Chinese tech workers are being pushed to document their workflows for AI agents—and the backlash is revealing deeper anxieties about identity, dignity, and the future of work
A satirical GitHub project has struck an unexpectedly raw nerve among China’s tech workforce. Called Colleague Skill, the tool allows users to “distill” a coworker’s professional habits and personality into a reusable blueprint for an AI agent. Though its creator built it as a joke, the concept resonated because it reflected something many workers were already experiencing firsthand: pressure from management to document their own workflows for automation.
As explained here, the tool works by pulling chat histories and files from popular Chinese workplace apps like Lark and DingTalk, then generating detailed manuals that an AI agent can follow—right down to a person’s communication quirks and punctuation habits. Tianyi Zhou, a Shanghai-based engineer who built it, told Chinese media that the project was a direct response to AI-related layoffs and the growing expectation that employees should automate themselves out of a job.
“It even captures their little quirks”
Amber Li, a 27-year-old tech worker in Shanghai, tried the tool out of curiosity, using it to recreate a former colleague. The result was unsettling in its accuracy. Within minutes, she had a detailed profile of how that person worked—their habits, their reactions, even their writing style. She now uses the AI stand-in to help debug code and get instant responses.
“It is surprisingly good,” Li said. “But it also felt uncanny and uncomfortable.”
Her experience is becoming less unusual. Since AI agent tools like OpenClaw exploded in popularity across China, employers have been encouraging—and in some cases explicitly pushing—workers to experiment with automation. One anonymous software engineer described the process of training an AI on their own workflow as deeply dehumanizing: their years of expertise and judgment had been reduced to a set of interchangeable modules.
On Chinese social media platform Rednote, workers have turned to dark humor to process the shift. One user quipped that “a cold farewell can be turned into warm tokens”—suggesting that documenting a coworker’s skills before your own might buy you a little more time before the axe falls.
What companies are really after
While the Colleague Skill trend may seem like corporate trend-chasing, academics argue there’s a more calculated logic at play. Hancheng Cao, an assistant professor at Emory University who studies AI in the workplace, points out that workflow documentation gives companies far more than just automation tools.
“Firms gain not only internal experience with these tools, but also richer data on employee know-how, workflows, and decision patterns,” he explains. “That helps companies identify which parts of work can be standardized—and which still require human judgment.”
In other words, the exercise of asking employees to document themselves isn’t just about building better AI. It’s about giving companies a clearer map of where humans can eventually be cut out of the equation.
Fighting back: The “Anti-distillation” tool
Not everyone is going along with it quietly. Koki Xu, a 26-year-old AI product manager in Beijing, spent about an hour building a deliberate countermeasure: an “anti-distillation” tool designed to sabotage the workflow documentation process from the inside.
Xu’s tool lets workers choose between light, medium, and heavy sabotage modes, depending on how closely management is watching. In each mode, the agent rewrites documentation into vague, generic language—plausible enough to pass a quick review, but useless for training a meaningful AI replacement. A video she posted about the project racked up over five million likes across platforms.
Xu, who holds both undergraduate and master’s degrees in law, is quick to point out the legal murkiness underlying all of this. While employers can reasonably claim ownership over work files and chat logs created on company devices, tools like Colleague Skill also capture something harder to define: a person’s tone, judgment, instincts, and personality. Who owns those? The answer, she argues, is far from settled.
“I originally wanted to write an op-ed,” Xu said, “but decided it would be more useful to make something that actually pushes back.” She hopes the conversation around these tools will spark broader debate about how to protect workers’ dignity as AI becomes more deeply embedded in professional life—even as she herself runs seven AI agents across her personal and work devices.
The gap between hype and reality
For now, at least, mass replacement remains more threat than reality. Li says her company has yet to find a reliable way to substitute AI for actual employees—the tools still require too much supervision and produce too many errors to operate independently. “I don’t feel like my job is immediately at risk,” she says.
But the psychological damage may already be done. “I do feel that my value is being cheapened,” Li adds, “and I don’t know what to do about it.”
That tension—between AI’s current limitations and its rapidly expanding ambitions—sits at the heart of what China’s tech workers are navigating right now. The tools may not be ready to fully replace them yet. But the process of preparing for that replacement has already begun to change how workers see themselves, and how their employers see them.
There is a certain irony in the current landscape: many AI communicators and enthusiasts—those who every day promote tools, write guides, and fuel collective excitement around AI—may be among the first to be replaced by the very technology they celebrate. Those who produce content about how to use AI, who explain workflows and optimizations, are unknowingly helping to build the manual for their own replacement. It is exactly the same dynamic experienced by the Chinese tech workers in this article, just wearing a different name tag.
But framing AI as an enemy to fight risks becoming a distraction. The real problem is not the technology itself—it is the social and economic context in which it is adopted, and the complete absence of any adjustment to working conditions in the face of an epochal transformation.
Until we seriously address issues such as reduced working hours, a basic or transitional income for those displaced by automation, adequate wages that reflect the real productivity gains generated, also thanks to AI, and a fairer redistribution of the profits that automation produces, no “anti-distillation” tool will be able to change the rules of the game. It will only be symbolic resistance within a system that remains structurally unchanged.
The Chinese workers who document their colleagues to survive a few more weeks, and those who sabotage their own workflows to avoid being replaced, are both reacting in an understandable way to a system that pits them against a machine while offering no safety net. The real question is not “how do we stop AI?”—it is “who benefits from this transformation, and who pays the price?”
As long as the answer to that last question remains unchanged, the unease emerging from Shanghai to Berlin, from Beijing to Milan, will find no solution in GitHub repositories—but in public squares, collective agreements, and public policy.
