The hype around "lobster farming" (meaning "using OpenClaw or something alike") has finally subsided.
Previously at the Morgan Stanley conference, NVIDIA CEO Jensen Huang spoke highly of OpenClaw, saying, "OpenClaw is probably the single most important release of software, you know, probably ever. If you look at OpenClaw and the adoption of it, Linux took some 30 years to reach this level. OpenClaw, in 3 weeks, has now surpassed Linux."
OpenClaw's defining feature is its ability to connect large language models with local systems, enabling cross-application automated execution and feedback. This selling point has thrilled internet giants, as not everyone is a tech geek capable of handling local deployment. Selling cloud-based deployment subscriptions not only helps amortize the massive upfront capital expenditure on AI but also allows companies to pitch an "AI productivity" narrative to the capital markets.
Around the Spring Festival, amid enthusiastic coverage from tech influencers, OpenClaw became a hot topic in China's internet circles.
It must be said that these course-selling gurus truly know how to capture internet traffic. Open-source and one-click deployment features were marketed as "free to use" and "zero barrier to entry." The technical principles behind AI agents were glossed over, and the high-level permissions required were downplayed, while one-sided emphasis was placed on "high business returns" and "AI dividends." These targeted talking points fueled AI anxiety among ordinary users and allowed for a significant harvest of "course fees."
But now, "if you're slow to learn, you don't need to learn at all." In less than a month, hardly any ordinary people were still talking about OpenClaw, because they had no use for it.
What truly ignited the "lobster farming" craze in China was Tencent.
Founder Pony Ma promoted the lobster on his WeChat Moments, and other companies like Alibaba, ByteDance, and Xiaomi quickly also rolled out products and services adapted for OpenClaw. Events offering free deployment assistance were held across the country. Major tech firms used this opportunity to deeply integrate users into their own ecosystems, aiming for paid conversions.
With frequent coverage from major media outlets, OpenClaw rapidly broke into the mainstream in just one week. According to official data from Tencent Cloud, the number of cloud-based "lobster farmers" exceeded 100,000.
As the hype grew, Tencent launched SkillHub, a localized mirror platform for OpenClaw, synchronizing a large number of official skills from ClawHub overseas. This drew dissatisfaction from OpenClaw creator Peter Steinberger, who complained on social media that Tencent had increased the load on servers without providing corresponding support to OpenClaw in return for the surge in attention.
Tencent explained that Tencent Cloud independently bore the vast majority of SkillHub's traffic, and it was not a case of "taking without contributing." At the same time, Tencent officially joined the OpenClaw community as a sponsor, providing substantive support through funding, computing power, and ecosystem resources. Steinberger later described the resolution as "a good redemption arc."
By design, OpenClaw is a tool aimed at tech enthusiasts. Deployment involves specialized steps such as environment configuration and API key management, and its use is far from free—it operates on a cumulative billing model based on "input tokens + output tokens," with both implicit and explicit costs being considerable.
As a result, although OpenClaw garnered public attention overseas, it remained largely within tech communities like GitHub. Huang's high praise was based on the fact that the implementation of AI agents validated the need for building large-scale AI infrastructure, which would bring more predictable demand for NVIDIA's products.
In contrast, Tencent's effort to turn "lobster farming" into a nationwide phenomenon reflected a different line of thinking.
From a logical perspective, the rapid progression of "lobster farming" was clearly not an unplanned development. Even though the AI landscape changes rapidly, such a comprehensive product lineup could not have been improvised.
OpenClaw seemed to serve as a convenient vehicle for Tencent. Its open-source nature allowed Tencent to naturally reference its architecture to develop its own "comparable" products, and the concept of AI agents aligned perfectly with Tencent's goal of leaping forward in the AI era.
In January, CEO Pony Ma publicly admitted at an employee conference that Tencent had been "slow to act" in the AI field.
