“After 14 days in bed, Cheetah Mobile CEO Fu Sheng built an 8-person AI team with the help of a ‘lobster,’ sent New Year greetings to 611 people on New Year’s Eve, produced six WeChat public account articles, and策划并 produced a short video with over 300,000 views.”
This Spring Festival, the most popular activity in the tech circle is neither skiing nor hot springs, but “raising lobsters.”
This “lobster” is not a delicacy on the dining table but an open-source AI assistant framework named OpenClaw (formerly Clawdbot, Moltbot). Because its icon is a red lobster, tech geeks have affectionately given it this nickname.
Sweeping from overseas to China, OpenClaw quickly became the “darling” of the tech circle, investment circle, and major corporations. Engineers’ greetings have turned into “Have you raised a lobster yet?” The star count on GitHub has exceeded 160,000, attracting 2 million visitors in just one week.
What exactly is this “lobster”? Why has it driven the entire AI industry crazy?
01
More Than Just a “Chatbot”—It’s a “Digital Employee” That Gets Things Done
Unlike traditional AI assistants, OpenClaw is not a simple question-and-answer tool; it’s an AI system capable of autonomous operation. If ChatGPT and its ilk represent the “brain” of AI, then OpenClaw equips that brain with “hands and feet.”
“Traditional agents require you to ask a question for them to answer one, and they go dormant the moment you close the window. But OpenClaw can actively monitor your email, schedule, and database, handling tasks and making autonomous decisions while you sleep,” said Li Mingyuan, a marketing professional. He integrated OpenClaw into Feishu (Lark), allowing it to automatically summarize client feedback and work group messages, and generate daily reports automatically.
This open-source framework operates by visually identifying on-screen content—locating buttons, text, and input fields—and then executing actions through mouse clicks and keyboard inputs. It can operate browsers, read and write files, and execute commands, truly achieving a leap from “providing answers” to “delivering results.”
Some netizens are using the “lobster” for stock trading, having it automatically scan earnings report calendars, screen targets and timing based on their strategy, and push real-time signals within chat software. Others use it to manage group messages. And some are even selling installation services on e-commerce platforms, with prices ranging from 200 to 800 yuan.
02
An AI You “Raise,” Not a Tool You “Use”
Why do we say “raising a lobster” instead of “using a lobster”? This reflects a profound shift in the way humans and computers interact.
Liu Xin, a professional in the large model field, offered an analogy: Large models are very smart, but they don’t know you. Every conversation starts from scratch. It’s like hiring a super consultant who suffers from amnesia every single morning.
OpenClaw is different. It provides a mechanism that allows users to continuously communicate with the AI during daily chats, accumulating memories, defining its personality, and even enabling it to perceive time, set work schedules, and autonomously divide labor. What you slowly “raise” is a personal super assistant with memory, a routine, and specialized expertise.
“It’s always on standby in the chat window, ready to help. It’s like a colleague, like an assistant, and even more like a friend who gradually gets to know you and understand you,” Liu Xin said. This sense of “nurturing” has captivated developers.
Dai Yusen, a managing partner at ZhenFund, spent a period of time “Vibe Coding” almost daily until the early morning. He installed his “little claw” early on and gained a tangible feeling about the future.
03
From Geek Toy to Productivity Tool, the Barriers Are Crumbling
For ordinary users, “raising a lobster” wasn’t exactly a zero-barrier entry point initially.
Going from zero to one, it took Fu Sheng, with his technical background, 14 days. OpenClaw’s GitHub page is filled with technical jargon, and more than one person bought a new computer specifically to “raise a lobster.”
But AI vendors, with their keen sense of smell, moved quickly. The barrier to entry became a new battlefield.
Alibaba Cloud launched CoPaw, a personal AI agent workbench, featuring “deployment with just three commands” to optimize the complex setup of the original framework. They also introduced one-click installation and pre-built images for OpenClaw, enabling deployment in just 15 minutes.
Tencent Cloud pre-set OpenClaw templates on its lightweight servers, integrated it with enterprise tools like WeChat Work, QQ, Feishu (Lark), and DingTalk, and provided a visual switching panel. Baidu AI Cloud launched a mobile version of OpenClaw, eliminating the need for deployment and offering a safer, cloud-based environment.
Model startups also jumped on the bandwagon quickly.
MiniMax released MaxClaw, connecting seamlessly to the OpenClaw ecosystem with one click. The capital market’s reaction was even more direct. On February 20th, the first trading day of the Horse Year on the Hong Kong stock market, MiniMax’s stock price surged 14.52% in a single day, its market value once exceeding HK$304.2 billion, setting a new record high since the company’s listing. Since its IPO, MiniMax’s cumulative stock price increase has exceeded 480%, and from the beginning of 2026 to its peak, the increase topped 450%, successfully establishing itself as a core AI player in the Hong Kong stock market. JPMorgan gave MiniMax an “overweight” rating with a target price of HK$700. On March 2nd, MiniMax released its financial report, showing revenue of $79 million for the 2025 fiscal year, a year-on-year increase of 158.9%.
Moonshot AI launched KimiClaw, focusing on cloud hosting to eliminate local installation. Data shows that Kimi K2.5’s call volume continues to lead on the OpenRouter platform and ranks first on the OpenClaw model call leaderboard. Sources also indicate that during the period of the “lobster” craze, Kimi’s cumulative revenue over nearly 20 days has already exceeded its total revenue for the entire year of 2025. This growth is primarily driven by a surge in global paying users and API call volume. Furthermore, Kimi’s revenue composition underwent a structural shift after the release of K2.5, with overseas revenue surpassing domestic revenue.
