A cold farewell can be turned into warm tokens.
That’s the dark joke circulating on Chinese social media right now — and it captures exactly what’s happening inside tech companies across the country. Workers aren’t just watching AI creep into their workflows. Their bosses are asking them to hold the door open.
A Joke That Hit Too Close
It started with a GitHub project called Colleague Skill. The premise was satirical: feed it a coworker’s chat history and profile from workplace apps like Lark or DingTalk, and it generates a detailed manual replicating that person’s tasks, tone, even their punctuation habits.
The creator, Tianyi Zhou, built it as a spoof — a dark commentary on AI-related layoffs rattling the tech industry. It went viral anyway. Not because workers found it funny, but because it felt uncomfortably real.
“It is surprisingly good,” Shanghai-based tech worker Amber Li told MIT Technology Review. “It even captures the person’s little quirks.”
This Isn’t Hypothetical Anymore
Chinese tech bosses, riding the wave of OpenClaw — an open-source AI agent tool that became a national obsession this year — have started pushing employees to formally document their workflows. The explicit goal: use that documentation to train AI agents to handle the same tasks.
The spread of OpenClaw agents happened so fast that government agencies and state-owned enterprises began warning staff not to install the software on work devices, citing cybersecurity risks. That warning didn’t slow the corporate appetite for automation. If anything, it sharpened it.
Hancheng Cao, an assistant professor at Emory University who researches AI and labor, put the dynamic plainly: firms gain not just tool experience from this exercise, but “richer data on employee know-how, workflows, and decision patterns.” In other words, the documentation isn’t just for automation — it’s an intelligence grab.
This Is What Job Replacement Actually Looks Like
Most coverage of AI and jobs focuses on the abstract — future unemployment projections, macro studies, think-tank reports. The Chinese tech worker story is different because it’s specific and immediate.
One anonymous software engineer told MIT Technology Review that training an AI on their workflow felt reductive — as if years of judgment, context, and craft had been flattened into a set of interchangeable modules. That feeling matters. It’s not just psychological discomfort. It’s a preview of how the company leverages shifts once those modules exist.
This isn’t new territory, either. U.S. startup Mercor has spent years hiring unemployed professionals — laid-off journalists, lawyers, video editors — to train AI models on the very jobs they used to hold. That the same dynamic is now playing out in China, at scale, through employer mandates rather than freelance gigs, suggests we’re past the pilot phase.
The countermeasures are telling, too. Koki Xu, a 26-year-old AI product manager in Beijing, published an “anti-distillation” skill on GitHub — a tool specifically designed to make a person’s workflows harder to replicate. That such a thing exists, and that people are using it, says more about the stakes than any labor report.
What It Means
Here’s what to watch: whether this workflow-documentation trend migrates from China’s tech sector into other industries and geographies. The tools are already global. OpenClaw and similar agent platforms don’t respect borders, and multinational companies are watching these experiments closely.
The worker who said “I don’t feel like my job is immediately at risk, but I do feel that my value is being cheapened” may have identified the actual first stage of AI displacement — not mass layoffs, but a quiet erosion of leverage. When your entire role can be described in a manual, negotiating power goes with it.
The question isn’t whether companies will use that manual. It’s when.
Related: After the OpenClaw Surge, Reality Sets In: Why AI Experts Aren’t Impressed
