{"schemaVersion":"1.0","version":"2026-10-21.journeys.2","checkedAt":"2026-10-21","scope":"这里的职责、岗位和团队材料是局部案例，不代表职业目录；职业资料已独立维护于/careers。四个工作领域、三种阅读目的的编辑参照。交付契约、当前能力和未来情景引用工作单元模块；公开岗位与合成团队材料不能代表个人能力或真实组织收益。","sources":[{"id":"J-CANONICAL","title":"Canonical：Software Engineer, Sustaining Engineering","url":"https://job-boards.greenhouse.io/canonical/jobs/3326693","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"This role entails / What we are looking for / Apply for this job","scope":"2026-10-16可见的单一公司岗位样本","limitations":"原始发布日期未知；Worldwide不保证所有国家雇佣资格，申请前核对；此岗位不证明总体需求或AI导致新增招聘。"},{"id":"J-AMAZON","title":"Amazon：Senior Financial Analyst, NACF Planning，10559730","url":"https://www.amazon.jobs/en/jobs/10559730/senior-financial-analyst-nacf-planning","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"Description / Key job responsibilities / Basic Qualifications / Job details","scope":"美国Bellevue的高级财务分析岗位样本","limitations":"原始发布日期未知；岗位职责样本，不代表全球财务需求、当前录用概率或投资建议。"},{"id":"J-ALLEGIANT","title":"Allegiant：Optimization Engineer – Commercial","url":"https://jobs.lever.co/allegiantair/f3650a2f-7122-4d36-86a2-ef8bd52cd9a4","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"Minimum Requirements / Preferred Requirements / Job Duties","scope":"美国现场工作的商业优化工程岗位样本","limitations":"原始发布日期未知；数学任务只占岗位一部分，不能用数学题成绩推适配。"},{"id":"J-PR","title":"AI参与的修复PR为什么未合并？","url":"https://arxiv.org/html/2602.00164v1","publishedAt":"2026-01-29","observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"§2.1–2.3，表1，§3.2","scope":"AIDEV-POP中8106个修复PR及326个关闭未合并样本","limitations":"观察研究，项目/代理分配不随机；合并不等于同质量节省人时，未合并不全是技术失败。"},{"id":"J-PROV2","title":"SWE-Bench Pro V2版本修订","url":"https://labs.scale.com/leaderboard/swe_bench_pro_public_v2?tab=full","publishedAt":"2026-09-22","observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"Update September 22, 2026","scope":"发布方对公开split及隔离评测协议的修订说明","limitations":"官方声明，本轮未复跑审计；V1和V2任务与协议不同，不能直接连成绩计算进步。"},{"id":"J-QWEN","title":"数学推理训练能迁移到其他任务吗？","url":"https://arxiv.org/html/2507.00432v2","publishedAt":"2025-07-01","observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"表1；§2.2；附录A.3","scope":"Qwen3-14B及数学训练方案比较；v2修订2025-10-20","limitations":"相同问题数据不等于相同总算力；不是实际建模交付实验。"},{"id":"J-GDPVAL","title":"GDPval公开任务：虚构巡演损益表","url":"https://huggingface.co/datasets/openai/gdpval/blob/11e7900cdcac61bc4daf59e65feb238acda98fbf/README.md","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"task_id 7b08cd4d-df60-41ae-9102-8aaa49306ba2；train；参考工作簿与rubric","scope":"固定revision 11e7900cdcac61bc4daf59e65feb238acda98fbf；会计任务产物参考；该revision发布于2026-02-10，单项任务首次公开日未知","limitations":"本轮核对题目、rubric和工作簿表头，未执行任务；未找到明确许可，只链接与摘要，不复制文件或rubric。"},{"id":"J-FRONTIERFINANCE","title":"FrontierFinance论文与金融研究评价","url":"https://arxiv.org/html/2608.11683v1","publishedAt":"2026-08-12","observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"§3与§5.1、表4","scope":"220道查询与来源化验收项的金融研究评测","limitations":"条款满足率不是完整交付率，模型裁判有误差，未测本团队成本。"