Soon after his sudden passing on March 24, 2026, renowned Chinese education consultant and influencer Zhang Xuefeng, known for his frank and sometimes contentious advice to students and parents, reappeared—not in person, but in cyberspace.
Based on his public interviews, writings, and a timeline of his life, a model known as “zhangxuefeng.skill”—a digital avatar designed to mimic his speaking style, reproduce his thinking, and offer advice to students in a recognizably personal way—has appeared on GitHub, a hosting platform for software development projects.
But Zhang’s “cyber resurrection (赛博复活 sàibó fùhuó)” has raised a new set of questions: Is it ethical to monetize the dead or, in a more critical turn of phrase coined by netizens, perform this kind of “cyber tomb raiding (赛博盗墓 sàibó dàomù)?” Should a person’s knowledge and experience be extracted, replicated, and continuously reused after they’re gone?
Find out more AI-related trends:
- An AI Guru Has Become a New Battleground Between Young and Old
- Are China’s Schools Ready for the Onslaught of AI?
- Overseas, Optimized: How AI Is Driving China’s Next E-commerce Push
With the rapid advancement of AI, many tech companies and independent developers are beginning to test these boundaries. Giants like Amazon, Meta, and IBM frequently lay off employees while simultaneously advancing AI initiatives and delegating their work to machines.
Others have begun requiring employees to package their work into reusable “skills”—advanced prompts that bundle context, instructions and other information into transferrable files that can be used to teach AI agents how to behave and complete tasks—raising concerns about the impact of a new “silicon-based life (硅基生命 guījī shēngmìng)” on the orthodoxy of “carbon-based life (碳基生命 tànjī shēngmìng).”
The origins of this bold new wave of cyber-human replication can be traced back to late 2025, when Anthropic introduced Agent Skills for Claude, allowing users to load skills to improve the model’s performance on specialized tasks. After evolving into a cross-platform standard, it was adopted by Microsoft, OpenAI, and more than 20 other AI labs.
But what truly brought the term into Chinese public discourse was the concept of “refining a colleague (炼化同事 liànhuà tóngshì),” driven by a recent project on GitHub called “colleague.skill (同事.skill tóngshì.skill),” which gained over 6,600 stars in just five days before quickly going viral.
Many joke that office workers no longer need to worry about the colleague who understands the business best leaving for another company—they can simply “distill (蒸馏 zhēngliú)” that person into a “colleague.skill” and grant them a kind of cyber immortality within the workplace. Simply by uploading materials such as emails, Feishu messages, DingTalk documents, and WeChat screenshots, along with a little additional context, users can “instantly refine (顷刻炼化 qǐngkè liànhuà)” their departed collaborators into an AI agent capable of replicating their job.
When a model can code the way they do, answer questions as they would, and even anticipate exactly when they might pass the buck, it’s described by netizens as:
Large language model + colleague.skill + memory plugin = your colleague
大模型 + 同事.skill + 记忆插件 = 你的同事
Or as the project description itself euphemistically puts it:
Turn the coldness of parting into a warm skill; welcome to cyber immortality.
将冰冷的离别化为温暖的skill，欢迎加入赛博永生。
More than a dozen derivative projects have emerged since “colleague.skill” first appeared, including “ex.skill (前任.skill qiánrèn.skill)” and “boss.skill (老板.skill lǎobǎn.skill).” One viral comment riffs on the well-known Chinese saying “When gathered, we are a blazing fire; when scattered, we are stars across the sky (聚是一团火，散是满天星 Jù shì yì tuán huǒ, sàn shì mǎntiān xīng),” joking:
Colleagues—apart, they’re tokens; together, they become a skill.
同事，散是token，聚是skill。
This discourse reflects a growing anxiety about having to compete with more efficient and lower-cost AI counterparts in the workplace, prompting a common refrain among employees:
My skill has been uploaded, and my desk has been cleared.
我的skill已上传，我的工位已清空。
Those lucky enough to remain at their companies have also turned to humor to process the sense of loss over their former work friends:
Laid-off colleagues haven’t disappeared. They’re still on the server; they’ve just been distilled into tokens to keep you company at work.
被毕业的同事并没有消失，其实他留在了服务器里，被蒸馏成了token继续陪着你。
Some netizens, however, wryly insist that not everyone is worthy of distillation:
Some colleagues, once distilled, would only drag down the AI’s performance.
有些同事，蒸馏出来只会拉低AI的效果。
As the line blurs between “colleague.skill” and “colleague.kill,” more people are coming to terms with the unsettling reality of a living, breathing coworker being reduced to a file. When experience becomes a replicable data asset, anyone can become the next to be “optimized out (优化 yōuhuà).”
Legal considerations have also entered the chat. Questions abound as to whether personal data and attributes such as chat logs, tone of voice, and ways of thinking count as intellectual property, and if they should be available for public use without restriction. As is common in the wake of rapidly advancing technology, the regulatory framework lags behind, leaving a grey area that is open to interpretation.
Chen Tianhao, a tenured public policy professor at Tsinghua University, notes that tacit knowledge developed by workers in the course of their work should, in principle, belong to the workers themselves. According to Chen, labor laws and related regulations may need to be revised, and future employment contracts should clearly define who has the right to use such tacit knowledge—and where the boundaries of that use lie.
While these legal and ethical discussions are timely, some feel the imminent threat posed by AI tools like “colleague.skill” is overblown. Even in the case of a public figure like Zhang, several developers of different Zhang Xuefeng simulators told 21st Century Business Herald in April that the resemblance achieved by such models is currently only 50 to 60 percent accurate. Zhang’s individual knowledge and his long-term social experience are uniquely his own; they do not exist publicly on the internet, nor can anyone simply go out and recreate them.
For many employees, this may come as reassurance. There is comfort in the notion that it is our personal and emotional judgment, our unique composite of social interactions and lived experience, and, especially, those fleeting moments of intuition and inspiration that are our most valuable qualities, and remain the most difficult to replicate. At least for now.
Related Articles