For a long time, Tencent's strategic focus remained on mature cash-cow businesses like social media and gaming, leading to a conservative approach to prioritizing AI large models and agents. Management's "take it slow" mindset translated into a wait-and-see attitude at the execution level.
AI research and development were scattered across multiple business units, with data silos, inconsistent models, and significant resource waste. It was not until the end of 2024 that Tencent's AI products, Yuanbao and the Hunyuan Large Model, officially began productization. By then, competitors like ByteDance's Doubao and Alibaba's Qianwen had already captured critical market opportunities.
WeChat, as a well-established super-ecosystem, also became a "sweet constraint."
Out of concern for crossing user boundaries, AI products were long embedded only as plug-ins within WeChat, limiting functional breakthroughs. Since this year, Tencent has ramped up promotional efforts for its AI product Yuanbao on WeChat, but user complaints have led to restrictions on certain permissions.
Still, the contrast between "massive user base" and "AI conversion potential" means that finding the right entry point could trigger exponential growth in user value. The concept of AI agents aligns perfectly with Tencent's need to resolve this dilemma—by transforming WeChat into an AI gateway without altering the core user experience.
One of the major challenges OpenClaw faced in overseas markets was permissions; it requires extensive authorization to deliver value. When integrated with more niche products, the information an AI agent can access and the actions it can perform are clearly bounded. But WeChat, with its extensive and comprehensive coverage across scenarios, offers the broadest possible scope for AI agent development.
Thus, once AI agents presented a way for Tencent to balance "relying on WeChat" with "breaking through constraints," an eager Tencent naturally seized on the concept, integrating various AI assistant features it had previously experimented with and launching a series of conceptual products.
Tencent is not pursuing a single technological breakthrough but rather transforming AI capabilities into an "intelligence enhancement layer" for WeChat, evolving the platform from a "social tool" into an "intelligent task hub."
The biggest issue with the current "lobster farming" craze is that it elevates an open-source execution framework to the level of AI agents, using narrative to create anxiety.
Industry consensus holds that true AI agents must possess autonomous capabilities across a full "perception-planning-decision-execution-reflection" loop, enabling them to break down goals, adapt to dynamic environments, utilize multiple tools, and optimize iteratively without continuous human intervention.
OpenClaw, however, relies heavily on training data coverage and manually defined execution processes for information retrieval and decision output. It can only perform repetitive tasks in predefined scenarios without understanding the underlying intent behind tasks; once removed from a preset environment, faced with dynamic changes or novel problems, its execution can deviate significantly, making it unable to complete complex, open-ended tasks without human oversight, let alone possess autonomous awareness.
Furthermore, as a "weekend project" developed by an independent creator, OpenClaw has significant shortcomings in architecture design and default configurations—85% of instances are exposed to the public internet, and sensitive information is stored in plaintext. If maliciously exploited, such open-source AI frameworks could lead to large-scale attacks, posing a serious risk given the immense volume of sensitive data within the WeChat ecosystem.
Now, network and data security are not only technical requirements but also essential for development and represent legal red lines. When industry participants in China become overly enthusiastic, it's no surprise that regulatory authorities issue successive safety warnings to cool things down.
The nationwide lobster farming craze is a classic example of industry hype—there's no need to avoid or justify that fact.
This hype stems from the industry's impatience for progress in the AI arena. Whether it's tech giants eager to stake their claims or service providers and developers of AI applications, everyone hopes to leverage yet another wave of viral attention to accelerate technology adoption, attract capital, and stimulate market activity.
The value of the "lobster farming" hype lies in making the concept of AI agents accessible and engaging to a broader audience through a fun, low-barrier approach. It transforms the highbrow field of AI into an easy-to-understand, participatory "pet-raising" experience.
The frenzy will eventually fade, and how it concludes will be the real test for all involved. Fad followers may exit hastily amid a wave of uninstalls, but industry leaders will have prepared their next moves—gathering valuable insights from the chaos, identifying potential bubbles, eliminating missteps, and capturing the fundamental user demand for efficient, intelligent execution tools.
Comment