},{"id":"J-SAMAYA","title":"FrontierFinance运营方榜单：9月28日快照","url":"https://samaya.ai/benchmarks/frontier-finance","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"System Performance；Updated September 28, 2026；Notes","scope":"发布方榜单的指定系统、工具预算与费用口径","limitations":"动态页面，本次读取仍标9/28；不同系统/日期不能直接相减为模型进步。"},{"id":"J-MATH-TRANSLATION","title":"Navier–Stokes形式翻译的技术质疑（v1）","url":"https://arxiv.org/html/2610.08144v1","publishedAt":"2026-10-06","observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"active","locator":"Introduction Disclaimer；§3.1.1","scope":"对指定自然语言与Lean材料的语义对应质疑","limitations":"作者不裁决自然语言证明是否正确；未复核整个数学争议或后续所有修订。"},{"id":"R-JOBBENCH","title":"JobBench：专业证据推理与委派偏好","url":"https://arxiv.org/html/2605.26329v1","publishedAt":"2026-05-25","observedAt":"2026-10-19","checkedAt":"2026-10-19","status":"active","locator":"§2.1–2.3；附录E；§3.1","scope":"选题与链式评分方法","limitations":"偏好用于选题；基准没有测量员工采用、满意度或净省时。"},{"id":"R-HARNESS2","title":"Harness²：用执行对照改进运行方法","url":"https://arxiv.org/html/2610.14224v1","publishedAt":"2026-10-09","observedAt":"2026-10-19","checkedAt":"2026-10-19","status":"active","locator":"§2.1–2.3；§3.1；§4","scope":"冻结执行模型下的候选运行系统、行为对照与经验复用","limitations":"作者研究，本站未复现；行为对照不能验证正确性，多次运行和编辑器成本需计入。"}],"domains":[{"id":"analysis","title":"数据与分析","summary":"从指标报表、组成变化到实验分析：AI可以执行更多查询与复算，交付仍要回答数据口径和结论究竟支持什么。","scope":"三种分析交付的编辑参照，覆盖指标工程与实验解释方向；没有已核查的本领域招聘样本，也不代表数据岗位整体。","caseId":null,"forecastId":null,"taskRef":null,"forecastRef":null,"roles":[{"id":"analytics-delivery","title":"分析交付与指标工程","summary":"把业务问题变成可重放的查询、指标表和解释；重点是数据与口径能被接续。","duties":["维护分母、去重、时区与数据快照的定义","交付查询、结果和可追溯说明，定位刷新差异"],"aiChanges":["Agent可以沿数据定义查找、写查询并运行检查，交付单位从一段SQL扩大为可复核报表包","稳定且规则明确的重复刷新优先使用固定程序"],"humanContribution":["决定指标怎样对应业务问题","接收数据权限、口径变更和异常处理"],"entryRequirements":["能写并核对SQL或数据程序，识别join倍增与缺失值","能说明业务口径，保留版本和重放方法"],"portfolioEvidence":["一份含输入快照、查询、基线切片、结果和异常解释的报表包"],"notCovered":["这是按交付组织的职责方向，不是已核查的招聘需求或岗位匹配结论。","公开实践没有验证你的数据权限、口径和业务决策收益。"],"sourceIds":[],"resourceRefs":[{"module":"workunits","type":"source","id":"WU-DATA-AGENT","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"source","id":"WU-DATA-APP","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A02","version":"2026-10-21.workunits.2"}],"evidenceStatus":"active","evidenceWarnings":[]},{"id":"experimentation-analysis","title":"实验分析与决策解释","summary":"判断观察变化来自什么，并区分描述性分解与可支持的因果结论。","duties":["检查分组、分配机制、暴露与流失","计算效应及不确定性，或指出为什么不能识别"],"aiChanges":["AI可执行分解、生成分析和对照备选解释","是否满足因果识别条件仍需由研究设计和实际数据支持"],"humanContribution":["定义决策问题与容许风险","核对随机化、混杂、污染和结论外推范围"],"entryRequirements":["理解比例、加权平均及不确定性","能从分配机制判断可以回答什么，缺什么证据"],"portfolioEvidence":["一份组成变化诊断，明确不将观察性分解冒充因果","一份实验核查记录，保留随机化、流失和假设检查"],"notCovered":["这是按交付组织的职责方向，不是已核查的招聘需求或岗位匹配结论。","公开实践没有验证你的数据权限、口径和业务决策收益。"],"sourceIds