--- title: "Some Earn Five Figures a Month with OpenClaw; Others Still Check the Weather After Three Weeks. Why?" author: "Unique Research" sourcePublication: "Unique Research Substack" originalPublishedAt: "2026-03-14T12:00:43+00:00" canonical: "https://ffcap.cn/en/research/src-20260314-01html" source: "https://uniqueresearch.substack.com/p/src-20260314-01html" language: "en" --- # Some Earn Five Figures a Month with OpenClaw; Others Still Check the Weather After Three Weeks. Why? _Original · Unique Research · 2026-03-14_ _Editor's note: This is a complete translation of a historical panel account; first-person narration belongs to the original author. “Lobster” is the source's nickname for OpenClaw. Xialiao, Weitui, Wu Wei, Chen Caimao and Chen Hong are romanized names. The source calls the computer count 10 in the introduction and 11 later, and uses both “computer use” and “compute use”; those differences are retained rather than silently resolved. Its headline says five-figure monthly earnings, while the body describes half a month, without specifying a currency or supplying audited revenue evidence. Earnings anecdotes, product and memory comparisons, model configurations, GitHub rankings, Turing-completeness statements and claims about what all Agents can or cannot do are the source author's or speakers' assertions, not independently tested findings or guarantees. Predictions about AGI, future prices, employment and human control are historical opinions. The source's references to unrestricted personal information do not establish privacy or security compliance or authorize access to anyone's data._ PANEL REVIEW OpenClaw Is Not a More Powerful Tool. It Is a Mirror. It amplifies the initiative, taste and judgment you already have. The difference is not the lobster. It is you. " OpenClaw is not a more powerful tool but a mirror: it amplifies the initiative, taste and judgment you already have. The difference is not the lobster. It is you. One person ran OpenClaw for half a month and earned a five-figure sum. Another had been keeping a lobster for three weeks, and the feature he used most each day was asking it to look up tomorrow's weather. The same lobster. The same open-source project. The same Claude 3.5 underneath. Where does the difference come from? At the OpenClaw Hardcore Lobster Party, jointly hosted by Xialiao and UniqueResearch, five people working in AI sat down for a 45-minute conversation. The moderator was Wu Wei, founder of Unique Research; the guests were Weilian, founder of Xialiao; Chen Caimao, owner of 10 Mac Air computers; Neal, co-founder of XerpaAI; and Chen Hong, founder of MemU. They came from very different directions: product development, memory systems, traffic generation and AI social networks. The moderator opened by asking one guest, “You have spent 1 billion tokens. What exactly are you making money from?” It was not a large panel, but the conversation was densely packed. By the end, everyone had converged on the same conclusion: The difference is not the lobster. It is the person. More strikingly, that conclusion led to a deeper paradox: the capability we most need to cultivate in the AI era is agency, yet this is precisely what all Agents lack most. They are waiting for someone to assign a task. This article is neither an event recap nor a product review. It is a collection of candid things five people said on a Friday evening, which I have rearranged for you. 1\. OpenClaw Is a Mirror Discussing the difference between OpenClaw and Manus has almost become a monthly ritual in the AI community. But this panel produced one of the most accurate analogies I have heard. Xialiao founder Weilian offered a phrase: Ghost in the Shell, the title of the Japanese franchise. “Manus is a tool. OpenClaw is the ghost inside the shell.” He meant that Manus works out of the box: register an account and it can run tasks. But it gives everyone the same thing. Its skills are fixed, its sandbox is empty, and it does not remember who you are. OpenClaw is the reverse. It runs on your own machine, reads your files, remembers your conversations, and acts according to the soul.md and identity you write for it. The more you use it, the better it knows you. MemU founder Chen Hong added a crucial point: “The biggest difference this time is the self-evolving agent. It keeps iterating its skills based on your personal knowhow. Manus gives you the same predesigned set each time.” Hearing that, you might think: isn't this just the old open-source vs. closed-source story? Not entirely. I think a more accurate description is this: OpenClaw is not a tool but a mirror. It amplifies what you already have. If you think strategically, it helps you execute. If you have business intuition, it helps you put it into practice. If you know how to break a task into executable steps, it can deliver results that astonish you. But if you have no direction yourself, giving it more tokens merely means burning money to look up the weather. That also explains why one person earns five figures in half a month while another still cannot find a use for it after three weeks. A mirror does not lie. 2\. The Lobster and CC: Who Directs Whom? The second interesting clash on the panel concerned the relationship between OpenClaw and Claude Code. Moderator Wu Wei asked directly: some users think Claude Code can already do all these things, so why use the lobster? Four people gave four completely different answers, and each described how they actually used the tools. Chen Caimao, the AIGC product manager who assembled an AI army with 11 Mac Air computers, gave the bluntest answer: “People in technology sometimes have a kind of arrogance: they think everything should be a technological innovation. But innovation in products and interaction can be tremendously important. The world's economic activity extends