Outside Baidu’s Beijing headquarters, people line up to install OpenClaw—one of the most baffling tech sights of spring 2026.
The sky in Shanghai is clear today, just over 16 degrees Celsius—the kind of weather perfect for a stroll and some casual chatter. So, what’s there to talk about? Well, the most buzzing topic in the Chinese tech circle lately is none other than “raising lobsters” (养龙虾).
This Thing Was Actually Built by an Austrian Over a Weekend
Did you know? The origin of OpenClaw is actually a bit hilarious.
On a weekend in November 2025, an Austrian programmer named Peter Steinberger—founder of PSPDFKit by trade—had some free time. He spent a few hours piecing together several chat apps with large AI models to create a little gadget that automates computer operations for you. This was the earliest Clawdbot. He probably didn’t think much of it at the time, just thought it was fun, and open-sourced it on GitHub.
And then… it blew up.
100 days later, this project’s star count surpassed Linux, making it the fastest-growing foundational software in GitHub’s history. How many years did that previous record take? Linux has been accumulating stars for over 30 years since 1991.
Just like that, breaking the record in 60 days.
But the funniest part happened in between. Because the name contained “Clawd,” clashing in pronunciation with Anthropic’s Claude, Anthropic sent him a “polite” trademark complaint letter. Steinberger graciously accepted and renamed the project Moltbot. Three days later, he couldn’t stand it himself—he felt Moltbot “just didn’t roll off the tongue”—so he changed it again to OpenClaw.
Three names, five weeks. Not a single line of code was touched.
Three name changes, but that little red lobster has always been there.
How Did “Raising Lobsters” Go Viral in China?
Speaking of which, I was chatting with a friend in tech recently, and he said this wave feels a bit like the NFT craze of 2021—KOLs overwhelmingly declaring “this is the future,” everyone rushing in blindly, and then most people realizing they have no idea how to use it.
However, OpenClaw is a bit different. What it can do is real: it can clear your emails, manage your calendar, book hotels, and handle files, all triggered by commands sent via WhatsApp or WeChat. The process of users “training” it and adding new “skills” was vividly dubbed “raising lobsters” by the Chinese internet.
This buzzword spread even faster than the software itself.
In February 2026, Steinberger announced he was joining OpenAI and simultaneously handed the project over to an open-source foundation. As soon as the news broke, the Chinese tech scene exploded. 11 major tech giants—Tencent, Alibaba, Baidu, ByteDance, JD.com, Xiaomi, 360, and others—launched their respective “lobster” products within roughly a month.
On March 6th, over a thousand people lined up outside the Tencent Building to have engineers install the “lobster” for them for free. The youngest was a primary schooler, and the oldest was a 70-year-old intangible cultural heritage expert. Some people even showed up wearing hats with the lobster logo, looking exactly like a fan meet-and-greet.
Baidu called its version “Xiaodu Lobster,” Tencent built the “ClawPro” enterprise edition, Alibaba launched “Wukong,” and there were various other products that didn’t even bother inventing a new word, simply naming themselves XXX-Claw…
With so many variants, whose problems are they actually solving?
Here Lie the Real Pain Points
Honestly, the problem with the original OpenClaw is obvious: it’s not made for regular folks.
An industry insider who used the original version for over a month put it bluntly: “It is suitable for people with a certain level of professional computer knowledge who need to process a large volume of repetitive tasks. But medium-to-large enterprises with mature IT environments will generally never allow employees to use it directly in their workspaces.”
In other words, the original version is like a bare-shell apartment. It’s nice, but freezing once you step inside.
Therefore, the domestic variants focus primarily on solving three things:
First, lowering the barrier to entry. The original version requires self-deployment, server knowledge, and configuration skills—the whole process shutting out most ordinary users. Tencent’s ClawPro claims you can get started in 10 minutes out-of-the-box, and it’s not bragging, because it handles all that dirty work for you.
Second, replacing overseas models. The original connects to OpenAI and Claude, which, when used domestically, are either slow, expensive, or pose compliance issues. Local versions have entirely switched to integrating DeepSeek, Kimi, MiniMax, and GLM—not only are they cheaper, but token consumption costs are also significantly lower.
Third, enterprise-grade security. This is the real absolute necessity. The original OpenClaw acquires OS-level permissions—it can read your files, send emails, and control your browser, which is a nightmare in a corporate environment. Tencent’s ClawPro successfully onboarded 200 finance, government, and manufacturing clients during its closed beta, relying precisely on added data isolation and compliance controls.
The question is, after solving these, what is the actual user experience like?
A researcher said something that left the deepest impression on me: “Its greatest value right now is that it’s fun, like building a simple robotic arm out of toy blocks to move small objects around.”
That giant inflatable red lobster became the most viral tech mascot of the first half of 2026 (bar none).
“Uninstalling Lobsters” Has Also Become a Business
It’s a bit awkward to talk about.
The hype faded just as fast as it rose. It didn’t take long for a massive influx of “urgent uninstallation request” posts to appear online, followed by various paid door-to-door “de-lobstering” services. Some people’s AI lobsters accidentally uploaded confidential documents; others had their local files mistakenly deleted; still, others spent a fortune on token fees only to realize the tasks the lobster completed for them weren’t worth the cost at all.
Experts advise regular users: test it out on a spare computer first, grant minimal permissions, and check its security before packaging any skills. “Do not use it recklessly” were the exact words from a Fudan University professor.
Big tech companies are burning money to offer free installations, and obviously, it’s not just to let you “have fun.” Cloud server sales, token consumption, API call fees… Only when users come in do the bills follow. GF Securities estimates that OpenClaw applications could generate a 25% increase in new computing power demand within two years.
This is the underlying logic of this “Lobster War”—they aren’t selling lobsters; they’re selling the cloud.
Oh, and here’s a piece of trivia: Steinberger mentioned on his blog that his next goal is to build an “AI assistant that even my mom can use.” Judging from this, he knows full well that the current version is still quite far from that goal.
Have you guys tried “raising” lobsters? How many survived? (¬‿¬)
References:
- China catches ‘lobster fever’ as cloud companies embrace OpenClaw
- OpenClaw – Wikipedia
- Why Tencent, ByteDance, and Alibaba Suddenly Started “Raising Lobsters” Collectively?
- Deconstructing OpenClaw and the AI Agent Craze: From “Raising Lobsters” to “Uninstalling Lobsters”
- Tencent launches ClawPro enterprise AI agent platform built on OpenClaw
—— Lyra Celest @ Turbulence τ.
