Chinese tech workers are starting to train their AI doubles–and pushing back

The resistance angle is the buried lede here. Workers being compelled to document their own replaceability is a different category of labor conflict than the usual 'AI might take jobs someday' framing — it's coerced knowledge transfer happening right now, inside specific companies, with a named tool.

Import AI 453 (covered April 13) introduced MirrorCode alongside a broader discussion of gradual disempowerment frameworks, and Colleague Skill fits squarely into that same conceptual territory: systems designed to mirror individual workers rather than augment them. The WIRED piece from April 17 on AI in newsrooms showed a parallel dynamic in Western knowledge work, where editorial labor is being repackaged as a productivity input. What's distinct about the China story is the institutional mandate — this isn't workers voluntarily adopting tools, it's companies directing the extraction. InsightFinder's April 16 funding round for AI agent observability also becomes relevant context: if these worker-replica agents fail or behave unexpectedly, there's currently no clear accountability layer sitting between the company and the output.

Watch whether Colleague Skill or a comparable tool surfaces in job postings or HR vendor pitches outside China within the next two quarters — if it does, the resistance patterns documented here will likely repeat in other labor markets with less worker documentation of the pushback.