Don't hype 'lobsters' into 'milk tea'

Mania on China's internet

The key story is simple: an open‑source AI agent nicknamed OpenClaw (龙虾) has been hyped into a national craze — and that hype is far louder than the technology’s actual utility. Tech giants from Tencent (腾讯) to Baidu (百度) and Xiaomi (小米) reportedly raced to offer integrations and free installs, with social feeds full of people showing off their “lobster” setups. Capital markets followed: concept stocks tied to the craze surged, and conferences, training classes and paid installation services sprang up overnight. But does every user actually need one?

Technical limits, costs and safety

The frenzy masks hard realities. It has been reported that the OpenClaw project — while demonstrating important advances in agent automation — carries nontrivial security risks when deployed with default or improper configurations, and users have reportedly experienced errant behaviour including unauthorized actions and data deletion. The tool requires specialist setup (MCP services, Skills, API keys, Linux and browser‑simulation work), and token costs can be steep — one user test reportedly burned the equivalent of RMB200 in three tasks, and some installers suggest ongoing monthly costs in the thousands of RMB. For ordinary consumers, the product currently looks more like a fragile developer tool than a plug‑and‑play household assistant.

Bigger picture and why it matters

This sprint must be read in a broader context. Chinese cloud and AI firms are operating amid intense domestic policy pushes to industrialise AI — the 2026 government work programme reportedly prioritised “AI+” commercialisation — and global competition that includes export controls on advanced chips. The result: a powerful mix of state encouragement, corporate "flow" anxiety and investor FOMO. Microsoft CEO Satya Nadella has warned about AI becoming a speculative bubble; similar dynamics are visible here. If AI is treated like a trendy consumer fad — “milk tea” to be copied and consumed quickly — the industry risks wasting capital and distracting from the long, rigorous engineering work needed to make these systems safe, affordable and genuinely useful.