China's AI displacement is showing up in rate cards, not just layoff notices

What ended Fei Zhaojun's job was a manager's read rather than a benchmark. His bosses judged AI close enough for coding work while he still doubted it ^[2], and he has since come round on the substance, saying mid-level coders are essentially replaceable in most cases ^[3]. His reason for using the tools is not a claim about quality at all: even if it proves a disaster, he said, at this stage you have to use it because everyone else does ^[4]. Adoption driven by that logic does not pause for verification, and it does not unwind when the output disappoints.

The enterprise diffusion figure works out to a rise of roughly five times, 37.9 percentage points inside a single year ^[1]. It is also a survey of what firms say about themselves. The collapse in live-action short series reaches the report secondhand, through Chinese media, and sits next to generative AI moving into creation, production and distribution in that same industry ^[10]. Soft numbers, both of them. The price signal is harder: a rate cut of that depth means a translator has to at least double output to stand where he stood before ^[2], and the arithmetic holds whether or not any survey respondent is telling the truth.

Where the displaced land is the part the evidence cannot settle. Shujing He of Plenum describes people who have left traditional workplaces as eager to experiment with AI-enabled businesses and independent ventures ^[12]. Fei's version of that is vlog-style short videos about ordinary people's lives, which he says do not make him a living ^[16]. The same reporting carries economists who think displacement on this scale could eventually undermine the strength of the world's second-largest economy ^[14], and an International Labour Organization finding that women carry higher risk because their work concentrates in more automatable tasks such as electronics assembly ^[15]. Nothing in the material reconciles the optimistic absorption story with the demand-side warning.

For anyone pricing exposure to this, the distinction worth holding is between markets that adjust through headcount and markets that adjust through rate. Salaried work reprices in one afternoon and yields a number someone can count. Freelance and contract work reprices continuously and quietly, which is why the contract rate is the better instrument even though the layoff is the story. China is running this with the state pushing from one side and, on the reporting available, no organised resistance on the other, so it will produce readable numbers before anywhere else does.