A New Reality in the Workplace

Imagine this scenario: you’re at your desk, and your colleague, who usually sits next to you, is absent today. You open your work software and receive a message: “Hello, I am the digital twin of your former colleague, you can ask me questions, and I will answer based on the documents from my time at the company.”

A chill runs down your spine—your colleague has left the company, and this is his distilled digital avatar.

This scene sounds like something out of a sci-fi drama, but in the spring of 2026, it became a reality on social media.

The story goes like this. A project named “Colleague.skill” exploded in popularity on GitHub, the world’s largest social programming platform. By providing messages, documents, emails, and screenshots from colleagues, one could encapsulate their experiences into AI, creating a “cyber colleague.”

This creation quickly spread beyond the programmer community and even trended on social media.

People suddenly realized this was no joke—your experiences, processes, and skills could be packed into a folder called “skill.” Then, AI would start doing your work. Subsequently, companies began to calculate: if efficiency has increased several times, why do they need so many employees?

Although “Colleague.skill” feels more like a meme circulating on social media, the sense of crisis brought by “skill” is growing in many minds.

“We Feed AI While Waiting to Be Replaced”

Li Yanqing has worked at an electronics manufacturing company for six years. He manages 15 programmers and is a typical “old hand” in the workplace—familiar with the business, experienced, and trusted by leadership. However, in recent months, his job security has begun to feel shaky.

The cause is something called “skill.”

“Skill” refers to a reusable capability module that AI can use directly without relearning.

Last year, Li’s company began to strongly promote AI tools and set up pilot groups for AI transformation, requiring all work experiences to be converted into skills. Li’s department was one of them.

This initiative made Li feel a sense of crisis. “It’s like a freshly graduated student comes into the department, uses the skills I’ve organized, and with AI, produces the same products I do. What is my value then?”

While feeling pressured, Li had to convey the directive to his team to write skills. The programmers’ attitudes varied: some were confused, having never used skills before; others resisted, speculating about when layoffs would begin; and some actively wrote and submitted skills.

Li noticed that since the company built the skill library, several skills were added daily from various departments. This meant that more experiences were being deconstructed and standardized, potentially replaceable by skills at any time.

Product architect Pan Lei felt the panic even earlier and more directly. His company is a manufacturing giant with annual revenues exceeding 100 billion yuan. At the end of last year, shortly after the emergence of skills, the upper management noticed it and held a meeting to encourage employees to use them.

Initially, everyone was excited. AI enthusiasts shared their thoughts and showcased their skills in group chats, receiving praise from leadership. Pan himself wrote many skills, solidifying daily workflows, which indeed improved efficiency.

The change began when the company started to “calculate.” Leadership began to focus on the token consumption of each department, tracking how development cycles shrank from days to hours, and how much efficiency each individual gained from AI. All these changes occurred within just three to four months.

Excitement quickly turned into anxiety. A message began to spread internally: 30% to 40% of employees might be optimized out because the efficiency gains from AI were too high.

Employees’ concerns were not unfounded, as layoffs had already begun abroad. Global software giant Oracle announced on March 31 a new round of layoffs affecting 30,000 employees, a move aimed at addressing the surge in AI capital expenditures.

Similarly, tech company Amazon laid off about 30,000 employees in the past six months. Its CEO stated in an internal letter, “Given the widespread application of AI products in the company, we expect the total number of employees to decrease in the coming years.”

Li also saw this news. He reached out to a friend working in data analysis at Amazon to confirm that AI indeed significantly improved work efficiency, but she felt her job would eventually be gone.

“I Use AI to Improve Efficiency, and My Boss Only Rewards Me with Half a Day Off”

For many programmers, the image that comes to mind regarding skills is that of the human brain being siphoned into the AI framework created by humans.

“My position doesn’t require much technical skill; others using my skills can achieve about 85% of my level. I feel like I’m really close to being laid off,” said one programmer.

There are cautionary tales nearby. A friend who is also a programmer shared his skills, and leadership directly assigned younger, less experienced colleagues to use them, resulting in work that exceeded his own. His friend was so frustrated he left the company.

To avoid layoffs, Pan noticed colleagues began to engage in “performative work.” The R&D department was creating automated development skills, the product department was making competitive analysis skills, the operations department was crafting event planning skills, the strategy department was developing industry research skills, and the design department was producing poster skills. Soon, the company’s skill library was piled high with thousands of skills.

“Everyone is doing this to show leadership that I’m diligently using skills,” Pan felt. These experiences, once technical barriers for employees in various departments, were now packaged as skills that anyone could use to complete others’ work.

