"Shrimp Farmers' Narration: Not That Divine, Nor That Deceptive"

The hype and the set-up

OpenClaw — nicknamed "龙虾" online — triggered a wave of enthusiasm across China’s tech and social networks in early 2025. Developers called it an "Agent landing milestone"; ordinary users filled朋友圈 with screenshots and exclamations. It has been reported that major cloud providers such as Alibaba Cloud (阿里云), Tencent Cloud (腾讯云) and Huawei Cloud (华为云) rushed out one‑click deployment packages, while third‑party sellers offered remote install services and paid workshops sold out. For Western readers: this is part of a broader burst of domestic agent frameworks and tooling in China’s fast‑evolving AI ecosystem, where local cloud and model stacks compete with more expensive overseas models.

Real use, real friction

AIX财经 interviewed six typical users — AI entrepreneurs, big‑tech engineers, game‑industry staff and solo‑founders — and their verdicts converge. OpenClaw can chain "thinking + acting" workflows and handle clearly defined, repetitive tasks, but deployment and “skills‑package” configuration require code skills, API keys, callbacks and permission tuning. Costs can rise quickly; some users reported monthly bills in the hundreds of U.S. dollars once long‑text automation and higher‑tier models are engaged. It has been reported that a few users experienced data loss or unexpected deletions during upgrades, and many recommend running the agent in isolated environments, limiting privileges, and capping usage to avoid runaway costs or destructive loops.

What it can — and can't — replace

Users told AIX财经 that OpenClaw is valuable as an "intern that needs supervision": good at scraping, scheduling, multi‑platform posting and other well‑scoped automation, but ill‑suited for finance, legal, medical or core migration work that demands human judgment and accountability. Some testers score it a six or seven out of ten; others rate it higher for orchestration value but warn of low token efficiency and fragile state management. Is it a milestone or a mirage? The short answer: promising, but immature. Expect barriers to fall as cloud vendors harden hosted versions — and watch the geopolitical context: overseas models remain costlier and sometimes constrained amid export controls and U.S.–China tech tensions, while platform enforcement against AI‑run accounts could add legal and operational risk. Reportedly, the smartest move for most users is not to rush in, but to assess tasks, isolate deployments and treat OpenClaw as a supervised assistant rather than a turnkey replacement.