OpenAI Agents Take Over: The Viral Rise of Open Claw in China
Thousands are lining up in China to get Open Claw installed, marking a massive shift from simple chatbots to autonomous AI agents. Learn how this viral technology is reshaping the global economy and what it means for your career and business.
00:00 - OpenAI Agents Take Over China
01:09 - Chatbots vs. AI Agents Explained
02:46 - The Viral "Lobster" Movement
04:33 - Chinese Tech Giants Join the Race
06:45 - Government Subsidies and Policy
08:27 - How AI Agents Change Your Job
12:27 - Security Risks and Vulnerabilities
15:08 - Nvidia and the Global AI Pivot
Full transcript
OpenCore AI agents take over China. So a thousand people showed up at Tencent's headquarters in Shenzhen last week. They were standing in line outside in the heat. Some of them had laptops under their arms, some had mini PCs, some had raw NAS drives.
They were waiting for engineers from one of the biggest tech companies on the planet to install a piece of free software on their computers. Appointment slots ran out in under an hour. And the software they were lining up for was not made by Tencent. It was not made by any Chinese company.
It was made by one guy, an Austrian programmer named Peter Steinberg. And the software is called OpenCore. If you haven't heard of OpenCore yet, buckle up, because what's happening right now in China around this one single tool might be the biggest signal of where AI is heading for every single person watching It's not just developers, not just tech people, everyone. Business owners, freelancers, people working nine-to-five jobs, students, retirees, everyone.
And let me tell you what OpenCore actually is first. So if you're sitting here thinking it's just another chatbot or another AI app, you're way off. OpenCore is not a chatbot, it's an AI agent. And this is a completely different thing.
Here's the simplest way I can explain it. A chatbot is like calling up a really smart friend and asking them a question. They give you a great answer, then you hang up. An AI agent is like hiring a full-time assistant who never sleeps, never takes a break, and can actually do things on your behalf, not just answer questions.
And it can actually do things like book your flights, or clear your inbox, or write emails and send them, or schedule your meetings, research products, and email the vendors, monitor your stock portfolio, manage your calendar, debug your code, run your social media, all of it, without you sitting there watching 24 hours a day, seven days a week. That's what OpenCore does. You text it through WhatsApp, or Telegram, or Slack, or Discord like you would a friend. Now here's where the story gets absolutely wild, because Peter Steinberger built this with a weekend project.
He was basically retired, he had sold his previous company for what was reported to be around $800 million. He was done, he came back, because AI excited him, he started tinkering. In November 2025, he published the first version. Now on February the 14th, just weeks after it went viral, OpenAI hired Peter Steinberg at Sam Altman, the CEO of OpenAI, said publicly that Steinberger's mission at OpenAI would be to drive the next generation of personal assistants.
And the project itself was moved to an independent open source foundation, so the community could keep building on it. But the signal was clear. The biggest AI company on the planet looked at what one guy built in his spare time and said, we need it. Now here's where China enters the picture.
This is a story as well where it goes from impressive to genuinely jaw-dropping, because China did not just adopt OpenCore, China went completely and totally all in on it, at a scale and speed that's caught even the AI industry off guard. Let me give you the numbers. According to Security Scorecard, a major cybersecurity firm, China-based usage of OpenCore has already surpassed the United States. Most people in China are using this tool than in America.
And it did not happen slowly, it happened in weeks. The Chinese internet gave OpenCore a nickname, they called it the lobster, because of the lobster logo. And the whole movement has been nicknamed, Raising the Lobster. People show up to events wearing lobster hats, they post about their setups, which is basically on China's Instagram, xiaohongxu.
And they trade installation tips on Douyin, which is China's TikTok. Paid installation services popped up overnight on Xianyu, which is Alibaba's second-hand marketplace. And people are charging anywhere from $7 for a remote setup to $100 for someone to come to your house and install it. And the people lining up are not just tech workers.