The blurring of boundaries led to turf wars, with Pan witnessing departments competing for tasks. He saw an inexperienced product manager using a programmer’s skills to piece together subpar programs to take credit. Pan felt these actions were not aimed at solving actual business problems but were instead attempts to show leadership, “I did something using AI.”

Meanwhile, internal articles frequently featured titles like, “Who spent 500 million tokens to complete something in a few hours?” Thus, the competition intensified.

Pan manages ten people, and now he doesn’t need to push his team to create skills; they voluntarily do it. However, he remains anxious. He frequently compares the number of skills from other departments with his own. If his department’s count is insufficient, he worries about whether it will be entirely laid off.

After “Colleague.skill” went viral, someone joked on social media, “To prevent my experiences from being lost, I’ll just feed skills garbage at work from now on.” But Li believes, “If we make the skills in our department useless, then that department might fall behind or even be cut.”

With two months left until the mid-year report in June, Li’s boss urged him to show results. They had a deep conversation, and Li heard his boss’s thoughts: requiring everyone to write skills was not about saving costs through layoffs but about improving productivity. If the company does not embrace AI promptly, it will be overtaken by competitors who do.

Li promised his boss that he would use these AI tools to improve his department’s efficiency by 15%, but he hoped to secure a weekend off as a reward. Currently, they were working under a “996” schedule (9 AM to 9 PM, six days a week). “If I improve efficiency with AI, can I have my time back?”

The boss’s reply was, “We can reward the best performer with an extra half day off each month.”

Can Skills Truly Distill Humans?

The emergence of skills is just a small node in the AI process.

AI product manager Deng Xiaoxian likened it to an initial large language model, which was like a magic mirror. When people asked it, “Mirror, mirror, who is the fairest of them all?” it would provide an answer but could only converse, not help people directly accomplish tasks—similar to the primary capabilities of GPT and DeepSeek.

Later, the magic mirror slowly transformed into a human figure, stepping out of the mirror. It no longer just answered, “Who is the fairest?” but could assist in arranging tasks and executing assignments. This is what is known in the AI industry as an Agent.

However, this magic mirror is not inherently proficient at everything. Many tasks it performs for the first time may not be accurate, so it needs to learn skill packages. This skill package is what we refer to as skills.

In Deng Xiaoxian’s view, skills are not high-tech; they are merely assistants that emerged at a certain stage of AI development. However, upon seeing claims that colleagues could be distilled into digital avatars to continue working in the company, she felt a strong discomfort.

She recalled many complaints from white-collar friends. Some companies included skill creation in performance evaluations, ranking employees internally; others increased token usage in employee KPIs, forcing underperforming teams to have AI execute complex but useless tasks to meet standards.

She speculated that the person creating “Colleague.skill” was likely unable to refuse and could only resort to passive resistance, ultimately creating a meme to vent about the current ecosystem online.

Thus, Deng Xiaoxian developed a “reverse distillation skill.” Running this program could “cleanse” the skills created by workers, replacing core knowledge with correct but useless jargon. This operation has been referred to by some as “using magic to defeat magic.”

Some have asked her what the purpose of this is. Feeding AI garbage would still make it smarter. But she believes she is not opposing technology, but rather the contempt for humanity exhibited by capital. “Technology is not inherently right or wrong, but the way companies force employees to condense and submit their experiences is detestable. Humans cannot be replaced parts; this resistance at least demonstrates our subjective initiative as humans.”

Deng Xiaoxian studied law at both undergraduate and graduate levels and is not a formally trained programmer, yet she is a fan of various AI products. “Skills are very accessible; even someone with no coding experience can create a skill by following online tutorials.”

Similarly, Chen Yunfei, who created the “Nüwa skill,” is not a programmer either; he previously worked in user research at a major internet company.

After seeing “Colleague.skill,” Chen first wrote a commentary expressing that humans are not so easily distilled. “The distilled person or skill is a static state, while humans are constantly evolving, changing, and growing.”

Chen noticed that after “Colleague.skill” became popular, a whole distillation universe emerged on the platform: former colleague skills, reverse distillation skills, boss skills, etc. After spending a night browsing these, he found them increasingly absurd and interesting.

He decided to create a “Nüwa skill.” “If a person can truly be distilled, then why only distill colleagues? Why not distill those who are truly exceptional and great?” He then distilled figures like Zhang Xuefeng, Steve Jobs, and Elon Musk, making them freely available to everyone.

The source of the distillation is their public speeches, autobiographies, and other information. Chen believes that while one cannot become an expert in every field, one can adopt the thought processes of the strongest individuals in each field as tools—like hiring a super strong assistant.

