A couple of days ago, a friend sent me a WeChat message asking: Should you raise a lobster too?

My first thought was—has he switched to selling crayfish?

After clicking, I realized he was talking about OpenClaw. The AI agent that racked up 250,000 GitHub stars in three months and was hailed by Jensen Huang as 'the most important software release of our era.' Overnight, my WeChat feed filled with tutorials, monetization case studies, and one-person company guides—as if not raising a lobster meant getting left behind.

I do content and consulting solo. By today's narrative, I fit the 'one-person company' user profile perfectly. But I have a professional habit—when something goes viral, my first thought isn't 'Should I jump on this?' but 'Who's making money in this chain?'

So instead of installing Lobster immediately, I conducted due diligence on this gold rush.

 1. A Perfect Industrial Chain: Everyone Makes Money Except the Gold Diggers

Let's look at the data.

OpenClaw itself is open-source and free. An Austrian programmer's 'weekend project,' its code sits on GitHub for anyone to download. Founder Peter Steinberger was later poached by OpenAI, and the project was handed over to an open-source foundation.

How do you make money from something free? The answer: It doesn't make money itself, but it enables an entire industrial chain to profit.

Layer 1: Cloud Providers—A Perfect Opportunity to Clear Inventory

OpenClaw needs to run 24/7. Home computers go offline when shut down. The solution? Rent cloud servers. Tencent Cloud, Alibaba Cloud, and Volcano Engine all launched one-click deployment services almost simultaneously. Tencent even sent staff to set up booths for free installations.

This isn't charity. For the past year, inference-side computing power consumption had been sluggish, leaving low-spec lightweight servers gathering dust in warehouses. OpenClaw acts as a token shredder—a single task consumes hundreds of thousands or even millions of tokens, perfectly helping cloud providers clear inventory.

Layer 2: Large Model Companies—Finally Found a C-Side Consumption Scenario

OpenClaw doesn't come with a brain; it needs to connect to large models to function. The first thing users do after setting up Lobster is buy an API key. MiniMax launched MaxClaw with zero-configuration, zero-API-fee one-click deployment; Moonshot AI released Kimi Claw, a cloud version ready to use out of the box.

During the Chinese New Year red envelope wars, companies spent billions to acquire users who uninstalled apps after claiming bonuses. But OpenClaw's agent logic is different—it runs continuously, consuming resources nonstop. Using an open-source community project to drive model API calls for your own company is a cost-effective (cost-effective) deal by any measure.

Layer 3: Installation Services—The New 'Broadband Installation'

Search 'OpenClaw proxy installation' on Xianyu (China's Craigslist), and you'll find remote services for 100-200 RMB or on-site installations for 500-1,000 RMB. A team called SetupClaw claims to have hit $1 million in annual revenue. On the day Tencent offered free installations, queues stretched across exhibition booths.

Even more exaggeration (over-the-top) are the courses and training camps. 'Build a one-person company with OpenClaw,' 'Lobster generates daily gold'—sound familiar? This playbook matches exactly with tutorials promising 'open a Taobao store and earn 100,000 RMB monthly' from years past.

Layer 4: Government Subsidies—Even Policies Keep Pace

On March 8, Shenzhen's Longgang District released a draft proposal offering up to 2 million RMB in subsidies for OpenClaw-related projects, with application demonstration projects receiving 30% of actual investment as rewards (capped at 1 million RMB). Civil servants in Shenzhen's Futian District already use 'government affairs lobsters' to handle public complaints.

You see how perfect this chain is—cloud providers sell servers, large model companies sell tokens, installation teams sell services, training institutions sell courses, and local governments sell policies. Everyone makes money.

Except the person who installed Lobster, planning to run a one-person company. They type a command, and Lobster replies: 'I don't currently have browser control capabilities for you.'

I've seen this structure many times. A concept goes viral, forming a complete service chain where every link profits—except the consumer, entrepreneur, or follower who ultimately pays for the concept, often earning the thinnest margins.

In gold rushes, the richest are never the miners but the shovel sellers. This isn't new, but it rebrands every few years.

 2. 'One-Person Company' Isn't New—Just a Repackaged Old Story

OpenClaw's most compelling narrative isn't its technology but the claim: 'Future companies might have just two employees—you and your AI cluster.'

Honestly, that line resonated with me at first. As a solo content creator and consultant, I know how exhausting running everything alone can be. If a 24/7 AI could crunch data, draft content, and organize client materials, my ceiling would rise significantly.

But then my professional habit kicked in again—I started recalling where I'd heard this story before.

