The Digital Double Dilemma: Chinas Tech Sector Grapples with the Rise of AI Agents and the Automation of Professional Identity
The landscape of the Chinese technology sector is currently undergoing a profound and unsettling transformation as employers increasingly pressure staff to train artificial intelligence agents designed to replicate their own professional functions. What began as a series of experimental internal projects has evolved into a broader cultural phenomenon, sparking a wave of soul-searching among a workforce that has traditionally been among the most enthusiastic adopters of emerging technologies. At the heart of this shift is the emergence of "AI agents"—software entities capable of performing complex, multi-step tasks—and a viral GitHub project that has forced a national conversation regarding the dignity of labor, the ownership of personality, and the future of job security in an automated age.
The Catalyst: From Viral Satire to Corporate Strategy
The current discourse was ignited in early April 2024 when a GitHub repository titled "Colleague Skill" began circulating widely on Chinese social media platforms like Weibo and Rednote (Xiaohongshu). The project claimed to offer a method for users to "distill" the professional skills, decision-making patterns, and even the personality traits of their coworkers into a digital surrogate. By integrating with ubiquitous Chinese workplace communication tools such as ByteDance’s Lark and Alibaba’s DingTalk, the tool promised to generate comprehensive manuals and executable AI agents that could mimic a specific employee’s output.
While the creator of the project, Tianyi Zhou, an engineer at the Shanghai Artificial Intelligence Laboratory, later clarified that the project was intended as a satirical "stunt" to highlight the absurdity of AI-driven layoffs, the reaction from the tech community was far from humorous. For many, the tool felt less like a joke and more like an instruction manual for their own obsolescence. The project struck a nerve because it mirrored a growing reality in Chinese offices: bosses are increasingly mandating that employees document their workflows in granular detail to facilitate the training of agents using platforms like OpenClaw or Anthropic’s Claude Code.
Technical Mechanisms of Professional Distillation
To understand the anxiety surrounding Colleague Skill, one must look at its technical premise. The tool functions by ingesting vast quantities of proprietary data. A user provides the name of a colleague and basic profile details; the system then scrapes chat histories, shared documents, and project management logs from internal platforms. Using large language models (LLMs), it synthesizes this data to create "reusable manuals" that describe not only the technical duties of the role but also the "unique quirks"—such as specific punctuation habits, response times, and interpersonal tones—that define a human worker’s professional presence.
Amber Li, a 27-year-old tech worker based in Shanghai, recounted her experience testing the tool as a personal experiment. After recreating a former colleague’s persona, she found the results "uncanny." Within minutes, the AI had produced a functional profile that could debug code and respond to queries in a manner nearly indistinguishable from the original person. "It captures the person’s little quirks," Li noted, describing the feeling as "uncomfortable" and "alienating." This ability to turn a human being’s professional essence into a modular, digital asset is at the core of the current labor unrest.
The Corporate Logic: Standardization and Data Harvesting
From a management perspective, the push toward AI agents is framed as a logical evolution of operational efficiency. Hancheng Cao, an assistant professor at Emory University who specializes in the intersection of AI and labor, suggests that companies are pursuing these "work blueprints" for reasons that extend beyond simple headcount reduction. According to Cao, firms are seeking to convert "tacit knowledge"—the unwritten expertise and intuition held by experienced employees—into "explicit knowledge" that the company can own and scale.
By forcing employees to codify their decision-making patterns, companies gain a rich repository of data on internal workflows. This allows leadership to identify which aspects of a job are truly creative or judgmental and which are merely repetitive processes that can be standardized. In a market where China’s tech giants are facing increased pressure to maintain margins amidst a slowing economy, the ability to "de-skill" certain roles through AI distillation represents a significant cost-saving opportunity.
A Chronology of the AI Agent Surge in China
The rapid escalation of this trend can be traced through a series of technological and cultural milestones over the past eighteen months:
- Late 2022 – Mid 2023: The global explosion of generative AI leads Chinese tech firms (Baidu, Alibaba, Tencent) to launch their own LLMs. Initial use cases focus on chatbots and creative assistance.