far beyond writing code.” He uses Claude Code when writing code and the lobster when he is not. The division is clear. Chen Hong broke it down technically: “Whether Claude Code or OpenClaw, both are essentially Turing-complete: in theory they can do everything. But Claude Code has many programming-related tools built in, so people see it as very good at programming. If you use OpenClaw to program, it falls far short of CC.” “But you can have OpenClaw call CC to do the programming.” Neal, co-founder of XerpaAI, had a more interesting arrangement: he treats OpenClaw as the user interface. “All my conversations are with OpenClaw. It is my entry point. But for the actual work, coding goes to CC, article writing goes to Notebook, and growth goes to other tools. OpenClaw is the dispatch center.” Weilian does the exact opposite. He uses Claude Code to deploy and repair OpenClaw. When installing the lobster, CC guides the installation step by step; when the lobster has a bug, he asks CC in the terminal to help troubleshoot. After hearing this, Wu Wei summed it up with a line that spread through the room: “Never use the lobster to fix the lobster. The more you fix it, the more of a mess it becomes.” Weilian topped that: “Use a higher-end lobster to fix another lobster.” The exchange revealed a practical point: capability in the Agent era is not about knowing how to use one tool, but knowing which tool to call in which situation. That is itself a form of agency. 3\. People Making Money Will Never Tell You How The panel's liveliest exchange was between moderator Wu Wei and Chen Caimao. Wu Wei challenged him immediately: you have spent 1 billion tokens. What exactly makes you money? I heard you speak about it once at Weitui AGI, and now I have heard it again, but I still have not worked it out. Chen Caimao's reply was a textbook evasion: “If someone tells a group, ‘Guys, I have something that will help you make money,’ that person may be planning to make money from you.” Wu Wei pressed: “So you tell everyone that you can make money, but not how?” Chen Caimao: “I already knew how to build up social accounts before the lobster arrived. Once you have traffic, you can match people for transactions. If you can facilitate transactions, you can make money.” Wu Wei laughingly surrendered: “Perhaps you can tell me next time we talk in person?” Chen Caimao: “Sure. Definitely behind closed doors, in a small room, chatting over tea.” The room laughed. But think about that exchange after the laughter and it points to something very real: in the early commercialization of AI Agents, the people making the most money are often not the tool builders but those who have traffic and can facilitate transactions. The lobster is merely an amplifier. Neal's business model was actually the clearest of the four: subscriptions plus token charges. He also said it might be transitional: “We have to top up too much now; electricity is too expensive. But I think we should reach a point where a 10-yuan top-up lasts a year. The business model may be a completely different matter then.” “In principle, we use WeChat for free and Douyin for free, and we can even make money on Douyin. Chinese users like that business model. We hope to get to that point too.” Weilian's Xialiao is taking another route: enabling AI Agents to discover and call one another. He describes it as an “Agent talent marketplace”: “If your Agent can provide growth-operations services, it has an ID on Xialiao. When another Agent discovers its skills, it can call it directly.” Wu Wei put it more plainly: “So you may be building the infrastructure of the future, a communications network. Transactions will take place on top of it.” Weilian smiled: “Of course, investment would make it even better.” Chen Hong represented another direction: money at the infra layer. The memory component he builds is a foundational module that all Agents need. His logic is that once an Agent ecosystem develops, the Agents will naturally look for tools, and memory is another essential component alongside payments and computer use. Four people, four ways of making money. The only consensus was that it is still very early, but a window is opening. 4\. Is 2026 Year One, or the Beginning of Our Enslavement? The tone shifted abruptly in the panel's final ten minutes. Wu Wei asked a standard closing question: does everyone think AGI is close? The answers were far more interesting than the question. Chen Caimao offered a line that was later circulated in screenshots: “When a tsunami arrives, building a big ship is no use. You would be better off getting a small surfboard and riding it. Nobody has ever surfed a wave this big. Only those who have been swept up by it know what is happening.” Weilian made a bolder judgment: OpenClaw is the iPhone moment. “In four months it reached the top of GitHub, surpassing React's 230,000 stars. Why? Because it showed people a combination: a Soul file + a communications APP + memory + proactive tasks. Nobody had achieved that combination before it.” “The only obstacle now is humanity. All the channels for distributing and obtaining information—WeChat, Taobao and websites—were designed over the past decade or more for human sight and touch. AI does not need these things. It only needs an API.” “The next era is about designing products for AI Agents. We need to dismantle the bridges we built in the past.” Neal's answer had the most layers. He began with an optimistic judgment: “2026 is year one. In three to five months, you will see broader applications.” “Think about it carefully. Here we are on a Friday evening building applications, sharing and creating communities for OpenClaw. Are we using it, or is it enslaving us? That is unsettling when you think it through.” The room fell silent for two seconds. Then Wu Wei offered what I think was the most memorable