In November 2025, CCTV and Meituan released a three-minute short film about a graphic designer who quits her desk job to deliver food in Dali, Yunnan. The film showed clean mountain air, scenic routes, and a life of freedom. Viewers called it disgusting. The video was taken down within days.
I keep thinking about that backlash. It was not about the film itself. It was about the gap between the story and the reality everyone already knows. Food delivery riders in China do not glide through postcard villages. They sprint through apartment complexes, dodge security guards, and pray the elevator arrives before the timer hits zero. The algorithm does not care about scenery.
But something shifted in 2025. The government finalized the Live E-Commerce Supervision and Management Measures in late November, and earlier in the year, regulators pushed new national standards for platform algorithms. GB/T 46862-2025, which took effect this year, explicitly bans platforms from using fines as routine punishment for delivery delays. Platforms must now factor in merchant prep time, traffic, and weather. A delay caused by a slow kitchen is no longer automatically the rider’s fault.
I spoke with a rider named Wu Bin in March. He was 28, had worked in advertising, and switched to delivery for the flexibility. He told me he ran fifty orders a day and earned around 10,000 yuan a month. He also told me about the fear. One bad review, one late delivery, and the deduction stings. He showed me the neighborhoods where bikes are banned, where he has to park and sprint. “The algorithm chases you,” he said. “You are not choosing to be fast. You are surviving.”
By summer, the “delivery wars” between platforms had turned into a race to the bottom on price. The 2024 national ride-hailing report showed driver hourly income dropped from 31 yuan to 27 yuan. Some drivers slept in their cars. Others ran two or three phones to game the dispatch systems. A 45-year-old veteran rider in Nanjing told a reporter that during peak promotion periods, he could clear 1,000 yuan in a day. The catch was he barely slept.
Then the small changes started. Platforms began extending delivery windows during peak hours. Bad weather brought automatic penalty waivers. Some cities, like Suzhou, launched “algorithm negotiation meetings” where riders could voice concerns directly. The concept of “rider-friendly communities” spread, with designated rest stops, legal aid, and even summer camps for riders’ children. It is not enough. But it is something.
What strikes me is how quietly this is happening. There are no grand announcements, no viral hashtags about worker liberation. Just a gradual recalibration of the timer. A few extra minutes here, a waived fine there. The system is learning to wait, not because it grew a conscience, but because the cost of churn got too high and regulators started asking questions.
The CCTV film imagined delivery work as a dream. The reality is that riders like Wu Bin still spend their best hours staring at a phone screen, calculating routes, and hoping the elevator comes. The difference in 2025 is that someone, somewhere, is finally counting the seconds from the rider’s side too.
Will it last? I don’t know. Platforms have a way of finding new pressure points. But for now, the algorithm is learning to wait. And that is a strange, small thing to be grateful for.