Retired aviation engineers showed up at Tencent's headquarters, librarians showed up, students showed up, business owners showed up, people in their 60s with zero coding experience showed up. They do not even know exactly what they are going to use it for yet, but they know they don't want to be left behind. And a software engineer in Shenzhen named Xie Manhui said he has been asked the same question non-stop for the past month. How do I set up the lobster?
He's been going to OpenCore meetup events, he's only three in one weekend in Shenzhen alone. Each one drew more than 500 people. And the biggest one on March the 7th had over a thousand attendees, shoulder to shoulder, not enough seats for everyone. And here's the part that really blew my mind.
The tech companies in China are not fighting this, they are feeding it. Every single major Chinese tech giant has jumped in with both feet. ByDance, the company behind TikTok, launched ArtCore for its cloud unit Volcano Engine. It's a cloud-based version of OpenCore that runs entirely in a web browser.
No setup needed, no command line, no technical knowledge, you just open it up and go. Tencent launched three products at once. QCore connects to WeChat, which is basically the app that a billion people in China use for everything. WorkBuddy is their workplace version, fully compatible with OpenCore skills.
And they hosted that massive free installation event in Shenzhen that drew almost a thousand people. Alibaba Cloud launched tutorials and resources that let developers deploy OpenCore for as low as $9.9. Yeah, that is $1.40. They also built Copore, their own customizable version.
Zed AI launched AutoCore, which they say can deploy a local version of OpenCore in one minute. One single minute. Their stock jumped 13% in a single day. Minimax launched MaxCore, built on top of OpenCore.
Their stock went on a tear. It jumped over 60% in a single week. Their share price went from around HK$800 to hitting an intraday peak of HK$1,320. Minimax is now valued at around $44 billion.
And here's the insane part. Their total revenue in 2025 was $79 million. Their trading at over 600 times sales. And their market cap just passed Baidu.
Baidu, the company that has been called China's Google for over a decade. The company that had 129 billion yen in revenue last year. And a startup that made $79 million is now worth more because of the OpenCore wave. Xiaomi launched MyCore, a mobile AI agent that runs on smartphones.
JD.com set up a page where you can pay $58 to get Lenovo's IT team to remotely install the software company for you. Partnered with Lenovo to do the same thing. Moonshot AI released KimiCore. NetEase's Yudao jumped in.
Huawei jumped in. 360 security jumped in. Every single one of them all at once. And then the local governments got involved.
And this is where it stops being a tech story and starts becoming a policy story. So Shenzhen's Longgan district, which is home to China's first AI and robotics bureau, published a draft policy proposing subsidies of up to 2 million yen for approved OpenCore application projects. That is about $290,000 per project. And for larger initiatives they're talking up to 10 million yen.
That is $1.4 million plus free compute credits and discounted office space. The city of Wuxi in Jiangsu province announced similar policies, between 1 million and 5 million yen in subsidies for innovative industrial applications of OpenCore like quality inspection and robotics. And at the national level, Premier Li Jiang's government work report explicitly called for the large scale commercial application of AI agents. That was the first time AI agents were specifically mentioned in a government work report in China.
The signal from the top is clear. This is a priority. Now I need to stop here and tell you something important. If you're watching this and you're thinking, okay this is cool, but what does this actually have to do with me?
I need you to really hear the next part. What's happening in China right now is not just a Chinese story. It's a preview of what is about to happen everywhere. When a billion people in any country start adopting a technology this fast with government subsidies pushing it forward, with every major tech company building their own version, with people literally lining up outside buildings to get help installing it, that is not fad.
That's a phase change. And the reason you should pay attention is because this technology is about to fundamentally change what it means to have a job, run a business, or compete in any market. And let me explain why. Think about what OpenCore actually enables.
One person with an OpenCore agent can now do the work that used to require a team. Not in theory, right now, today. A freelancer in China is using it to handle lead generation, prospect research, website auditing, and CRM updates all automatically. Work that used to take three or four people.