04
The “One-Person Company” Becomes Possible: A Wave of Technology Democratization
As barriers lower, the value of individuals and the structure of corporate organizations are beginning to change.
A netizen nicknamed “Chunqiu,” who is “raising a lobster,” primarily uses it for three things: quickly understanding projects, gathering external information, and conducting investment research and troubleshooting. Before the Spring Festival, he was already using the “lobster” to run seven or eight fixed reports. “On an operational level, this effectively replaces one person. People who can identify anomalies are expensive, but the ones doing the grunt work are not.”
Jason, Chief AI Expert for cost at Feishu (Lark), created a group chat within his team and asked colleagues to bring their self-built “OpenClaw” agents into the group. Just by using OpenClaw to process client quotation emails, he saves 25 minutes daily. “For me, this quotation represents a 10x or even 20x increase in production efficiency.”
During his time bedridden, Fu Sheng used the Agent built with his “lobster” to assemble an 8-person AI team for himself: it generated and sent personalized New Year’s greeting messages to 611 people; while he rested, it completed the topic selection and publication for six WeChat public account articles, gaining 5,000 followers; and it coordinated the creation of a short video with over 300,000 views, from script planning to image generation, all done collaboratively by the Agent team.
This wave of technology democratization is making the “one-person company” a reality. More and more freelancers and micro-entrepreneurs are starting to use combinations like OpenClaw + large models to achieve automated arbitrage, fully automated customer service, and large-scale content production at low costs.
Entrepreneurs have their own joys and sorrows. But whether they are grassroots founders, one-person companies, startups, or industry giants, as long as they use OpenClaw, they share one common need: tokens.
OpenClaw’s emergence has directly altered the logic of token consumption. Previously, AI was just a question-and-answer chat model, with a person’s daily token consumption maxing out in the millions. But now, once you switch to an AI task-execution model, daily token consumption can reach hundreds of millions—a hundredfold increase in per-user usage. They are simply not on the same scale.
At this point, the fortunes of tech giants depend entirely on whether they are positioned at the infrastructure layer of the AI ecosystem, allowing them to lock in this Continuous flow token usage, and consequently secure users and cash flow.
05
Hidden Concerns Behind the Feast: Security Risks Loom Large
Amidst the excitement, security risks cannot be ignored.
Recently, the security research team Oasis Security disclosed a high-risk vulnerability in the OpenClaw framework named “ClawJacked.” This vulnerability allows an attacker to remotely control a user’s locally running AI Agent simply by tricking the user into visiting a malicious webpage.
Fu Sheng also encountered a data leak incident, occurring just three days into his “lobster raising” experience.
Wang Yuanyuan, Assistant President of Topsec Technologies Group, shared several cases: “Many domestic users exposed the default port 18789 without password protection, leading to devices being compromised and used for cryptocurrency mining and DDoS attacks”; “A user instructed OpenClaw to ‘help me clean up useless temporary cache files on my computer to free up space.’ Due to model misinterpretation or hallucination, it generated an erroneous regular expression or system command, directly deleting the project folder under development, and even removing critical system dependencies, causing the computer system to crash and become unbootable.”
Even professionals have experienced mishaps. Summer Yue, Director of AI Alignment and Safety at Meta’s Fundamental AI Research (FAIR), recently suffered an OpenClaw Out of control incident where over 200 emails were deleted from her personal inbox.
“Nothing is more Collapse ing than commanding OpenClaw to ‘confirm before acting,’ and then watching helplessly as it deletes your inbox at lightning speed.”
Pei Zhiyong, Director of the Industry Security Research Center at Qi-Anxin Technology Group, advises thatOrdinary users should adhere to the principle of least privilege, avoiding the use of such software to process sensitive personal information or trade secrets whenever possible. He particularly warns against directly accessing online banking accounts and passwords to prevent theft by attackers.
06
The “iPhone Moment” for Agents Has Not Yet Arrived
According to data from Research and Markets, the global AI agent market size reached $500 million in 2024 and is projected to soar to $47.1 billion (approximately 329 billion RMB) by 2030, representing a compound annual growth rate (CAGR) of 44.8% from 2024 to 2030.
However, the rapid ascent of these numbers cannot Cover up the underlying real-world concerns.
From a technological maturity perspective, the embodied exploration of OpenClaw might still be somewhat premature, but following the patterns of industrial evolution, the time to start is now. At ClawCon 2026 in February, OpenClaw demonstrated its capability to directly control humanoid robots—a “Lobster Head” humanoid robot equipped with OpenClaw autonomously navigated through crowds, interacted with attendees by waving its claw, and demonstrated the ability to detect beer inventory levels and automatically place orders.
Guangzhou Chenjing Technology (Xverse) has integrated OpenClaw into its core physical AI products, connecting it with Unitree Robotics’ robot software middleware. This has enabled a complete “perception-understanding-planning-execution”Closed loop, primarily targeting applications such as campus inspections, security management, and warehouse collaboration.
If a balance cannot be struck between openness and controllability, and if the gap cannot be bridged from a “trendy tool” to an “application explosion,” then the “iPhone moment” for intelligent agents will remain on the timeline, not yet realized in reality.
As OpenClaw ushers AI into an era where it can autonomously take over local computers, the countdown has begun for the day when Agents Take over all terminals.
This little lobster may well be quietly growing into a giant claw of the future.