Artificial intelligence has already become part of everyday life for a lot of people, but one specific situation is making tech workers in China pretty uncomfortable.
A GitHub project called Colleague Skill went viral on Chinese social media with a concept that sounds like something straight out of a Black Mirror episode: using AI to replicate the skills, habits, and even the quirks of your coworkers.
The project was created as satire — a sharp joke about corporate culture and AI-related layoffs.
But it ended up hitting a nerve that a lot of people had already been feeling, they just didn’t know how to name it.
The thing is, this isn’t just fiction.
Several tech workers told MIT Technology Review that their bosses are actively asking them to document their own workflows to automate tasks using AI agent tools like OpenClaw or Claude Code.
In other words: training your own replacement.
And between dark humor, personal experiments, and even sabotage tools, this story is raising questions the corporate world still doesn’t have ready answers for. 👇
The Project That Became a Mirror for Collective Anxiety
Colleague Skill isn’t exactly a commercial product. It emerged as criticism disguised as a tool — a humorous yet bitter way of pointing out the absurdity of a situation that’s becoming increasingly common at tech companies. The core idea of the project is simple and unsettling at the same time: feed an artificial intelligence model with enough data about an employee — including their communication style, technical decisions, response patterns, and even their most frequent mistakes — so that an AI can simulate that professional in everyday situations.
In practice, Colleague Skill works like this: the user names the coworker whose tasks they want to replicate and adds basic profile information. The tool then automatically imports conversation history and files from Lark and DingTalk, two corporate communication apps that are extremely popular in China, and generates reusable manuals describing that colleague’s job functions and even their unique mannerisms — all ready for an AI agent to replicate the behavior.
Receive the best innovation content in your email.
All the news, tips, trends, and resources you're looking for, delivered to your inbox.
By subscribing to the newsletter, you agree to receive communications from Método Viral. We are committed to always protecting and respecting your privacy.
The practical result would be a kind of digital clone of a worker, capable of answering emails, reviewing code, participating in asynchronous meetings, and executing recurring tasks without that human needing to be present — or present at all. The reaction on Chinese social media was immediate and intense, mixing irony with genuine discomfort that revealed just how bottled up this discussion had already been.
The creator of the project is Tianyi Zhou, an engineer at the Shanghai Artificial Intelligence Laboratory. In an interview with Chinese outlet Southern Metropolis Daily, Zhou explained that the project was born as a provocation, motivated by AI-related layoffs and the growing trend of companies asking employees to automate themselves.
What gave the project its viral power was precisely the fact that it didn’t invent anything. It simply named something that was already happening behind the scenes at many tech companies. Managers in different parts of the world — not just China — are already actively exploring the possibilities of using AI agents to absorb routines that previously depended entirely on people. And the most delicate step in that process is exactly what Colleague Skill satirizes: collecting data about how a professional works, thinks, and solves problems. When you document your own workflow in detail, you are, in a way, writing the manual for your own replacement. It’s a cruel paradox, and the project’s dark humor hit that contradiction with a surgical precision that few serious analyses have managed to achieve.
Of course, satire doesn’t solve the real problem. But it plays an important role: pulling the discussion out of the abstract and placing it at the level of concrete human experience. When someone sees a GitHub repository that describes, almost step by step, how to create a digital version of themselves for corporate use, the emotional impact is different from reading a report about automation and job displacement. That’s what Colleague Skill did for thousands of tech professionals — and that’s why it resonated so strongly. Not because it showed something new, but because it clearly showed something that was already happening in their lives, just without a name.
When Satire Becomes a Real Experiment
Not everyone just sat back and watched. Amber Li, 27, a tech worker in Shanghai, decided to test Colleague Skill after seeing it on social media. She used the tool to digitally recreate a former coworker as a personal experiment. In a matter of minutes, the system generated a file detailing how that person performed their job.
The result surprised her. According to Li, the tool captured even the most subtle mannerisms of her colleague, like the way they reacted to certain situations and their punctuation habits in messages. With that generated profile, she started using an AI agent as a new virtual coworker that helped debug her code and responded instantly. The experience, however, was described by her as both strange and uncomfortable at the same time.