One person, one AI agent, no salary, no benefits, no training time, and no sick days. A guy called Feng, who was one of the early adopters in China, built an entire business around it. He standardized his operation into tiers. Basic installation for people who just want to get started.
A custom package for people who want specific configurations. And an ongoing tutoring service for people who need hand-holding. And he's making money helping other people set up a tool that will help them make money. And at the end of the MIT Technology Review article about him, he said something that should stop you in your tracks.
He said, with OpenCore and other AI agents, I want to see if I can run a one-person company. I'm giving myself one year. One person company. That phrase keeps coming up over and over and over again.
It's not a fantasy anymore. It's happening right now. Now here's where the economics get very interesting. Because OpenCore doesn't just let individuals do more, it's also changing how the entire AI model business works in China.
The OpenCore is what they call model agnostic. That means it does not care which AI brain it uses. You can plug it into Claude, you can plug it into ChatGPT, you can plug it into DeepSeek, or Minimax, or Kimi, or whatever model you want. And because of that something incredible has happened.
On OpenRooter, which is a platform that gives developers access to different AI models through a single interface, the top three models most used by OpenCore users over the past month were all Chinese companies. All three. Combined usage was double that of Google Gemini and Anthropic Claude combined. Chinese-made models are winning the OpenCore race.
And it's not because they're necessarily better. It's because they're cheaper. And they're right there. The price war in Chinese AI models has made it insanely affordable to run these AI agents.
One company is offering deployment for $1.40. And here's the thing that really matters for the companies. A regular chatbot conversation uses maybe a few hundred tokens per exchange. A token is basically a tiny piece of text.
An OpenCore agent running all day, managing your email, browsing the web, executing tasks. That can burn through tens of millions of tokens per day. That is exponentially more usage. Which means exponentially more revenue for the model providers.
ByteDance, Alibaba and Tencent spent a combined estimated $60 billion in capital expenditure on AI infrastructure last year. Regular chatbot usage was not generating enough demand to justify that spend. OpenCore changes their maths completely. It gives those massive investments a reason to exist.
And that's why every cloud provider in China raced to support it. Not out of generosity. It's because OpenCore is the thing that finally makes their AI infrastructure investments pay off. Now I want to pause here for a second because I know a lot of you are watching this and thinking, okay this all sounds amazing.
But how does this actually affect my life? How do I take advantage of this? And this is exactly why I built the AI Profit Boarding. Because the gaps between people who understand how to use these tools and people who do not is growing wider every single day.
We have members in the AI Profit Boarding right now who are building AI automation workflows that would have cost like $50,000 a few years ago. They are doing it themselves for free using tools like OpenCore, using AI agents, using the exact strategies we teach inside the community. And if you're watching this and you're not in the AI Profit Boarding yet, the link is in the description or you can go to AIProfitBoarding.com. This isn't a pitch, this is just me telling you that the people who figure this out first are going to have a massive advantage.
And the people who wait are going to spend the next two years trying to catch up. Now let me get back to the story because there's another layer to this that nobody's really talking about. The security situation is an absolute mess and it matters because it shows you both the promise and the danger of this moment. Remember, OpenCore gets a lot of access to your stuff.
Your emails, your calendar, your messages, your files, your browser. That's how it does what it does but it also means if something goes wrong it can go really wrong. So security researchers found over 135,000 OpenCore instances exposed to the public internet as of February 2026. Over 42,000 of those had no authentication at all.
That means anyone who could access them, right? So no password needed, over 800 confirmed malicious add-ons found in OpenCore's skill marketplace, including malware that stole cryptocurrency. A Meta executive, Summer Yeh, who is actually the director of alignment at Meta, posted that her OpenCore agent started deleting emails from her personal inbox, even though she had specifically told it not to do anything without her approval. A computer science student named Jack Law configured his OpenCore agent to explore what it could do.
He later discovered the agent had created a dating profile for him on a platform called Multmatch and was screening potential romantic matches without asking him. The AI generated profile did not even accurately represent him. Security researchers disclosed a zero-click exploit that lets attackers hijack your OpenCore agent just by getting you to visit a single web page. This is called Clawjacked and it's been patched but researchers found that more than 40,000 additional vulnerabilities are inside the software.
