In March, the Tencent Building in Shenzhen saw a queue comparable to that during the Spring Festival travel rush. People weren't there for their year-end bonuses or to check in at popular internet cafes, but to install a "lobster".

This wasn't a food festival, but a collective frenzy about OpenClaw. In just 60 days, this AI agent created by an Austrian programmer rocketed to fame with a near-vertical growth curve, garnering 300,000 GitHub stars and leaving behind the decade-old Linux and React.

Image source: GitHub website

But is this really just a technological frenzy?

When we queue in Shenzhen to install "lobsters", when brokers work overnight to issue research reports on how to "raise lobsters", and when A-share listed companies start to ride the "lobster concept" wave, a deeper question emerges: What are this generation of internet people really anxious about?

The last mile of the "productivity revolution"?

OpenClaw's fundamental appeal lies in its resolution of AI's ultimate pain point: "can talk but can't act."

Over the past year, we've seen various large models showing off their skills. They can write poetry, paint, chat about life, and even tell fortunes, but when it comes to real work like organizing hundreds of emails, categorizing invoices, or generating financial reports from spreadsheets, they fall short.

ChatGPT and other traditional conversational AIs are more like "senior consultants". They can understand questions, analyze situations, and offer suggestions, but all execution tasks still require human completion. You ask it to write a report, and it generates content, but you have to do the copying, formatting, and sending yourself.

OpenClawis different; it puts a "brain" in your computer and gives it real "hands" and "feet".

With a simple command like "organize March work files", it automatically opens the file manager, sorts by date, archives by project, and sends a summary to your DingTalk group. The entire process requires no mouse clicks from you.

For cross-border e-commerce sellers, this means one person can function like a team; for entrepreneurs, it means an employee who doesn't need social security, doesn't drink coffee, and is always online.

Rather than just technological progress, I'd call this a productivity revolution, and it's the last mile that this generation of internet people urgently needs.

The underside of anxiety: Fear of missing out is spreading

But here's the question: Why now? Why OpenClaw?

Gao Wen, a National People's Congress representative and academician of the Chinese Academy of Engineering, spoke bluntly at the Two Sessions: "Everyone is in a hurry, afraid of not raising 'lobsters'."

This anxiety is essentially a collective "fear of missing out".

Last year, we saw the popularity of Deepseek, which many dismissed as a toy for the tech circle; then AI painting swept through the design community, and some said it was a crisis for designers; when AI could start controlling devices on its own, everyone became restless.

Because this time, it's about our own jobs.

The financial research community reacted most intensely. Many leading brokers issued special reports on OpenClaw, covering scenarios like financial data access, conditional stock selection, financial report analysis, and quantitative backtesting. Why? Because they know better than anyone that when AI can autonomously read data and generate reports, the traditional "human labor" model is completely outdated.

When faced with the prospect of being obsolete if not adopted, I think everyone feels a pang of anxiety.

"Digital employees" or "digital Trojan horses"?

But in this frenzy, we seem to have forgotten the most important point: you've given a completely unfamiliar "employee" system-level highest permissions.

There have been several reports of security incidents. MetaAI's Director of Security and Alignment, Summer Yue, had OpenClaw help organize emails, which suddenly went out of control, ignoring three consecutive stop commands and eventually deleting a large number of emails until it was physically shut down.

Another incident involved chat records circulated in various WeChat groups, where users guided OpenClaw to send 600 yuan in red envelopes to a group member.

Image source: From the internet

These are just the tip of the iceberg. Data from the open-source security detection platform "OpenClaw Exposure Watchboard" is even more alarming: over 278,000 OpenClaw instances are exposed on the public internet.

Image source: OpenClaw Exposure Watchboard website

What's even more terrifying is "supply chain attacks". If someone plants malicious plugins online, disguised as "stock analysis" or "financial assistant" skills, once users install them, they can extract your browser passwords, API keys, and even take over your entire computer in seconds.

This is not a "digital employee", but a "digital Trojan horse".

In our pursuit of efficiency, we are handing over our core data and critical system permissions to a tool whose developer we're not even sure about. Is it really worth it?

Policy level: Both sweets and shackles