Is this an actual prediction?这是预测吗？
No. It is a structural read by one analyst against a public methodology. The numbers are hand-set, calibrated against the four sub-scores per the formula above, and capped at one analyst's confidence. Use it as a discussion frame, not as advice for your specific career. Real outcomes depend on geography, regulation, employer choices, and the speed of capability shifts that nobody currently models well.
否。这是一名分析者依公开方法所作的结构性研判。数字按上文公式与四项子分手工设定与校准，上限止于一名分析者的信心。请作为讨论框架使用，不应作为针对你个人职业的建议。真实结果取决于地理、监管、雇主选择，以及目前无人能很好建模的能力推进速度。
Why hand-curated instead of model-generated?为何手工编制而非模型生成？
Because every published auto-rated job-exposure list I've seen is dominated by the model's own training-data biases. Hand scoring forces a transparent rubric, an explicit per-row rationale, and a small enough surface that any wrong row can be challenged and revised by name. ~90 occupations is the size where one analyst can actually defend each entry. A 20,000-row auto-rated list is much bigger but defends none.
因为我所见到的每一份自动评分的职业暴露榜单都被模型自身的训练数据偏见所主导。手工评分迫使评分量表透明、每行有明确理由、表面足够小以便每一行都可被指名质疑与修订。约 90 项是一名分析者真正能为每一条目辩护的规模。两万行的自动榜单虽大，但没有一行是可辩护的。
What's the most over- and under-rated category here?这里最被高估与最被低估的类别是？
Overrated by the public (= my scores are lower than the panic): teacher, nurse, therapist, surgeon, plumber, electrician. Each requires presence and judgment in unstructured settings — exactly what current AI is worst at. Underrated by the public (= my scores are higher than the comfort): paralegal, contract reviewer, insurance underwriter, junior copywriter, junior financial analyst. Each is mostly text-and-rules, which is exactly where current AI is strongest.
公众高估者（即我的分数低于恐慌）：教师、护士、心理治疗师、外科医生、水管工、电工。每者皆要求在非结构化情境中的临在与判断——正是当前 AI 最差之处。公众低估者（即我的分数高于安慰）：律师助理、合同审阅者、保险核保员、初级文案、初级金融分析师。每者皆以文本与规则为主，正是当前 AI 最强之处。
What does "exposed" actually mean for an individual?'暴露'对个人到底意味着什么？
It means three structurally distinct things, often conflated. (a) Augmentation — your output multiplies and your job changes shape; pay can go up. (b) Headcount thinning — the same total work is done by fewer people; some get cut. (c) Total displacement — the role disappears as a market category. The same composite score can map to any of the three depending on demand elasticity, regulation, and employer choices. A high score means "watch this category"; it does not mean "you specifically will be replaced." Mid-career workers should optimise for the part of their job that requires NR-COG and NR-INT.
它意味着三件结构上不同的事，常被混为一谈。(a) 增强——你的产出倍增，工作形态改变；薪酬可能上升。(b) 人员削减——同样的总工作由更少的人完成；部分被裁。(c) 全面位移——岗位作为市场类别消失。同一综合分依需求弹性、监管、雇主选择，可对应三者中任一。高分意味着'关注此类别'；并非'你个人将被替代'。中段从业者应优化工作中需要 NR-COG 与 NR-INT 的部分。
What jobs are missing — and what would change the scores?遗漏了哪些职业？什么会改变这些分数？
Missing: most blue-collar field roles (firefighter sub-types, fishing, agriculture), most narrow specialists (actuaries, surveyors, audiologists), and most jobs that exist mostly outside the BLS taxonomy (gig work, content creators, professional gamers, ritual workers). The dataset is biased toward roles legible to the US labour-statistics frame; readers should adjust accordingly. Things that would change the scores fast: a single robust tabletop physical-task model (would lower R-PHY and raise composite for plumber, nurse aide, etc.); regulatory carve-outs for specific professions (would freeze scores in legal, medical, financial advisory); and any genuine self-improving model release (would raise NR-COG-resistance bar significantly). Watch those three signals.
遗漏：多数体力现场岗位（消防员细分、渔业、农业）、多数窄类专家（精算师、测量员、听力学家），以及主要存在于 BLS 分类之外的职业（零工、内容创作者、电竞选手、仪式工作者）。本数据集偏向美国劳动统计可识别的岗位；读者应据此调整。能快速改变这些分数的事项：一个真正可靠的台架物理任务模型（将降低水管工、护理员等的 R-PHY 并抬升综合分）；针对特定行业的监管豁免（将冻结法律、医疗、财顾的分数）；以及任何真正自我改进模型的发布（将显著抬高 NR-COG 抗替代的标杆）。关注这三个信号。