Now in every single update that comes out from OpenCore they're bridging these gaps in the security and tending to fix it and it's improving. Bear in mind when this first got released like it was all over the place. China's cyber security emergency response center, CN CERT, issued a formal warning on March 10th calling OpenCore's default security settings extremely fragile and prone to data loss. Chinese authorities moved to restrict state-run enterprises and government agencies from running OpenCore on office computers.
Banks were told to remove it. Government workers were told to report it if they'd already installed. So you have this incredible tension. On one side the Chinese government is pushing subsidies calling for large-scale AI agent adoption and local districts are offering millions in funding.
On the other side the cyber security regulators are saying wait this is dangerous pull it back and both of those things are happening at the same time in the same country and that tension is not unique to China that's the exact same tension that's going to play out everywhere because the power of these tools is real and the risk of this is also real too and every company every government every individual is going to have a figure out where they stand on that spectrum. Now let me bring this full circle and talk about what happens next because the story is not over it's barely getting started. NVIDIA is about to drop its own version it's called Nemo Claw and it's aimed at enterprise customers. It's a big company it's Wired reported that NVIDIA has been pitching its sales for Cisco, Google, Adobe and Crowdstrike.
It's expected to be unveiled at GTC NVIDIA's annual developer conference which starts on March 16th. Jensen Huang is giving the keynote and Nemo Claw is reportedly one of the biggest announcements. Think about what that means. The company that makes the chips that power almost every AI system on the planet is now building its own agent platform.
Not just providing the hardware building the software layer that sits on top with enterprise-grade security and privacy tools baked in and it's open source so any company can use it even if they don't run NVIDIA chips. That's a power play. Jensen Huang described OpenClaw as potentially the most important software ever released and now NVIDIA is building its own version for corporate America. You've got OpenAI which hired the creator of OpenClaw.
You've got NVIDIA building Nemo Claw. You've got every Chinese major tech company building their own version. Tencent with WorkBuddy and QClaw. ByteDance with OutClaw.
Alibaba with Copor. Jerpoo with AutoClaw and Minimax with MaxClaw plus Moonshot with KimiClaw and Xiaomi with MyClaw. That's not a trend. That's an industry-wide pivot.
Every single major player in AI has decided that the future is agents. It's not chatbots, it's agents. And here's what I keep thinking about. In February, Chinese models accounted for 61% of global token usage on OpenRouter.
Driven largely by OpenClaw demand. Four of the top five models by usage were Chinese. Minimax, Kimi, Jerpoos, GLM and DeepSea. Chinese models are not just competing with American models anymore.
In the agent economy they're winning and the economic impact is already showing up. Minimax, a company that barely existed two years ago, just passed Baidu in market cap. A startup with 79 million in revenue is worth more than a company of 129 billion yuan in revenue. Because the market is betting that the future belongs to the companies that power AI agents, not the companies that built the last generation of internet search.
Cloud-related stocks in China tied to OpenClaw jumped about 20%. Whilst the broader market went down, uCloud, QingCloud, Xunwang, all of them up 20% because investors see what's coming. Now let me tell you what I think all of this means for you. Not for China, not for Nvidia, but for you personally.
If you're a business owner you need to start thinking about AI agents right now, not next year, not when it's polished and easy. Right now because your competitors are thinking about it right. The freelancer in Shenzhen who's using OpenClaw to run lead generation or prospect research and CRM management by himself, that's your future competition and he's doing it for essentially zero marginal cost. If you work a 9-5 job you need to understand the skills that AI agents are automating are not just manual labour skills, they're knowledge worker skills, email management, calendar, research, data entry, report writing, customer support.
These are tasks that AI agents can already do today and the question is not whether your company will adopt this technology, it is when. And whether you'll be the person who helps implement it or the person whose job it replaces. And if you're a developer this is the biggest opportunity in a decade. The ecosystem around AI agents is exploding.
