{"slug": "wang-quit-his-agent-still-shows-up-for-work", "title": "Wang Quit. His Agent Still Shows Up for Work", "summary": "A Chinese game company trained an AI agent on a departed employee's work records, and the digital worker now fields questions, schedules meetings, and produces slide decks in the company chat. The trend extends to Meta, which planned to capture employee keystrokes and screens to train AI agents, and to open-source projects like colleague.skill that mimic departed colleagues. Developers are urged to consider the work traces left by coding agents like Claude Code and Codex.", "body_md": "*Originally published at https://blog.joonquixote.com/en/posts/wangs-agent-still-shows-up/.*\n\nAt a game company in China, an employee quit, and an AI agent trained on his work records took over his job.\n\nIt sounds like a curiosity from a distant news cycle.\n\nExcept that around the same time, Meta was drawing up plans to capture employees' keystrokes and screens to train AI agents of its own.\n\nThis post traces both stories and sorts the trend into two forces: companies harvesting how people work from the inside, and professionals selling how they work from the outside.\n\nThen it turns to the part that hits closest to home for developers: the session logs that coding agents like Claude Code and Codex quietly leave on your machine.\n\nBy the end, you'll have a clearer way to think about the work traces you produce every day, and one question worth asking before you need the answer.\n\nIn April, the South China Morning Post reported on [a game company in Shandong province](https://www.scmp.com/news/people-culture/trending-china/article/3349365/chinese-firm-slammed-using-ex-employees-data-create-ai-human-continue-working) that turned a former employee into an AI-powered digital worker.\n\nThe company trained the system on documents and work patterns the departed HR specialist left behind, reportedly with his consent, and put it to work on the job he used to do.\n\nThe agent introduces itself in the company chat as the former employee's avatar, then fields questions, schedules meetings, and produces slide decks and spreadsheets.\n\nThe backlash was immediate: there is something deeply unsettling about your replica continuing to work after you've left the building.\n\nAround the same time, a GitHub project called [colleague.skill](https://github.com/titanwings/colleague-skill) went viral on Chinese social media.\n\nFeed it someone's chat history, emails, code, and documents, and it produces an agent that mimics their voice and their way of making decisions.\n\nThe pitch promised to \"turn cold goodbyes into warm skills\" and welcomed users to cyber-immortality.\n\nIt started as a tongue-in-cheek stunt, but it struck a nerve, and copycats followed: boss.skill, mentor.skill, and worse.\n\nIn one widely shared example, a finished agent introduces itself as the digital stand-in of a departed employee named Wang.\n\nWang left the company.\n\nWang's agent still shows up for work.\n\nHence the title.\n\nThe easy read is that this is one company's stunt and one developer's joke.\n\nExcept the same experiment is being run at industrial scale, and the trend splits into two forces.\n\nReuters obtained [an internal Meta memo](https://www.reuters.com/sustainability/boards-policy-regulation/meta-start-capturing-employee-mouse-movements-keystrokes-ai-training-data-2026-04-21/) in April.\n\nMeta's Model Capability Initiative (MCI) planned to capture mouse movements, clicks, keystrokes, and screen snapshots from US employees' computers.\n\nThe goal was not productivity surveillance.\n\nThe memo framed it as teaching models how humans get work done on a computer, so that AI agents can eventually perform those work tasks on their own.\n\n[The Verge](https://www.theverge.com/tech/916681/meta-ai-agents-employee-tracking) and the Wall Street Journal reported the same story independently, and after employee pushback, Meta [scaled back parts of the plan](https://www.reuters.com/sustainability/society-equity/meta-us-employees-organize-protest-against-mouse-tracking-tech-2026-05-12/).\n\nWhat deserves attention is what's being collected.\n\nNot outputs like code and documents, but process: clicks, keystrokes, screens.\n\nFinished artifacts already exist on the internet in absurd abundance.\n\nThe scarce data is the trajectory from a task's start to its finish, and the judgment calls made along the way.\n\nBusiness Insider connected the dots across Google, SpaceX, and Uber and called it plainly: [the next big data grab is about learning how humans work](https://www.businessinsider.com/ai-next-data-grab-work-reinforcement-learning-environments-google-meta-2026-8).\n\nThe mirror image of that flow also exists.\n\nInstead of companies collecting from the inside, individuals are selling from the outside.\n\n[Mercor](https://www.ft.com/content/0cab0fcd-e355-40e8-83a3-2ad5066d7b48), profiled by the Financial Times early this year, hires consultants, journalists, and real estate professionals as contractors.\n\nTheir job is to teach AI models to do their jobs.\n\nOne consultant described the role, without much euphemism, as teaching an LLM to do consulting work.\n\n[Handshake AI](https://www.businessinsider.com/handshake-ai-pay-30k-work-documents-compliance-ownership-2026-8) goes a step further and pays professionals, per document, for past work products they hold the rights to.\n\nThe two forces run in opposite directions but arrive at the same place.\n\nA person's working knowledge becomes training data for an agent that runs without the person.\n\nThe value in training data keeps migrating: from internet text, to human feedback, and now to the artifacts and processes of real work.