the largest seed - round financing in history is held by Thinking Machines Lab, which completed a $2 billion financing in July last year with a valuation of $12 billion. In addition, the seed - round financings of Unconventional AI and LMArena are also quite substantial.

Although warnings about the AI bubble are widespread, investment institutions are still keen to bet on teams with outstanding resumes and prominent backgrounds. However, a strong founding lineup doesn't necessarily mean sustainable product implementation and long - term competitiveness. We need to continue to monitor how these startups will fulfill their promises in the future.

This article is from the WeChat official account “TechSugar” (ID: zhidxcom), written by Chen Junda and edited by Li Shuiqing. It is published by 36Kr with authorization.