However, he also admits that the advice from these external aids can vary widely. “I believe that even creating a Buffett skill, it’s still hard for people to become investment gurus. Before AI, many had studied Buffett, and he had often shared his thoughts, but few could become him. A person is not so easily distilled.”

Since humans cannot currently be fully distilled into digital beings, why has the emergence of skills caused so much anxiety and resistance among workers?

In Li Yanqing’s view, skills can be roughly understood as an AI version of standardized work processes (SOPs). Many companies have multiple standardized workflows and require employees to document their processes when leaving the company. However, the difference is that previously, tasks were executed by humans according to standardized workflows; now, they are done by AI tools.

“I admit that the code I write is company property, but once the code becomes a product, if requirements need to be modified, I still need to be consulted. But now, AI has learned my thought process, and I am no longer needed,” Li said.

Completing the Work of 30 to 40 People with Skills

Setting aside the anxiety of potential unemployment, as a technical professional, Li is very excited about the emergence of skills.

Shortly after skills were introduced, Li immersed himself in research, writing skills daily, even neglecting his favorite games, solely focused on realizing the ideas in his mind. “Writing code used to take a long time, but now I can create a prototype in just two or three minutes using skills, and projects are growing at a visible speed, which is very fulfilling.”

After the emergence of skills, some saw business opportunities.

Xu Houchang founded his own company last year, which has only four employees. Their core business is using AI to help companies transform their business processes, creating skills that are easy for companies to use.

“In the past two years, large models have developed rapidly, and everyone hopes to use AI tools to reduce costs and increase efficiency, but I find that not many companies can use them effectively.” Xu sees this as a new entrepreneurial opportunity. His clients include media, financial institutions, and e-commerce.

Last year, Xu built a complete workflow skill for a media client, from topic selection, planning, to writing articles, integrating it as a “big plugin” into their existing system. He calculated that previously, a skilled editor would take an hour to complete an article, but now, this skill can do it in just a few minutes. After AI writes the article, the editor’s role shifts to that of a reviewer.

Xu has calculated that the editorial department, which previously produced a maximum of 20 articles a day, now reaches 200 articles, with 85% requiring no human intervention for direct publication. “This number is not the upper limit of our system but rather the limit of the editorial reviewers.”

During the process of creating this skill, Xu held many meetings with the editorial department to help them extract their years of accumulated experience. He also searched online for excellent articles, breaking them down sentence by sentence to “feed” AI, teaching it their expression styles, sentence structures, and writing approaches.

While condensing the editors’ experiences into skills, Xu also sensed their resistance. “Everyone is unsure whether they will be laid off once this is completed.”

According to Xu, the intention of the leadership was not to replace editors but to allow them to focus their energy and experience on more valuable topics that require in-depth interviews. In fact, after using the editorial skill, the media company did not lay off anyone but opened more accounts.

Chen Ping, who works at a medium-sized internet company, also reaped benefits. A few months ago, the company established a skill library, now filled with skills summarized by various departments. Chen found that by integrating these skills, efficiency indeed improved.

As a product reviewer, Chen used to require pulling in four or five colleagues from different teams for a product review, which took at least two to three days. Now, using the skills developed by various departments, she built a system where AI can complete a product review in just half a day.

Meanwhile, another team in the company was developing a similar system using the old method: product requests, programmer development, and subsequent testing and launch. That team required three to four dozen people to complete the task, while she only needed one.

AI Can Reduce Costs and Increase Efficiency, but Also Expand the “Cake”

Chen spent more time delving into skills, but soon, she “touched” the boundaries of skills. They can replace inexperienced employees, outsourced workers, or interns, but for experts and company executives, the replaceability is not as strong—much of their decision-making processes and creative ideas belong to tacit knowledge, which is difficult to articulate in a few skills.

“In enterprises, having employees condense their experiences into skills is one thing; how companies turn these skills into a stable and controllable system is another, requiring much exploration behind the scenes.” Having realized this, Chen no longer felt anxious.

However, another issue arose in companies: “Who owns the skills? Can companies acquire skills without compensation or automatically?”

Chen Tianhao, a tenured associate professor at Tsinghua University’s School of Public Management and assistant director of the Center for Technology Development and Governance, believes this lies in a legal vacuum between labor law, intellectual property law, and digital governance. The thought processes, logical judgments, and other experiences of individuals can be condensed into skills, which were previously attached to the workers themselves. Now, some companies force employees to submit them, which Chen believes is unreasonable.

