"I haven't even figured out how to rAIse shrimp (openclaw), and everyone is already raising horses?" This sentiment has recently taken over AI circles. Just a month ago, the tech community was buzzing about "raising shrimp," but almost overnight, "raising horses" has become the new trend.
What is this "Horse"?
This viral "horse" goes by the name Hermes Agent, an open-source autonomous AI agent framework launched in February by the US-based team Nous Research. Due to the phonetic similarity and shared name with the lUXury brand Hermès, Chinese developers affectionately nicknamed it "Ai Ma Shi" or simply "The Horse." In just two months since its open-source release, it has skyrocketed to 110,000 stars on GitHub, ranking first in the monthly trends with 98,000 new stars.
Major domestic tech giants are jumping on the bandwagon. On April 10, Xiaomi announced that its MiMo platform had integrated the Hermes Agent for a two-week free trial. Following suit, Tencent Cloud, Alibaba Cloud, and MiniMax have also integrated the Framework. Most excitingly for Chinese users, the agent natively supports WeChat, compatible with images, videos, files, and voice messages. This means ordinary users can start "raising horses" directly within WeChat without wrestling with complex configurations.
Advanced Tools vs. Autonomous Employees
Industry insider Zhang Xuan offers a fitting analogy: "Shrimp" (OpenClaw) ACTs more like an advanced tool requiring clear, explicit instructions. In contrast, the breakthrough of the "Horse" lies in its "employee-like" mindset and initiative. Specifically, the "Horse" boasts three key tricks:
Self-Evolution: It summarizes and updates its methods after every task, becoming smarter the more it is used.
long-term memory: It remembers user preferences and doesn't suffer from "amnesia."
Security Awareness: It proactively confirms high-risk actions, such as deleting files or mass messAGIng, before executing them.
Pang Fei, who works in tech investment and uses both systems, notes that "Shrimp" excels at Standardized workflows, like summarizing announcements or performance reports. Meanwhile, the "Horse" is better suited for cross-scenario complex tasks, such as deep analysis of startups. They aren't replacements for one another but rather form a digital Productivity "team": the shrimp handles the execution ("working"), while the horse handles the strategy ("thinking").
Security Concerns and Market Trends
Regarding security, the previous "shrimp" craze exposed significant vulnerabilities. Wang Yuanyuan from Topsec Technology Group provided data showing that between March and April alone, 155 vulnerabilities were recorded for OpenClaw, with over 40% being critical or high-risk, involving issues like rEMOte code execution. Currently, the vulnerability disclosure rate for the "Horse" APPears lower. However, she warns that no system is absolutely safe; users must ensure credential encryption and strict network permission controls to mitigate deployment risks.
As with any AI boom, opportunists are quick to cash in. e-commerce platforms are flooded with products like "Local Horse Deployment," "Technical Resource Libraries," and "installation Support," with prices ranging from 2 to 500 yuan.
Jiang Han, a senior researcher at Pangu Think Tank, advises that the decision to adopt this technology should depend on genuine personal needs. The technical bARRier for "raising horses" has lowered significantly, and its core value lies in solving long-term, complex, and personalized needs by automatically refining Skills from past tasks. If you are a deep content creator or workflow optimizer, it can be a powerful productivity tool. However, if your tasks are one-off and standardized, mature Lightweight tools may already suffice.
Comments & Questions (0)
No comments yet
Be the first to comment!