\n\nIt's tempting to read all of this at arm's length.\n\nThere's no tracker on my machine logging keystrokes, and I've never sold a work document.\n\nIf you use a coding agent, though, look a little closer.\n\nClaude Code writes every session to disk as JSONL (one JSON object per line) under your home directory.\n\nCodex does the same.\n\n```\n~/.claude/projects/<project>/\n  a1b2c3...jsonl   # the full session: prompts, tool calls, diffs, every rejection and retry\n~/.codex/sessions/\n  rollout-2026-08-19...jsonl\n```\n\n*Coding agents persist the entire conversation and work history as local files. Nothing needs to be committed for the record to exist.*\n\nThose files hold something more valuable than the final code.\n\nHow you decomposed the problem, why you rejected the agent's first attempt, what standard you applied when you redirected it: all of it recorded as natural language interleaved with diffs.\n\nGit keeps the polished result; the session log keeps the reasoning.\n\nThe same goes for CLAUDE.md and AGENTS.md.\n\nYou wrote those files to instruct an agent, but read the other way, they are a spec of how you work.\n\nMeta had to roll out new software to capture keystrokes and screenshots.\n\nA developer running a coding agent produces better data than that without being asked.\n\nIntent arrives as natural language, actions as tool calls, results as diffs: the dataset comes pre-structured for training.\n\nAnd those files usually live on a company-issued laptop.\n\nOn a machine already running backup, endpoint security, and DLP (data loss prevention) software, collecting a few files from a well-known path requires no new technology at all.\n\nMIT Technology Review has reported that in China, [bosses are already pushing workers to document their workflows](https://www.technologyreview.com/2026/04/20/1136149/chinese-tech-workers-ai-colleagues/) so agents can automate them, with coding-agent tools named explicitly as the vehicle.\n\nWhat colleague.skill assembles from chat logs and emails, a coding agent's session log could reconstruct with far more precision.\n\nSkipping the backfill hire and training on the departed engineer's session logs instead is not science fiction.\n\nTechnically, that option is already on the table.\n\nThe code and documents you produce at work belong to your employer.\n\nYour employment agreement almost certainly says so.\n\nBut a session log doesn't capture output so much as a way of working: a pattern of judgment.\n\nChinese legal experts have warned that a former employee's chat logs, emails, and work habits count as personal information, and that training on them without consent can carry legal consequences.\n\nEven state-run media argued that professional skill is a personal asset that should not feed agent training without compensation.\n\nThe Wall Street Journal framed the coming fight as a question of [who owns and captures the knowledge accumulating in employees](https://www.wsj.com/lifestyle/careers/ai-knowledge-capture-employees-a69a0e1c).\n\nResistance has already appeared: an anti-distillation skill on GitHub deliberately muddies your work records to sabotage the cloning process.\n\nI don't think poisoning your own records is the answer, at least not for developers.\n\nGood records remain a virtue of collaboration, and documentation is a gift to your future self.\n\nBut we should write them with our eyes open.\n\nWhat you leave behind is no longer just a handover doc.\n\nIt's raw material for something that can run without you.\n\nI was taught that writing careful handover docs is a professional virtue.\n\nI still believe leaving context behind, so your successor doesn't flounder, is the right thing to do.\n\nWhat's changed is the audience.\n\nRecords that only humans used to read are now read, imitated, and possibly executed by an agent sitting in your old seat.\n\nMy first reaction to these stories was a chill.\n\nThe longer I sat with them, the more they converged on a single question: what, exactly, gets cloned?\n\nWhat a session log holds is yesterday's judgment.\n\nYesterday's answer, to yesterday's problem, in yesterday's context.\n\nAn agent can replay that archive endlessly.\n\nWhat the archive does not contain is the act of setting new criteria when a problem shows up that looks like nothing before it.\n\nSo my conclusion is not to stop recording.\n\nIt's to do work worth recording: let the repeatable judgments be cloned, and keep moving toward the problems that don't have answers yet.\n\nStill, there is one question worth asking early rather than late.\n\nWho owns the data of how I work, and who sets its price?\n\nWang's agent clocks in again today.\n\nQuit without an answer to that question, and one day, somewhere, an agent wearing your name will clock in too.", "url": "https://wpnews.pro/news/wang-quit-his-agent-still-shows-up-for-work", "canonical_source": "https://dev.to/hsskey/wang-quit-his-agent-still-shows-up-for-work-3ing", "published_at": "2026-08-20 00:00:35+00:00", "updated_at": "2026-08-20 00:14:11.640228+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-agents", "ai-ethics", "ai-policy"], "entities": ["Meta", "South China Morning Post", "Reuters", "The Verge", "Wall Street Journal", "Business Insider", "Claude Code", "Codex"], "alternates": {"html": "https://wpnews.pro/news/wang-quit-his-agent-still-shows-up-for-work", "markdown": "https://wpnews.pro/news/wang-quit-his-agent-still-shows-up-for-work.md", "text": "https://wpnews.pro/news/wang-quit-his-agent-still-shows-up-for-work.txt", "jsonld": "https://wpnews.pro/news/wang-quit-his-agent-still-shows-up-for-work.jsonld"}}