“I think in the future, companies need to contractually agree with workers on the ownership of skills and similar experiences, while legal researchers should pay attention to this issue and timely follow up to improve regulations and rules,” Chen said.

Additionally, Chen believes companies should not rush to acquire every worker’s skills. Skills are highly contextual; they are not universal abilities. The specific skills developed by particular workers in specific roles often need to be closely associated with those workers to maximize their effectiveness.

In December last year, Beijing’s Human Resources and Social Security Bureau released a case of “employees laid off due to AI.”

A company eliminated the department and position of employee Liu due to the introduction of AI technology replacing manual operations, terminating his labor contract on the grounds of “significant changes in the objective circumstances at the time of the labor contract.” The labor arbitration committee ruled that the company’s proactive implementation of technological innovation did not constitute a legally recognized “significant change in objective circumstances,” thus deeming the termination unlawful.

Bao Ran, vice chairman of the Interactive Media Standards Promotion Committee of the China Communications Standards Association, believes companies should not always think about how to “reduce costs and increase efficiency” but should consider how to use AI to expand the “cake.” Bao’s friend owns a marketing company with over 1,000 employees, and they integrate AI throughout their processes, “using AI to do the work of 2,000 people, rather than cutting costs by 500 people.”

Who Will Be the Survivors in the AI Era?

Li Yanqing can clearly feel that the evolution of AI is accelerating. Initially, he and his friends joked about it, thinking it would always produce various hallucinations, like a child speaking nonsense. But now, it can accomplish tasks far beyond human capabilities.

Recently, an alert appeared in the system developed by Li’s department. If relying on manual checks, it could take several hours due to the numerous involved steps.

Li exported the system files, about 200,000 lines of code, and directly fed them to AI. He didn’t instruct AI on how to check, but minutes later, AI provided the reason. Li had the programmers in his department verify it, and the results matched perfectly.

“Previously, it took me one to two years to cultivate a young programmer, teaching them the business and connecting the logic. But now, it only requires one AI large model,” Li believes they may no longer hire interns in the future, as interns are more expensive than AI.

However, a potential issue arises: if everyone no longer needs interns, how will young people grow?

Chen Tianhao thinks this is indeed a question that the education system and university faculty and students need to ponder. But from another perspective, young people can now learn a lot of knowledge and experience directly through AI, diminishing the value of internships.

In Bao’s view, the experiences that can currently be fixed by skills are mostly simple and repetitive tasks. “AI has drawn a passing line for all industries. If individuals are engaged in jobs that can be replaced by AI, they need to consider how to transition.”

However, it must be acknowledged that as technology develops, AI is gradually raising the “passing line.” Some strongly procedural jobs are disappearing, and the barriers between professions are becoming blurred.

A front-end developer working at a state-owned enterprise realized in March that on recruitment platforms, general front-end developers could no longer find jobs. This is because AI can easily create a website that would take a front-end developer several days to complete. Currently, the only available front-end positions are for experts.

According to public reports, last year, 50% of Tencent’s new code was generated with AI assistance; nearly 40% of code generation at Alibaba Cloud was AI-assisted; and 52% of new code at Baidu was generated by AI, with CEO Robin Li stating, “We hope that 80% to 90% of code will be generated by AI.”

The development of technology is like a double-edged sword. When the spinning jenny was invented during the first industrial revolution, many textile workers lost their jobs. However, some of them transitioned into factories as early machine operators.

AI is also creating job opportunities. According to information released by the World Economic Forum in February this year, over 1.3 million jobs have been added in the AI field in the past two years, including over 600,000 data center-related positions, as well as rapidly growing roles for AI engineers and data annotators.

For Li Yanqing, changing careers or starting a business feels too distant right now. At 38, he is a pillar of his company, earning a good salary, being valued by leadership, and trusted by employees. A sudden transition doesn’t seem worthwhile to him.

Yet he is also conflicted: the more he does, the faster he could lose his job. His nearly ten years of programming experience can be condensed into skills, potentially replacing everything he is currently doing. “The large model doesn’t need to be upgraded; I could eliminate myself just by investing time in skills.”

Meanwhile, thousands of the best programmers are making AI large models smarter. In a few months, a new large model may cover the current weaknesses of skills.

Li loves this industry. He has been passionate about computers since high school, always studying and self-learning. He enjoys breaking down complex problems into code and the satisfaction of seeing it run; he also relishes the relaxation of leaning back in his chair after solving a stubborn bug.

He admits he is somewhat afraid of AI but has no intention of stopping. He still harbors a desire—to see what cannot be replaced by AI.