This kind of account shows that the line between satire and practical application got blurred really fast. What started as a joke about corporate culture suddenly became an uncomfortable mirror of a real trend. Chinese internet users also found humor in the idea behind the tool, joking about automating their colleagues before getting automated themselves. In a comment on Rednote, a Chinese social media platform, one user wrote that a cold farewell can be turned into warm tokens — suggesting with irony that whoever distills their coworkers into tasks first might survive a little longer at the company.
Training Your Own Replacement Has Become Routine at Some Companies
The situation Colleague Skill satirizes is not hypothetical. Several tech workers shared accounts — many anonymously out of fear of professional retaliation — that their managers are formally asking them to document their processes with a level of detail that goes far beyond what would be needed for onboarding or project continuity. In some cases, the requests include creating step-by-step guides for how certain decisions are made, which tools are used at each stage, how priorities are set, and even what personal criteria guide the resolution of complex problems.
Since OpenClaw became a national craze, bosses in China have been pushing tech workers to experiment with AI agents. And asking employees to create manuals describing the most granular details of their daily work — exactly the way Colleague Skill does — is one way of trying to bridge the gap between what these agents promise and what they can actually deliver.
This kind of documentation, when combined with modern automation tools based on large language models, can in fact be used to create agents that replicate a large portion of an individual’s professional behavior in repetitive or semi-automatable tasks.
Hancheng Cao, an assistant professor at Emory University who studies AI and labor, believes companies have concrete reasons for encouraging the creation of these work maps beyond simply following a trend. According to him, companies gain not only internal experience with the tools but also richer data about employees’ know-how, workflows, and decision-making patterns. This helps companies identify which parts of the job can be standardized or coded into systems and which still depend on human judgment.
What makes this especially sensitive is the context in which these requests are happening. Many of the companies where this kind of request has been reported have already gone through rounds of layoffs in recent years, and the employees who survived those cuts are hyper-aware of any signal that they could be next. When a manager asks you to document everything you know and how you do what you do — in an environment where automation has already eliminated real jobs — the most immediate interpretation isn’t that the company wants to preserve your knowledge. It’s that the company wants to preserve what you know how to do, without necessarily preserving you.
A software engineer who spoke with MIT Technology Review anonymously due to concerns about job security reported that he trained an AI on his own workflow and that the process felt reductive. According to him, it felt like his entire job had been flattened into modules in a way that made him easier to replace. That reading may not always be correct, but it’s completely understandable within this context — and ignoring the psychological impact on professionals would be a serious mistake in any honest analysis of the topic.
Creative Resistance: Sabotage as Protest
The pressure to create AI agents based on your own work also sparked some clever countermeasures. Frustrated by the idea of reducing a person to a replicable skill, Koki Xu, 26, an AI product manager in Beijing, published a tool on GitHub called anti-distillation skill on April 4.
The tool, which took about an hour to build, was specifically designed to sabotage the process of creating workflows for AI agents. Users can choose from three sabotage modes: light, medium, and heavy, depending on how closely the boss is supervising the process. The agent rewrites the material into generic, non-actionable language, producing an AI substitute that’s far less useful than intended.
A video Xu posted about the project went viral, racking up more than 5 million likes across different platforms. In an interview with MIT Technology Review, she said she had been following the Colleague Skill trend from the start and that it got her thinking about alienation, disempowerment, and broader implications for the job market. According to Xu, she originally wanted to write an opinion piece but decided it would be more useful to create something that actively resisted the trend.
With undergraduate and graduate degrees in Law, Xu also raised important legal questions about the topic. While a company might argue that work chat histories and materials created on a corporate laptop are company property, a tool like Colleague Skill also captures elements of personality, tone, and judgment — making the question of ownership much less clear-cut. She said she hopes the project will spark more discussion about how to protect workers’ dignity and identity in the age of AI.
Xu emphasized that it’s important for employees to keep up with these trends so they can actively participate in how they’re used. And she herself is a tech enthusiast, with seven OpenClaw agents set up across her personal and work devices.
