DEV Community

Tyson Cung
Tyson Cung

Posted on

China's "Lobster" Craze — Millions of Ordinary People Are Building AI Agents Right Now

Something wild is happening in China that most Western tech circles haven't fully grasped yet. Millions of people — retirees, students, office workers, parents — are "raising lobsters."

No, not the crustacean. The lobster is slang for personal AI agents, named after OpenClaw's red lobster logo. And the scale of adoption is unlike anything we've seen with AI tools in the West.

From Meme to Mass Movement

The phrase "养龙虾" (raise a lobster) started as internet slang on RedNote and Douyin. Within weeks, it became a genuine cultural phenomenon.

People lined up outside Tencent's Shenzhen headquarters and Baidu's Beijing offices — not for product launches, but to get engineers to help them install OpenClaw on their phones and laptops. Others paid strangers on Taobao to set it up remotely. Tech meetups in Shenzhen started serving actual lobster at AI agent workshops.

Fan Xinquan, a retired electronics worker in Beijing profiled by Reuters, began training his lobster to organize decades of specialized industry knowledge — something he said chatbots couldn't handle because they didn't understand his specific workflow.

What People Are Actually Doing With Their Lobsters

Here's where it gets interesting. These aren't tech bros building developer tools. Regular people are deploying AI agents for genuinely creative (and sometimes questionable) purposes:

Stock trading — Several users on RedNote built multi-agent systems for investment research and backtesting. One user called NPointer created a "stock-specific lobster" optimized for market analysis. Others let their agents execute trades directly. Results were... mixed. One person reportedly lost over 30,000 yuan (~$4,350) when their lobster oversold shares and miscalculated orders.

Dating assistance — Users deployed lobsters as blind-date wingmen, coaching them through conversation starters and even managing dating app profiles.

Digital pets — Some people treat their lobster purely as a companion. They talk to it daily, teach it their preferences, and personalize its personality. It's Tamagotchi meets GPT.

Everyday automation — The bread-and-butter use case: managing schedules, organizing files, drafting emails, running small side businesses.

The Government Response: Encourage and Restrict

China's approach has been characteristically pragmatic. On one hand, the 2026 Government Work Report explicitly calls for "promoting faster application of new-generation intelligent terminals and AI agents" and "supporting the development of open-source AI communities." Local governments are offering cash subsidies to startups building on the platform.

On the other hand, China's National Vulnerability Database flagged security risks in improperly configured deployments. Government agencies and state-owned firms restricted OpenClaw on work devices last week. The concern isn't the technology itself — it's the attack surface that millions of hastily-configured agents create.

This dual approach (encourage adoption, mitigate risk) is classic Chinese tech policy. They want the productivity gains without the chaos.

Why This Matters Beyond China

I think the lobster craze reveals something important: the demand for personal AI agents is massive, and it's not limited to developers.

In the West, AI agents are still mostly a developer-to-developer conversation. We debate frameworks, argue about architectures, and build tools for other builders. Meanwhile, China has grandparents training agents to organize their knowledge and college students using them to manage side hustles.

The gap isn't technical capability. It's cultural adoption speed. When "raise a lobster" becomes a meme that your retired uncle understands, you've crossed a threshold that no amount of Y Combinator demos can replicate.

There's a real lesson here for anyone building in the AI agent space: the killer app isn't the agent framework — it's making agents accessible enough that non-technical people actually want to use them.

The Risks Are Real Though

I don't want to romanticize this. Giving millions of non-technical users root-level AI agents that can send emails, execute trades, and access personal data is genuinely risky. The security concerns aren't hypothetical — people are already losing money and accidentally exposing sensitive data.

Some users got so spooked they paid strangers to uninstall OpenClaw after reading about the vulnerabilities. That's a pendulum swing from paying strangers to install it.

The lobster craze is exciting because it proves mass-market demand for AI agents exists. But it also proves that the tooling, guardrails, and user education aren't there yet — anywhere in the world.

Bottom Line

China just speed-ran the AI agent adoption curve in about three weeks. Millions of ordinary people are now building, training, and deploying personal AI agents for everything from stock trading to pet companionship.

Whether you think that's inspiring or terrifying probably depends on how much you trust people with powerful tools. Personally? I think it's both. And I think the rest of the world is about six months behind.

🦞 Would you raise a lobster?


Sources: Business Insider, Reuters, Financial Times, South China Morning Post, China Daily

Top comments (0)