这股热潮也凸显劳动市场压力。文章指出，全球高管对 AI agents 的炒作已持续一年多，而中国今年将进入劳动市场的大学毕业生人数，甚至超过比利时总人口。所谓「一人公司」的政策想像，强化了 AI 可能缩减而非创造职位的担忧；在青年失业率持续偏高的背景下，这使自下而上的采用看起来更像职涯焦虑的放大器。

风险同样以复合方式上升。OpenClaw 可读取电子邮件、传送讯息并执行金融交易，符合研究人员所称网路风险「致命三要素」：敏感资料存取、对外通讯能力与接触不受信任内容。尽管其开源模式可让数千、甚至数十万开发者并行改进，带来更快除错与迭代，但中国如今实际上正在进行全球最大规模的 AI agents 现实测试，而普通人的金钱与生计正成为测试成本。

Over the past week, OpenClaw has shifted from a niche tool into a national craze in China. Hundreds queued outside Tencent's Shenzhen headquarters last Friday for installation, with users ranging from schoolchildren to retirees, showing extremely fast adoption. Tencent, Zhipu, and MiniMax quickly launched compatible products, while local governments including Shenzhen's Longgang, Wuxi, Hefei, and Suzhou drafted subsidies, showing that corporate and state resources are scaling this agentic AI expansion at the same time.

The frenzy also exposes labor-market strain. The article argues that global executives have hyped AI agents for more than a year, while the number of university graduates entering China's labor market this year exceeds Belgium's total population. The policy vision of one-person companies reinforces fears that AI may shrink rather than create jobs; against persistently high youth unemployment, bottom-up adoption looks more like an amplifier of career anxiety.

Risk is rising in compounded form as well. OpenClaw can read emails, send messages, and execute financial transactions, matching researchers' cyber-risk lethal trifecta: access to sensitive data, external communication ability, and exposure to untrusted content. Although its open-source model lets thousands, possibly hundreds of thousands, of developers improve it in parallel, bringing faster debugging and iteration, China is effectively running the world's largest real-world trial of AI agents, with ordinary people's money and livelihoods absorbing the cost.