Dignity, Identity, and What It Means to Be Irreplaceable
At the core of this entire discussion is something that goes beyond the jobs themselves. It’s about workers’ dignity and identity. For many tech professionals — especially those who spent years developing expertise in specific areas — work isn’t just a source of income. It’s a significant part of who they are. The way a software engineer solves a complex problem, the creativity a designer brings to a difficult interface, the intuition a data analyst uses to find patterns that models alone couldn’t prioritize — all of that is part of a professional identity that takes years to build. When a company suggests all of that can be documented and then replicated by an AI, it’s not just threatening a job. It’s questioning the singular value of that person.
This is a point that the public debate about job displacement by artificial intelligence often overlooks. Economic analyses talk about reskilling, new opportunities, sectors that will grow while others shrink. And there’s truth in those projections — nobody is saying the world is going to end. But they tend to treat workers as interchangeable resources that can be moved from one function to another as market demand shifts. What that perspective misses is that people don’t experience work that way. They’re not gears that switch positions without emotional cost. When a skill you spent a decade developing becomes potentially obsolete because of a language model, the impact goes far beyond finances. It hits self-esteem, sense of purpose, and the perception of value within a professional community.
The discussion about dignity at work needs to fully enter the conversation about automation and AI — not as a sentimental afterthought, but as a central component of the debate. Companies that are asking employees to document their own processes as a way of feeding AI agents need to be honest about what they’re doing and why. Transparency doesn’t solve the problem, but it respects the intelligence and humanity of those being impacted. And workers, for their part, have every right to question, negotiate, and understand the real terms of what they’re being asked to do. That’s not resistance to progress. It’s a legitimate demand for respect within a working relationship that’s being profoundly transformed by forces most people are still trying to understand.
The Real Limits of AI in the Corporate World
Despite all the noise, it’s important to stay grounded about what AI agents can actually do today. While they can take control of your computer, read and summarize news, answer emails, and even make restaurant reservations, tech workers who deal with these tools on a daily basis say their practical usefulness has still proven limited in real business contexts.
Amber Li herself, the tech worker in Shanghai who tested Colleague Skill, says her company still hasn’t found a way to replace real workers with AI tools, mainly because they remain unreliable and require constant supervision. According to her, her job doesn’t seem to be in immediate danger. But the feeling that lingered is different: the sense that her value is being cheapened, and she doesn’t know what to do about it.
This gap between the promise and the reality of AI agents is precisely what makes the situation so confusing for those caught in the middle of it. The technology still isn’t good enough to truly replace people in most complex roles. But the intention to get there is clear, and the intermediate steps — like asking employees to obsessively document their processes — are already having a real impact on how people feel about their own work.
What Comes Next in the Relationship Between AI and Work
The Colleague Skill episode is, in many ways, a cultural thermometer. It shows where we are in the process of absorbing artificial intelligence into the workplace: at a point of real tension, where the promises of productivity and efficiency from automation are starting to collide with the human realities of the people who operate these systems and are affected by them. AI tools evolve at a speed that leaves little time for social, labor, and psychological structures to adapt. And when the pace of technological change outstrips the pace of human adaptation, the result is exactly the kind of tension we’re seeing emerge now in China’s tech sector — and with less visibility but the same intensity, at companies all around the world.
The trend is for this conversation to deepen in the coming years. As AI agents become more capable of executing complex tasks autonomously, the pressure on tech workers to demonstrate what makes them irreplaceable will increase. This could be a positive catalyst for people to develop increasingly sophisticated and creative skills — the ones that models still can’t replicate with quality. But it could also generate a cycle of anxiety and insecurity that undermines productivity, mental health, and the quality of relationships in the workplace. The balance between these two scenarios will depend, in large part, on the choices companies make about how they implement these technologies and how they communicate that to their teams.
The legal questions raised by professionals like Koki Xu are also poised to gain prominence. If chat history and documents created in a corporate environment are company property, where does the line fall when the tool captures subjective elements like tone of voice, judgment style, and personality traits? This is virtually uncharted legal territory that will demand serious attention from lawmakers and labor attorneys in the years ahead.
What Colleague Skill leaves us with, beyond the bitter laughs and heated debates, is a question worth holding onto: where is the boundary between optimizing processes with artificial intelligence and turning people into disposable data? That line exists, and it matters. Not just for the workers living through this transition right now, but for the kind of corporate culture and society we’re building together as this technology matures. 🤔
