Beijing Official Builds AI Disaster App for Less Than 30 Yuan

A Beijing district official used AI tools to build a disaster-prevention mini program for less than 30 yuan (US$4.42) in token costs, drawing wide online attention after the system was used to track hazards and evacuations in Miyun District, Capital News reported yesterday.

Xie Yunshi, deputy director of the Miyun branch of the Beijing Municipal Commission of Planning and Natural Resources, developed the district-specific mini program, called Jiaoying, after buying about 1 billion tokens and spending nearly a month coding it himself.

The mini program records geological hazard sites, at-risk households and responsible officials across the district. It also updates mountain warnings, rainfall changes and the status of resident evacuations in real time.

Its popularity prompted questions online about what tokens are and how much 1 billion of them would cost.

"It looks like a huge number, but it really did not cost much," Xie said in an interview with Capital News. "I subscribed to a domestic large-model package that costs a little over 200 yuan a year. The mini program took a month and consumed an average of 1 billion tokens, but the apportioned cost was only more than 20 yuan, less than 30 yuan."

Xie said he built the tool to solve a recurring problem during the flood season. After warnings were issued, more than a dozen or even 20 staff members often had to call villages, towns and households at risk one by one, sometimes late at night. The calls could add up to tens of thousands of minutes and required large amounts of manpower and resources.

"The function of the mini program is to let village and township officials and community publicity workers carry out their tasks within their own areas," Xie said. "After evacuation at a hazard site is completed, they can update it directly in the mini program, without repeatedly calling to confirm whether people have left."

Programming as a hobby

Xie, 50, studied civil engineering at university but has kept programming as a hobby for more than 20 years. He said he once won first prize in a university computer competition and has continued updating his computer knowledge.

He said similar projects once required several months of writing code line by line. With AI, he said, once the architecture is in place, tasks such as database design, back-end work and front-end development can be handed over to AI.

The mini program was tested for half a month before it went online, and Xie said grass-roots officials responded enthusiastically. As users raised new needs, he adjusted prompts to better fit their work habits.

The most direct change after launch was fewer phone calls, Xie said. Information also became easier to see: Officials can open their phones and immediately view how many hazard sites each town has and how many people are involved.

After the mini program went viral, some internet users raised concerns about possible leaks of sensitive information.

Xie said the system has four levels of access: administrators, district-level investigators, township-level users and village-level users. Each group can see only information within its authority. The hazard site locations are based on public information, and the base map uses Tianditu, China's official map service.

"I served as a grass-roots official in a township and handled evacuations, so I know what townships need and I am familiar with grass-roots work processes," Xie said. "That makes the tool easier for people to use."