":[],"resourceRefs":[{"module":"workunits","type":"source","id":"WU-DATA-AGENT","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"source","id":"WU-DATA-APP","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A02","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A03","version":"2026-10-21.workunits.2"}],"evidenceStatus":"active","evidenceWarnings":[]}],"jobSamples":[],"teamScenario":{"fictional":true,"businessGoal":"合成演练：解释周报中转化率的变化，形成可重放的指标和需要补充的决策依据。","currentTeam":["分析员维护报表与解释","数据工程成员维护输入和定义","业务负责人提出问题并接收行动建议"],"workload":"演练只含两期、两个分组的1000次访问；真实团队的周期工作量和人时保持未知。","constraints":["总体比例与组内比例必须分别核对","描述性分解不等于因果识别","使用授权快照，保留分母、去重与缺失规则"],"options":[{"id":"keep-tools","title":"把稳定指标做成可重复运行的服务","rationale":"适合重复取数与口径确认占据主要时间，而决策问题相对稳定的团队。","now":["固定指标契约和基线切片；Agent辅助初次查询与异常诊断","通过验收后的周期刷新使用固定程序，分析师接收口径变更"],"oneYear":["连续多个周期维持数据与口径质量后，扩大到同类指标服务"],"threeYears":["若语义治理和异常恢复稳定，把更多探索查询纳入受控委派，按新增业务需求调整配置"],"assumptions":["指标定义有责任人，数据快照和权限可追溯"],"tradeoffs":["前期维护数据定义有成本；自动刷新并不自动给出因果解释"],"trigger":"错误若主要来自口径冲突，先补定义和数据责任；不要只增加Agent并发。","evidenceBasis":[{"sourceRefs":[{"module":"workunits","type":"source","id":"WU-DATA-AGENT","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"source","id":"WU-DATA-APP","version":"2026-10-21.workunits.2"}],"aiChange":"内部数据Agent会利用指标定义、查询历史和组织上下文处理分析问题。","handoffChange":"从人工反复传SQL转为接收一包查询、结果与口径说明。","organizationalInference":"本站的组织建议：先补可查询定义与重放检查，保留明确责任人。","condition":"能够对照独立切片与历史基线，权限和语义变化可追踪。","limit":"来源为特定组织的数据Agent实践与实现示例，没有测本合成团队的可靠率、净人时或编制。"}]},{"id":"train-reassign","title":"培养组成诊断与实验判断能力","rationale":"适合团队已有取数能力，但解释容易混淆分母、组成和因果。","now":["用相同数据让成员分解总体变化，再说明不能推出什么","将节省的例行取数时间用于异常复查和实验设计"],"oneYear":["在未见数据和口径变体上通过后，安排更独立的分析解释职责"],"threeYears":["随着工具承担更多查询与复算，按研究设计、解释和接收需求重配职责"],"assumptions":["可以获得练习数据和反馈，允许指出需求不能识别"],"tradeoffs":["培养需要业务导师；熟题答案不能证明未知情境判断"],"trigger":"以未见变体是否能守住分母和因果边界改判，不能按答案流畅度升级职责。","evidenceBasis":[{"sourceRefs":[{"module":"workunits","type":"source","id":"WU-DATA-AGENT","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"source","id":"WU-DATA-APP","version":"2026-10-21.workunits.2"}],"aiChange":"数据Agent可以缩短查询与生成分析材料的执行路径。","handoffChange":"人更早看到候选解释，需要检查比较口径和设计依据。","organizationalInference":"本站的组织建议：把培养重点从重复操作转向可核对的解释与设计。","condition":"先测完整分析质量与复核时间，确认执行省时没有转成更多返工。","limit":"来源为特定组织的数据Agent实践与实现示例，没有测本合成团队的可靠率、净人时或编制。"}]},{"id":"recruit-gap","title":"按缺口补数据治理或实验设计成员","rationale":"适合持续积压来自数据契约或识别设计，而内部培养时限不足。","now":["分别核对数据工程、指标语义与实验设计缺口","岗位样例采用同一交付契约，并核查候选人的真实贡献"],"oneYear":["新成员维护口径和设计审查，检查是否减少异常往返"],"threeYears":["按持续需求、复杂度和工具能力决定职责组合，不从模型榜单换算人数"],"assumptions":["业务量、预算与现有成员能力证据充分"],"tradeoffs":["只招会写查询或会提示词的成员可能补不上识别与治理缺口"],"trigger":"没有数据或人员材料时先完成需求和评估准备，不给出自动录用或人数建议。","evidenceBasis":[{"sourceRefs":[{"module":"workunits","type":"source","id":"WU-DATA-AGENT","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"source","id":"WU-DATA-APP","version":"2026-10-21.workunits.2"}],"aiChange":"可执行数据Agent依赖可访问的数据、上下文和检查设施。","handoffChange":"能够生成查询后，数据契约和结论接收更容易成为瓶颈。","organizationalInference":"本站的组织建议：分别定义治理与实验职责，再决定招聘、培养或工具投入。","condition":"存在持续需求且能提供授权材料核查本人贡献。","limit":"来源为特定组织的数据Agent实践与实现示例，没有测本合成团队的可靠率、净人时或编制。"