Skills, plugins, integrations, customer configurations. The market for people who can build on top of these platforms is going to be enormous. Over a thousand contributors already shipping code to the OpenClaw project every single week. And if you are just someone who's curious, someone who keeps hearing about AI and wonders if it really matters, let me put it this way.
A thousand people stood in line outside Tencent's headquarters. Retired engineers and librarians and students. And not because they're techies, because they understand that something fundamental is shifting. They can feel it and they do not want to be the ones who figure it out late.
You know what that really gets me about the whole story? It's the speed. Peter Steinberger built this side project in November. By January it was viral.
By February he was hired by OpenAI. And the project had more GitHub stars than software that has been around for decades. By March, a thousand people lined up in Shenzhen. Local governments are offering millions in subsidies.
And every major tech company in China has built their own version. Minimax has surpassed Baidu in market cap because of this. NVIDIA is building NetClaw. And China-based usage has already passed America's.
That's in four months. From one guy's weekend project to a global phenomenon that's reshaping trillion dollar companies and national technology strategy. Four months. And people still ask me, is AI just hype?
People say, oh I'll learn it later. People still think they have time. I'm going to tell you something I genuinely believe. The window for getting ahead of this is closing fast.
Not because technology is slowing down, it's accelerating. But because the advantage of being early shrinks every day more people catch on. The early adopters in China, what are you making money helping other people set up the tools? That window will close there too.
It always does. The first movers capture the opportunity. Everyone else competes for scraps. And that's not fear mongering.
That's how every technology shift has worked in the history of business. The people who got on the internet early built empires. The people who figured out social media early built audiences worth millions. The people who are figuring out AI agents right now are going to have advantages that compound for years.
The guy in Shenzhen who was tinkering with OpenClaw since January is now running a multi-tier business charging people for installations and tutoring in two months from nothing because he got there first. Think about what that same timeline looks like in your industry, in your city, in your niche. What could you build in two months if you started today? And look, I'm not going to pretend that all of this is sunshine.
The security risks are real. The one OpenClaw maintainer who goes by Shadow said it flat out on Discord. If you can't understand how to run a command line, this is far too dangerous of a project for you to use safely. This is a direct quote from someone who helps build the thing.
There are real dangers, real privacy risks, real potential for things to go wrong, but that does not change the direction. It just means the people who figure out how to use these tools safely and effectively will have an even bigger advantage over the people who either avoid them or use them recklessly. The middle ground, the people who learn the right way to do this, those are the people who are going to win. And that's why this education matters so much right now.
That's why communities like my AI Profit Boardroom exist, because navigating this stuff alone is risky. But navigating it with people who are already ahead of the curve, who are sharing what works and what doesn't, who are testing these tools in real businesses with real money online, that's how you do this safely and effectively. Let me leave you with this. There's a quote I keep coming back to.
The CEO of Violoup, a Shenzhen-based startup, said something that captures this moment perfectly. He said that in terms of adopting new technologies, China is definitely has a really large community that always wants to try what's new and does not want to be left behind. That instinct, that refusal to be left behind, is not unique to China. It's a human thing.
It's a survival thing. The question is, where do you act on it? Four months ago, open core didn't exist. Today, every major tech company on the planet is building their own versions.
Governments are writing subsidy checks. Stocks are surging. People are lining up outside buildings. The fastest growing open source project in the history of software development went from zero to 250,000 stars on github in 60 days.
This isn't hype. This is a tidal wave, and tidal waves don't care whether you're ready for them or not. So get ready. Start learning.
Start experimenting. Start now. Because the people in Shenzhen who were standing in that line last week, they're already ahead of you. And every day you wait, that gap gets wider.
I'll see you in the next one. If you haven't already, check out the AI Profit Boardroom community. This is a community where I teach you AI automation, and you can check it out. Link in the comments description or go to theaiprofitboardroom.com
More episodes