Chinese workers push back against AI 'double' project
SkimNews Take
The pushback from tech workers against training their AI doubles highlights an emerging tension where the very individuals who develop advanced AI are now directly confronting its disruptive potential in their own careers.
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Colleague Skill went viral on Chinese social media earlier this month, claiming to clone coworkers into reusable AI agents; the project was created as a spoof by Tianyi Zhou, an engineer at Shanghai Artificial Intelligence Laboratory.
- Colleague Skill automatically imports chat history and files from Lark and DingTalk, generating manuals that describe a coworker's duties and quirks for an AI agent to replicate.
- Amber Li used the tool to recreate a former coworker, producing a file that captured job tasks, punctuation habits, and quirks, and now employs an AI agent to help debug code, though she feels uncanny and uncomfortable.
- OpenClaw and Claude Code are AI agent tools that Chinese bosses are urging employees to experiment with, but workers report limited utility in business contexts.
- Hancheng Cao says firms gain richer data on employee workflows and can standardize or codify work, while workers feel alienated by the process.
- Koki Xu released an “anti‑distillation” GitHub skill that sabotages workflow creation with light, medium, and heavy modes; a video of the tool went viral, earning over 5 million likes.
Why it matters: Firms gain richer data on employee workflows and the ability to codify tasks, potentially standardizing work and reducing reliance on human judgment. Meanwhile, workers feel their expertise is being flattened into modules, threatening their sense of dignity and raising concerns about job security and ownership of personal traits.
Ask SkimNews