}]}],"capabilityGaps":["是否有稳定指标契约与可重放输入：需核查","谁能判断何时需要实验或更多数据：需明确"],"memberRequirements":["共同输入：前期A组900/90、B组100/20；后期A组100/10、B组900/180，数字依次为访问/转化。","计算总体与组内变化，按前期组内率分解组成项，并说明能否推出干预效果。","材料为编辑构造，未运行AI，也不是授权真实成员评估。"],"memberSamples":[{"id":"A-A","name":"成员甲（合成材料）","synthetic":true,"materials":["A-A-material：可逐行核对的构造分析记录。"],"confirmed":["材料正确区分总体增加8个百分点与组内率不变，并限制因果解释。"],"selfReported":["自述有业务分析经验；未核验。"],"unverified":["未知数据、实验诊断和业务沟通仍待验证"],"nextDecision":"追加缺失分组和变分母的未见样例，再核查本人真实贡献。","artifacts":[{"id":"A-A-material","title":"组成变化分析记录（编辑构造）","format":"text","provenance":"editorial_synthetic","execution":"not_executed","lines":["编辑构造的观察数据，不是模型运行。","前期：A组900次访问、90次转化；B组100次访问、20次转化。","后期：A组100次访问、10次转化；B组900次访问、180次转化。","总体前期110/1000=11%，后期190/1000=19%，增加8个百分点。","两期A组转化率均10%，B组均20%；按前期组内率分解，组内项0，组成项8个百分点。","没有随机分配、干预或混杂控制资料，因此不能把变化称为干预提升。"]}],"claimEvidence":[{"claimIndex":0,"artifactId":"A-A-material","lineStart":2,"lineEnd":6,"support":"可依据共同输入复算总体比例与组内变化，再核对结论范围。","boundary":"仅支持这份材料中的数值与表述判断，不证明真实身份、项目经验或一般分析能力。"}]},{"id":"A-B","name":"成员乙（合成材料）","synthetic":true,"materials":["A-B-material：可逐行核对的构造分析记录。"],"confirmed":["材料正确引用总体比例，但把总体变化错误外推为所有分组提升和策略效果。"],"selfReported":["自述有业务分析经验；未核验。"],"unverified":["是否会核对分母、改正结论和识别设计资料未知"],"nextDecision":"要求用组内率修订结论；一次错误不推出完整岗位能力结论。","artifacts":[{"id":"A-B-material","title":"组成变化分析记录（编辑构造）","format":"text","provenance":"editorial_synthetic","execution":"not_executed","lines":["编辑构造的结论文本，未调用模型。","总体转化率从11%升至19%。","这证明新策略使每个分组的转化效率都提高了8个百分点。","未核对分组访问量与组内转化率，也未提供分配机制。"]}],"claimEvidence":[{"claimIndex":0,"artifactId":"A-B-material","lineStart":2,"lineEnd":4,"support":"可依据共同输入复算总体比例与组内变化，再核对结论范围。","boundary":"仅支持这份材料中的数值与表述判断，不证明真实身份、项目经验或一般分析能力。"}]}],"decisionExample":"先复算相同输入，再讨论成员分工。甲的材料可支持继续验证解释能力；乙的当前结论应返工。只有真实贡献、未见变体和业务协作材料足够后，才能推进录用或重配决定。","actualHeadcountChange":null},"progress":[],"resourceRefs":[{"module":"workunits","type":"work-unit","id":"W-A01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A02","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A03","version":"2026-10-21.workunits.2"}],"workUnitRefs":[{"module":"workunits","type":"work-unit","id":"W-A01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A02","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-A03","version":"2026-10-21.workunits.2"}],"workRelations":[],"evidenceStatus":"active","evidenceWarnings":[]}],"moduleVersions":{"journeys":"2026-10-21.journeys.2","boundary":"2026-10-19.boundary.1","frontier":"2026-10-16.1","observations":"2026-10-19.observations.2","workunits":"2026-10-21.workunits.2","tasks":"2026-10-02.2"}}