Multiple times.

In 2015, during the 'mass entrepreneurship and innovation' era, the narrative was: 'The internet lets everyone become an entrepreneur.' Co-working spaces popped up everywhere with generous subsidies. On-demand beauty, massage, and car washes flooded screens as 'next unicorns.'

What happened next? A certain bike-sharing company still hasn't refunded deposits, and countless O2O projects vanished.

In 2020, live-stream commerce arrived. The story became: 'Everyone's a streamer; your phone is your store.' Suddenly, everyone went live, with training courses flooding the market.

What happened next? Headstreamers became increasingly concentrated, while 99% of participants just fed platform traffic.

Now it's 2026, with OpenClaw-powered one-person companies. The narrative upgrades: 'AI lets everyone start a company—no employees, no office, just one lobster handles everything.'

Spot the pattern? Every hype cycle follows the same script: New tool emerges → Lowers some barrier → Someone declares 'everyone can XXX' → Training/services profit first → Policies push it further → Most followers realize the tool is great, but they don't know what to do with it.

The problem never lies with the tool.

The problem is: Do you have something worth amplifying with this tool?

3. What Has OpenClaw Actually Changed? What Hasn't It Changed?

After all this, I'm not here to bash OpenClaw. It's genuinely important.

What it changed is clear: AI shifted from chatting to working. Previous ChatGPT versions resembled a 'brain in a jar'—smart but trapped in dialog boxes. OpenClaw gave AI 'hands and feet,' enabling file read and write (read/write), browser operation (operation), email handling, and script running. This marks a paradigm shift.

But what it hasn't changed is equally clear:

First, it hasn't lowered the barrier of 'knowing what to do.'

OpenClaw's safety pass rate is only 58.9%, with a 0% pass rate in 'intent misunderstanding and unsafe assumptions.' Translation: If your instructions are slightly vague, it fills in gaps and executes recklessly. Meta's AI safety director asked it to manage emails; it started mass-deleting messages. Three 'Stop' commands didn't work—only unplugging the power halted it.

This scenario feels familiar—like assigning tasks to a highly capable but business-illiterate intern. If you haven't clarified your needs, their stronger execution amplifies disasters.

Second, it hasn't lowered the barrier to 'sustained profitability.'

Using OpenClaw to auto-upload products to an independent store seems ideal. But what if suppliers run out of stock? Platforms introduce new CAPTCHAs? Exchange rates erase profits? Each 'what if' requires human solutions—from someone who understands the business.

That 36Kr article put it harshly but accurately: Most people think they're running companies but have just used AI to become more efficient wage workers. You yourself become the system's biggest bottleneck.

Third, it hasn't changed fundamental business laws.

AI can write code, design graphics, edit videos, and run automation. But it can't answer the core question: What do customers actually need?

Real customer needs form the foundation of all sustainable businesses.

4. What's the Real Moat for a 'One-Person Company'?

I've been a solo media creator and independent consultant for nearly six months. Honestly, I'm running this whole operation alone.

But what keeps me afloat isn't any AI tool.

It's the industry judgment I built over a decade in investment banking. It's clients saying, 'Talking to Linda changes how I view problems.' It's the data cross-validation and logical consistency behind every article. It's people paying for my cognition, not AI-generated reports.

OpenClaw can't give you these things.

I've seen many people invert the causality—they think tools enable company ownership, but actually, you need company-owning capabilities first for tools to matter.

Like a sharp knife doesn't make you a chef, but a good chef with a sharp knife can cook better meals.

OpenClaw is a sharp knife. But you need to confirm you have ingredients to cut.

5. Final Thoughts

Back to the original question—my friend asked if I should raise a lobster.

My answer: I'll install it. But not urgently.

I'll wait for early adopters to iron out kinks, for safety issues to resolve, for cloud versions to mature—then integrate it into my workflow for repetitive tasks like data organization, information retrieval, and draft generation.

But I'm crystal clear: It's a tool, not the business itself.

If you're hesitating about jumping on this trend, ask yourself three questions first:

Do you have a validated, client-paid skill? Do you have a clear business process AI can amplify? Do you have a business model that survives without AI but thrives with it?

If all three are 'yes,' congratulations—OpenClaw is a sharp knife for you. Go use it.

If all three are 'no,' you don't need a lobster. You need to figure out what 'dish' you want to cook.

Tools multiply efficiency by 1,000x. But 0 × 1,000 is still 0.

In every tech wave, survivors aren't the fastest runners but those who know their destination.