Typhoon Bavi grazed northern Taiwan on July 11 before making landfall in Taizhou, in China's Zhejiang province — exposing two starkly different approaches to disaster management across the strait.
The storm, roughly the size of France, brought heavy rain and wind to both sides of the strait and triggered panic buying across China's east coast, emptying shelves at Walmart, Yonghui and RT-Mart stores in Beijing. But the bigger story was the gap in how the two sides responded — and a viral tale from Beijing about a district official who built his own AI flood-warning app.
Taiwan Cancels Work and School, Evacuates Mountain Villages
Bavi had already brushed Japan's southern islands before reaching Taiwan without a direct hit at full strength. Still, Taiwan's Central Weather Administration warned northern mountain areas could see nearly a meter of rain, prompting typhoon-day closures of workplaces and schools, grounded flights at Taoyuan International Airport, and the evacuation of more than 14,000 mountain residents.
Social media footage from Taipei showed a different story at street level — pedestrians with umbrellas, convenience stores open, daily life largely intact. The storm still injured 113 to 134 people in the hills, mostly from falls and debris, though no major fatalities were reported.
China Orders Evacuations Across Four Coastal Provinces
China's response was far more centralized. Zhejiang, Fujian, Beijing and Shanghai evacuated nearly two million people combined, and the national meteorological center issued a red alert — its highest rainstorm warning — as forecasters cautioned Bavi's rain bands could linger over multiple provinces. In Wenzhou, residents stocked up on water and food, while state media highlighted grassroots cadres deployed to the front lines, a familiar "people first" governance narrative.
It reflects a broader pattern: fast, top-down deployment through a five-tier emergency response system. Some village committees in Beijing's outer suburbs have carried satellite phones since catastrophic flooding hit the Beijing-Hebei region in 2023, allowing officials to call for help if storms cut normal communications.
A Beijing Official's Self-Funded AI App Goes Viral
One story out of Beijing's Miyun district stood out. Xie Yunshi, deputy director of the district branch of the Beijing Municipal Commission of Planning and Natural Resources, spent his own money — enough for what he described as 1 billion AI tokens — to build a flood-response app called "Jiaoying," or "Call and Respond." It consolidates the district's hazard sites, at-risk residents and responsible officials, updating in real time as rainfall, landslide warnings and evacuation status change, and can navigate users directly to hazard zones.
The app went viral almost immediately, trending on Chinese social platforms as a model of grassroots officials using technology practically. Storm Media's review of current AI API pricing suggests 1 billion tokens would actually cost only a few thousand to several tens of thousands of yuan — a modest sum despite the striking headline figure.
The Viral App's Foreign AI Roots Raise Questions
Xie's own record complicates the story. His technical background shows he has previously built apps using tools including Claude, the AI system made by U.S. company Anthropic, and has written about a "plan first, build second" approach to development. His personal blog briefly went offline after traffic surged following the app's viral moment — an outage that fed online speculation about civil servants using foreign AI to handle sensitive data, and about officials possibly taking on outside technical work.
Should Xie later say he switched to a domestic AI model, the episode would fit a familiar script in Chinese officialdom: local officials responding to a leadership priority — here, the "governance modernization" push since the CCP's 20th National Congress in 2022 — with a visible, low-risk show of innovation. AI has moved from industrial policy into everyday governance over the past two years, showing up in document drafting, public messaging and city administration as local governments adopt homegrown tools. When a practice wins central approval, local governments tend to replicate it as a "model case" — making it plausible Xie's app could eventually be packaged as a governance achievement, or folded into cadre training.
Both Approaches Face Judgment in the Recovery Ahead
Taiwan's system leans on scientific forecasting, public communication and individual judgment — its typhoon-day closures are often joked about but appear to keep casualties low. China relies more on mass mobilization and grassroots execution, with AI tools now entering the mix as a genuine, if uneven, experiment in tech-driven governance. Xie's case captures the tension: enthusiasm for new technology running up against rules on handling sensitive information, where any official's self-funded "innovation" is likely to be read more than one way.
Taiwan opted for a low-key, practical response; China's grassroots officials answered with a modernization narrative. How well each holds up will become clear as recovery from Bavi continues.
