{"version":"2026-10-19.observations.2","reading":{"id":"reading-use","title":"使用份额变了，自己的工作该怎么判断？","deck":"先确定消息、企业还是任务作为分母，再把观察带回自己的验收。","editorialAt":"2026-10-15","sourceIds":["O-SIGNALS","O-AEI","O-CENSUS"],"sections":[{"title":"三种比例，分别回答一个问题","paragraphs":["Signals显示消费版消息的用途构成；AEI显示Claude chat/Cowork中被分类到某个O*NET任务的份额；BTOS问美国企业过去两周是否用过AI。消息可以变多而岗位采用不变，企业可以说用过而只有一个很小环节使用。不能把三种比例放在一起当作工作覆盖。"],"sourceIds":["O-SIGNALS","O-AEI","O-CENSUS"]},{"title":"24个月的主题变化要怎样读","paragraphs":["Signals v2.0每月抽样30万成人消费版消息，分类由模型完成并加权、加隐私保护噪声。主题图的分母是工作相关消息；工作相关份额图的分母是全部消费版样本消息。技术帮助包含数学、数据分析和编程；其份额下降不能解释为编程使用绝对量下降。历史估计曾被重述，所以这次固定v2.0，不能混接旧版本。"],"sourceIds":["O-SIGNALS"]},{"title":"让任务连接真的成立","paragraphs":["AEI 4/5月分类器用O*NET30.2，本站正式任务来自31.0。我们对这89项任务逐项核对Task ID和原英文全文；这个连接只证明使用分类对应，不证明本站子任务完成。","48项任务至少有一个月的公布记录：4月45项，5月46项。两个月合计91个有值的位置；其余87个未公布的位置不填零。","其中23个已公布记录显示0，可能只因取整。因此“公布零”和“没有记录”都不能解释为实际无人使用。"],"sourceIds":["O-AEI"]},{"title":"采用环境对试点有什么帮助","paragraphs":["BTOS是美国企业背景：相邻期并非同一批企业，六组双周轮换样本每12周复访，2026年7月27日有年度样本更新。2026年3月9–22日period96的API返回旧“生产商品或服务”题干，与相邻“任意业务功能”不同；原因未核实，本站排除该期并保留断点。金融与保险行业不等于财务分析岗位。再用自己的任务清单、质量与完整人时决定试点。"],"sourceIds":["O-CENSUS"]}],"actions":["写出每张图的分母和观察窗口","找到真正相同ID的任务；没有匹配就保留未知","以有限工作试点验证自己的收益"],"seriesIds":["signals-technical","signals-work","census-finance"],"taskIds":["C03","F01","M01"],"frontierIds":["F20","F08","F24"],"evidenceStatus":"active","actualCoverage":null},"sources":[{"id":"O-SIGNALS","title":"OpenAI Signals v2.0","url":"https://openai.com/signals/data-download/","version":"v2.0，2024-07至2026-06固定24个月","locator":"数据字典§2/5/6/7；ZIP topic/work_related monthly表","scope":"18岁以上消费版消息，每月抽样30万；加权与差分隐私；不含Enterprise和Codex","license":"CC BY 4.0；署名OpenAI Signals v2.0","limit":"消息分类份额，不能推断工作人员采用、工作完成率或绝对消息量","publishedAt":null,"readAt":"2026-10-15","collectedAt":"2026-10-14","status":"active","nextReview":"2027-01-13","methods":["https://cdn.openai.com/signals/data-dictionary.pdf"],"family":"O-SIGNALS"},{"id":"O-AEI","title":"Anthropic Economic Index June2026","url":"https://huggingface.co/datasets/Anthropic/EconomicIndex/blob/main/release_2026_06_26/data_documentation.md","version":"release_2026_06_26；Claude chat/Cowork GLOBAL 2026-04/05","locator":"schema、pct定义、缺失门槛；onet hierarchy_level=0","scope":"聊天/Cowork Free/Pro/Max；分类器O*NET30.2，本站31.0按ID+英文任务全文桥接","license":"数据CC-BY（原文未指定版本）；代码MIT；本站翻译/筛选改编并署名Anthropic","limit":"未公布不等于零；已公布0可能因两位小数取整；份额不是验收或工时","publishedAt":"2026-06-26","readAt":"2026-10-15","collectedAt":"2026-10-14","status":"active","nextReview":"2027-01-13","methods":["https://www.anthropic.com/research/economic-index-june-2026-report"],"family":"O-AEI"},{"id":"O-CENSUS","title":"美国Census BTOS AI使用调查","url":"https://www.census.gov/hfp/btos/api/periods","version":"固定period89–110；96题干断点不纳入","locator":"period/{id}/data；National或NAICS2；answer=Yes","scope":"美国非农雇主企业抽样调查，是否过去两周在任一业务职能使用AI；相邻期并非同一批企业；六组双周轮换样本每12周复访，年度样本更新","license":"美国联邦公开统计；只整理少量公布估计并链接原始端点","limit":"企业自报采用；不是员工/任务采用或生产率。2026-07-27样本更新，跨期人群不固定","publishedAt":null,"readAt":"2026-10-15","collectedAt":"2026-10-14","status":"active","nextReview":"2027-01-13","methods":[],"family":"O-CENSUS"}],"actualCoverage":null}