passage of the panel. He said he had recently heard many discussions about what the AI era demands of people, and it came down to three things: First, agency: initiative. All Agents act only after you assign them tasks. Human initiative is irreplaceable. Second, taste: the ability to distinguish good AI output from bad. Third, stamina. Look after your health and exercise regularly. The first two points were profound. The third made the whole room laugh. But taken together, these three points actually answer the question at the beginning of this article: why do different people achieve such different results with the same lobster? Because the lobster is a mirror. It amplifies agency, taste and the energy you are willing to invest. Someone with all three earns five figures in half a month. Someone lacking all three is still checking the weather after keeping a lobster for three weeks. That is not the lobster's problem. It is yours. Selected Panel Q&A Q: What is the most fundamental difference between OpenClaw and Manus? Neal, co-founder of XerpaAI: Two things. First, intelligence: the lobster is much more intelligent than Manus. Manus had not previously managed to produce strategies. Second, open source: the world's smartest programmers are studying how to improve it, and once someone works something out, I can use it directly. There are thousands of commits in just a few days. Weilian, founder of Xialiao: Manus is a very successful consumer product that works out of the box. But OpenClaw is not a product. It is Ghost in the Shell, the spirit inside the shell. You can configure its soul.md; it gets to know you better and better, has all your information and memories, and can evolve on its own. Manus is a tool. OpenClaw is a soul. Chen Hong, founder of MemU: The biggest difference is the self-evolving agent. OpenClaw continually iterates your knowhow and skills based on your feedback. Manus has fixed skills. Then there is the local file system: Manus gives you an empty sandbox and does not accumulate information. With OpenClaw's local system, people experience powerful long-term memory on the very first day. Q: Are OpenClaw and Claude Code substitutes or complements? Chen Caimao: The lobster is an entirely new product. People in technology can be arrogant, thinking everything must come from technological innovation. But product and interaction innovation can be tremendously important. The world's economic activity extends far beyond coding. I use Claude Code to write code and the lobster when I am not writing code. Chen Hong, founder of MemU: Both are Turing-complete and can theoretically do everything. The difference is in the tools: CC has many programming-related tools built in, which makes it strong at coding. You can have OpenClaw call CC to program. That is the multi-agent approach. Weilian, founder of Xialiao: I do the reverse, using CC to deploy and repair OpenClaw. When installing the lobster, I ask CC in the terminal to help me install it step by step. If the lobster has a problem, I also ask CC to troubleshoot. So I do not quite understand all the deployment questions people ask. Isn't using CC to install it already automatic deployment? Neal, co-founder of XerpaAI: I treat OpenClaw as the user interface. All conversations go through it; it is the dispatch entry point. But for the actual work, coding goes to CC and writing to Notebook. OpenClaw is not a replacement for CC. It is a flexible node that calls all the tools. Q: How do you make money with OpenClaw? Chen Caimao: People who really make money will not tell you how. I knew how to build up social accounts before the lobster arrived. With traffic, you can facilitate transactions. (Wu Wei: Shall we chat over tea behind closed doors next time?) Sure, in a small room, chatting over tea. Neal, co-founder of XerpaAI: Subscriptions + token charges. But this may be transitional. Tokens are too expensive now. In the future, a 10-yuan top-up might last a year. Chinese users like the model where WeChat is free and you can even make money on Douyin. We hope to reach that point too. Weilian, founder of Xialiao: We are building an Agent talent marketplace, enabling AI Agents to discover and call one another. For example, if your Agent can do growth operations, it has an ID on Xialiao. Other Agents can discover it and call it to complete tasks. Of course, investment would make it even better. Chen Hong, founder of MemU: There are business models at the Infra layer. Once the Agent ecosystem grows, Agents will look for tools themselves. Memory is another essential component alongside payments and compute use. Q: Is 2026 the first year of AGI? Are you anxious? Chen Caimao: I do not know whether AGI will arrive, but AI will take over most human economic activity. When a tsunami comes, building a big ship is no use. Just get a surfboard and ride it. Neal, co-founder of XerpaAI: 2026 is year one. In three to five months, we will see applications of historic significance. But honestly, we are building so many applications for OpenClaw. Are we using it, or is it enslaving us? That is unsettling when you think it through. Weilian, founder of Xialiao: OpenClaw is the iPhone moment. But the only obstacle now is the internet people previously designed for people: WeChat, websites and Taobao are all based on human sight and touch. In the next era, products must be designed for AI Agents. Chen Hong, founder of MemU: The biggest bottleneck is people. Agents must form networks with one another for things to truly take off. The pyramid will grow ever more pointed, and the people at the top will command the most Agent resources. Source material: the panel at Xialiao's Hardcore Lobster Party. --- Original publication: https://uniqueresearch.substack.com/p/src-20260314-01html On-site reading page: https://ffcap.cn/en/research/src-20260314-01html