Chinese tech workers are being ordered to train the AI agents that will replace them, and they're starting to say no.
The Summary
- Tech workers in China are being instructed by their bosses to train AI agents to replace them, using tools like the GitHub project "Colleague Skill" to distill their skills and personality traits.
- The practice is triggering a wave of soul-searching among workers who were otherwise enthusiastic early adopters of AI tools.
- This marks an inflection point: the people building the future are realizing they might not be in it.
The Signal
A GitHub project called Colleague Skill surfaced earlier this month with a straightforward pitch: workers could use it to replicate their colleagues' skills and personality traits with AI. The tool wasn't positioned as a threat. It was framed as productivity enhancement, digital preservation, knowledge management. The kind of language that makes dystopia sound like efficiency.
Then Chinese tech companies started mandating it. Not suggesting. Not piloting. Mandating. Managers told workers to feed their expertise into these systems, to train their digital replacements as part of their job duties. The people who build software for a living suddenly found themselves building their own obsolescence, one training session at a time.
"Workers could use it to distill their colleagues' skills and personality traits and replicate them with AI."
The backlash isn't coming from Luddites or technophobes. These are enthusiastic early adopters, the exact demographic that normally evangelizes new tools. They use AI assistants. They automate their workflows. They're fluent in the language of augmentation and enhancement. But being asked to actively participate in your own replacement crosses a line that automation theory doesn't prepare you for.
This is different from the typical "AI will take jobs" conversation in two critical ways:
- The timeline collapsed from theoretical future to mandatory present
- Workers aren't being replaced by AI, they're being conscripted to build it
- The agents aren't meant to augment human work, they're meant to archive it for future extraction
China's tech sector is ground zero because the power dynamics are explicit. There's less pretense about "human-AI collaboration" and more direct acknowledgment of what's happening: cost reduction through replacement. Western companies wrap the same dynamic in softer language, but the mechanism is identical. Train the model on your work. Document your process. Make yourself reproducible.
The Implication
This is the moment where the agent economy stops being abstract and becomes visceral. If you're a knowledge worker anywhere, watch what happens next in China's tech sector. The resistance patterns, the workarounds, the ways people protect their expertise or sabotage their digital doubles. That's the playbook for the next five years everywhere else.
The strategic question for companies: can you actually retain talent if your explicit policy is training their replacements? The strategic question for workers: what parts of your expertise are you willing to externalize, and what do you keep locked in your head?