Financial Observation: China's "Hundred Shrimp Compete," Is the Era of "One-Person Companies" Coming?

【Source: Global Times】

【Global Times report, Reporter Yang Shasha】“In the AI era, between 2026 and 2028, a ‘one-person unicorn company’ with a $1 billion valuation will emerge.” A prediction made by OpenAI CEO Sam Altman at the beginning of 2024 is now being acceleratedly fulfilled, thanks to a “lobster.” This “lobster” is a stand-in for the intelligent agent OpenClaw—it allows AI to operate a computer and carry out tasks like a human, making it possible for one person to manage a team and run a company. According to The Wall Street Journal, the “lobster”’s creator—Austrian engineer Steinböck, has now been hired by OpenAI with a high salary. This developer, who has almost single-handedly ignited the “one-person company” (One Person Company, abbreviated OPC) model, is pushing this change toward even farther-reaching territory. In China, a startup wave driven by “lobsters” has also already begun. From policy support to big players entering the fray, “one-person companies” are moving from concept to reality.

A policy-driven upsurge in entrepreneurship

“The single-person entrepreneurship era has arrived in China.” Hong Kong’s South China Morning Post reported that, driven by OpenClaw’s explosive growth and a series of supportive policies aimed at nurturing “one-person companies,” China is seeing a new wave of entrepreneurship. Guangdong Province was the first to roll out a comprehensive action plan, becoming the first provincial-level government to support AI-enabled “one-person companies.” Hubei Province followed suit, with both regions providing resources to support computing power and Token (tokens) output. Multiple municipal and district-level government departments in Shanghai, Wuhan, Qingdao, and other places have also issued similar policies.

Chinese tech companies responded quickly. After the 2026 Spring Festival, several big tech firms released large numbers of OpenClaw-like products at an unusually fast pace, a development industry insiders call “a hundred shrimp competing in the race.” Baidu was the first to launch “a mobile version of OpenClaw,” RedClaw, allowing users to command AI to handle tasks on their phones, and then also rolled out DuClaw with zero-deployment services. This month, Feishu’s AI work platform under ByteDance launched and upgraded multiple intelligent-agent products, including Feishu’s “lobster” Feishuaily intelligent partner, Feishu Miaoda, and more, targeting office/workplace users. Alibaba, Tencent, Huawei, Xiaomi, Zhipu, MiniMax, the Dark Matter of the Moon, NetEase, and others have also all released “lobster” products for local deployment or cloud deployment.

What’s the difference between domestic intelligent agents and foreign products? Tech investment blogger Bao Kun, who has 3 million followers across the whole internet, told reporters from Global Times that he has tried intelligent agents such as Tencent WorkBuddy, Feishu aily, and Alibaba’s “Wukong.” “These domestic intelligent agents are integrated really well, and their security is also guaranteed. But on convenience you have to give up something, because most of them are deployed in the cloud rather than locally. And OpenClaw’s localized deployment is one of its biggest features.”

Whether it’s a domestic “lobster” or an overseas “lobster,” at its core it’s constrained by the “impossible triangle”—it’s hard for one product to be cheap, secure, and convenient all at once. Bao Kun explained that security and convenience are often in conflict, so domestic “lobsters” lean more toward security, sacrificing some convenience. But no matter which platform you use, Token still has to be consumed behind the scenes. Many big companies now install “lobsters” for free; in essence, it’s like a water company offering to lay pipes for free, but you still have to pay the water bill yourself. Token is the “water” that the “lobster” intelligent agent must consume— the more complex the usage, the higher the water bill.

How “one-person companies” operate

“Today, I’ve gotten used to having the AI do a round of work for me first.” Multiple product managers at Feishu, when demonstrating how intelligent agents work to media outlets including Global Times, all said that AI can help you “get things done like a colleague.”

“The most impressive thing about aily is that it can learn your way of working.” A Feishu product manager demonstrated how to teach the AI a working method: with the goal of “interviewing a CEO,” like teaching a real colleague, get aily to learn how to research the background, come up with questions, and organize an interview handbook. After it’s taught, aily is asked to “turn this process into a skill.” The product manager interpreted it this way: “In the future, if you run into a similar interview scenario, aily will automatically prepare the background research, question design, and interview techniques all based on the approach you taught it, and the output format will be completely consistent with what you had before.” Through repeated training, aily will consolidate memory—becoming smarter the more you use it, moving from “you make it do it” to “it helps you do it proactively.”

In Baidu’s ecosystem, many developers have already used intelligent agents to realize one-person entrepreneurship. Wang Wei, technical head of Baidu Miaoda, shared a typical case with reporters from Global Times: an entrepreneur who came out of a big tech firm built a fully automated business process by combining OpenClaw and Baidu Miaoda. First, they used the Miaoda platform to generate sentiment analysis and AI-assisted application-related outputs via natural language. Then, OpenClaw served as an autonomous AI agent, calling external platform APIs to automatically complete application publishing, listing, and basic operational actions, forming an automated closed loop from generation to monetization. There are many similar cases as well—for example, individual developers making games with Miaoda, distributing and monetizing through private-domain channels or among friends, which has become a common phenomenon.

Huang Rongsheng, chief architect at Xiaodu Technology, said that currently all software is designed for “people,” and the APIs (application programming interfaces), MCP (model context protocol), and skills designed for intelligent agents are still scarce. Huang Rongsheng said, “Building capabilities for intelligent agents is an excellent entry point for entrepreneurship. Many designers and front-end developers already have successful cases.” Zhou Hongyi, founder of 360 Group, also raised a similar view: “All software, all products on the internet must be reworked using the thinking behind intelligent agents.”

AI is more like an amplifier, infinitely multiplying your abilities

If “lobsters” hadn’t appeared, the Chinese market likely wouldn’t have entered the intelligent-agent era so quickly. Huang Rongsheng said candidly, “I need to solve a huge trust problem. Would users be willing to give Xiaodu their email authorization code? I think that’s unrealistic. But they would be willing to give their own ‘lobster,’ because it’s their own digital asset.” He believes this is not just a technical issue, but how to combine consumer habits and industry ecosystems to form a consensus.

The founding of “one-person companies” in the U.S. is also drawing a lot of attention. Tailor Brands, a U.S. business services company, recently cited U.S. Census data, stating that in November 2025, about 535,000 business applications were submitted— the highest monthly total in the past three years. Among all small businesses in the U.S., 85.8% are “one-person businesses” with no employees, and 55% are home-based businesses. Other data shows that in the first half of 2025, the share of “one-person company” startups in the U.S. was 36%, a proportion that increased by 53% over six years.

Bao Kun now doesn’t want to sleep every evening; he gets excited early in the morning. “AI hasn’t made me lazier—it’s made me busier and more fulfilled. It’s a tool, but more like an amplifier that infinitely multiplies your abilities.” He said, “What makes us most happy is that this generation has accumulated management experience, professional knowledge, and data resources, and we also happened to catch the AI boom—talk about getting a boost at just the right time. This is the times. One generation can only solve the problems of one generation—we can only move forward.”