- Late 2023: Shift in focus from general-purpose AI to "AI Agents." Unlike chatbots, agents are designed to execute actions, such as booking travel, managing calendars, or writing and deploying code autonomously.
- Early 2024: The "OpenClaw" framework becomes a national craze in China. OpenClaw, an open-source project designed to give AI agents control over computer interfaces, prompts a gold rush among developers seeking to automate office tasks.
- April 2024: The "Colleague Skill" GitHub project goes viral, followed closely by the "Anti-Distillation" counter-movement. The debate moves from technical circles to mainstream social discourse.
Resistance and the Anti-Distillation Movement
As the pressure to automate increased, so did the ingenuity of the resistance. Koki Xu, a 26-year-old AI product manager in Beijing with a background in law, responded to the trend by publishing an "anti-distillation" skill on GitHub. This tool was specifically designed to sabotage the process of creating AI workflows.
Xu’s tool allows workers to feed the distillation process "poisoned" or generic data. It offers three modes of operation—light, medium, and heavy sabotage—depending on the level of managerial oversight. The tool rewrites employee manuals into vague, non-actionable language that remains superficially professional but is functionally useless for training an AI agent. A video demonstrating the tool garnered over 5 million likes across Chinese platforms, signaling a massive undercurrent of frustration among tech workers.
Xu argues that the trend raises significant legal and ethical questions regarding the ownership of "professional identity." While a company owns the copyright to the code or documents an employee produces, the "tone," "judgment," and "personality" of a worker occupy a legal gray area. "I believe it’s important to keep up with these trends so we as employees can participate in shaping how they are used," Xu stated, emphasizing that the goal is not to reject AI, but to protect human dignity within the system.
The Socio-Economic Context: 996 Culture and Labor Overhang
The intensity of the AI agent debate in China is inseparable from the country’s unique labor environment. For years, the "996" work culture (working 9 a.m. to 9 p.m., six days a week) has been the standard in the tech sector. However, as the industry matures and growth slows, many workers feel they are being squeezed from both sides: they are expected to maintain grueling schedules while simultaneously building the tools that will eventually render them redundant.
For an anonymous software engineer who spoke on the condition of anonymity, the process of training an AI on their workflow felt "reductive." The engineer described the sensation of seeing years of experience "flattened into modules." On social media, this sentiment has manifested as "bleak humor." One popular comment on the platform Rednote suggested that workers should "distill" their colleagues first in hopes that by becoming the "distiller," they might survive the next round of layoffs longer than those being "distilled."
Implications for the Global Tech Industry
While the most visible manifestations of this trend are currently in China, the implications are global. Organizations worldwide are exploring "agentic workflows" to increase productivity. The Chinese experience serves as a bellwether for several key issues:
- The Erosion of Tacit Knowledge: If AI agents successfully replicate the output of senior staff, companies may lose the "institutional memory" and mentorship capabilities that human experts provide.
- Privacy and Surveillance: The use of internal chat logs to train agents represents a significant escalation in workplace surveillance, blurring the line between professional output and personal expression.
- The Devaluation of Expertise: As workers like Amber Li have noted, the ability of AI to mimic their "quirks" leads to a feeling that their unique value is being "cheapened."
- Legal Battles Over Digital Twins: We are likely to see future litigation regarding whether an AI agent trained on a specific person’s data constitutes a "digital twin" that the individual has rights over.
Current Limitations and Future Outlook
Despite the anxiety, the immediate threat of total replacement remains tempered by technical limitations. Tech workers on the ground report that while AI agents are excellent at summarization and basic coding, they remain unreliable for complex, high-stakes business decisions. They require constant human supervision to prevent "hallucinations" or logical errors.
"I don’t feel like my job is immediately at risk," Amber Li concluded, "but I do feel that my value is being cheapened, and I don’t know what to do about it."
The current standoff in China’s tech hubs suggests that the next phase of the AI revolution will not just be about what the technology can do, but about how much of themselves humans are willing to surrender to the machines they build. As "anti-distillation" tools and "Colleague Skill" clones continue to proliferate, the boundary between the worker and the tool remains more contested than ever.
