A viral Chinese tool called Colleague Skill lets workers recreate coworkers as AI agents from job descriptions, sparking debate about worker dignity and replaceability.
A tool called Colleague Skill has gone viral in China, allowing workers to input a coworker's job description and generate an AI agent that mimics their work style, quirks, and decision patterns. Users report the output is surprisingly accurate, capturing punctuation habits and behavioral patterns. The trend is being driven partly by bosses pressuring employees to experiment with AI agents following the OpenClaw craze in China. Academics note companies benefit by capturing employee know-how into codifiable systems — effectively turning tacit knowledge into structured data.
The interesting technical signal here isn't the tool itself — it's the method: structured job documentation as fine-tuning data. Colleague Skill essentially turns tacit expert knowledge into agent-ready prompts by capturing decision patterns, communication habits, and task sequences. That's a workflow capture layer most enterprise AI stacks are missing. Developers building internal tooling or agent pipelines should note this as an architectural pattern, not just a culture story.
Document your own debugging or code review workflow in structured form this week using a plain markdown file, then test whether Claude or GPT-4o can replicate your decisions on 3 real past tickets — measure where it fails to identify your actual automation ceiling.
Go to claude.ai and open a new conversation
Tags