{"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":"computing","title":"计算机 · 软件维护与验证","summary":"把代码变更送到真正被接收的一步：比较偏修复实现与偏质量验证的工作，辨别团队积压究竟卡在哪里。","scope":"编辑组合的两种软件职责方向，连接C03局部维护；不覆盖全部软件职业。实际招聘要求以所核查JD为准。","caseId":"C03","forecastId":"software-maintenance","taskRef":{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},"forecastRef":{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"},"roles":[{"id":"maintenance-engineering","title":"软件维护与修复","summary":"喜欢追查现象、读旧代码并交付能回滚的修复；评价对象是被接收的变更。","duties":["复现用户问题并定位代码与依赖","实现补丁、补回归、解释未覆盖行为","与支持、审阅者协调并跟进实际接收"],"aiChanges":["在仓库、复现条件和验收可用时，可委派Agent连续定位、修改、执行检查并修正，交付一包待接收的变更","开工前确认问题未被解决；将审阅响应、失败恢复和完整复核时间纳入交付"],"humanContribution":["把模糊反馈变成可检查的变更契约","解释为何原行为保持、何时回退，回应审阅问题"],"entryRequirements":["能读懂一种项目主语言和依赖","会复现错误、Git分支和测试","按岗位核对Linux/云基础设施、语言和协作经验"],"portfolioEvidence":["一份真实或明确标为练习的issue→复现失败→补丁→回归→审阅记录","标清本人定位、修改和接收贡献，以及Agent运行与被拒绝的错误建议"],"notCovered":["C03不验证大型迁移、值班或客户沟通的全部能力","没有个人材料时不能判断你是否适合"],"sourceIds":["J-CANONICAL","J-PR"],"resourceRefs":[{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"},{"module":"observations","type":"reading","id":"reading-harness","version":"2026-10-19.observations.2"},{"module":"observations","type":"reading","id":"reading-delivery","version":"2026-10-19.observations.2"},{"module":"workunits","type":"work-unit","id":"W-C01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-C02","version":"2026-10-21.workunits.2"}],"evidenceStatus":"active","evidenceWarnings":[]},{"id":"quality-validation","title":"工程质量与验证","summary":"喜欢设计反例、诊断不稳定测试并帮助团队判断能否接收变更。","duties":["为需求和历史事故设计检查","诊断回归、环境差异和难复现问题","给出可执行的接收或补验依据"],"aiChanges":["AI可以阅读变更、设计测试候选、运行检查并按失败继续审查，输出可定位的问题和覆盖范围","验证器本身也要检查：正确替代实现能否通过，关键缺陷能否被拒收"],"humanContribution":["找出“能运行却没有解决问题”的变更","维护独立反例和可靠的测试环境，说明残余风险"],"entryRequirements":["能写可维护的测试并读产品实现","能区分测试失效、环境问题和真实缺陷","具体岗位可能要求分布式系统、性能与生产经验"],"portfolioEvidence":["一组能在旧实现失败、修复后通过的独立测试","一次包含触发条件、影响、复现与修订结果的审查记录"],"notCovered":["这是职责方向，不是统一岗位名称","Pro V2是评测维护参考，不证明此类人才需求上升"],"sourceIds":["J-PROV2","J-CANONICAL"],"resourceRefs":[{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"},{"module":"observations","type":"reading","id":"reading-harness","version":"2026-10-19.observations.2"},{"module":"observations","type":"reading","id":"reading-delivery","version":"2026-10-19.observations.2"},{"module":"workunits","type":"work-unit","id":"W-C03","version":"2026-10-21.workunits.2"}],"evidenceStatus":"active","evidenceWarnings":[]}],"jobSamples":[{"id":"canonical-3326693","title":"Software Engineer, Sustaining Engineering","company":"Canonical","url":"https://job-boards.greenhouse.io/canonical/jobs/3326693","region":"Home based – Worldwide；具体雇佣资格需核对","level":"要求职业软件工程经验，未写统一职级","publishedAt":null,"observedAt":"2026-10-16","checkedAt":"2026-10-16","status":"observed_open","statusReason":"核查时一手页面仍显示申请入口；没有提交申请或验证招聘方实际招募进度。","requirements":["职业软件工程经验；Linux经验以及Ceph/OpenStack/Kubernetes等至少一种云技术的较强经验。","在Linux上使用Python/Go/C/C++至少一种语言的较强开发经验，能够调试并使用Git。","JD列出突出学业表现；申请表要求说明高中数学/母语成绩依据与大学成绩，也明确允许无大学学历者填写no degree。","需参与国际线下活动；申请表具体询问每年2–4次、每次1–2周的承诺。"],"duties":["处理开源基础设施问题、回移补丁、与上游协作","审阅代码并和支持/产品团队沟通"],"roleIds":["maintenance-engineering","quality-validation"],"sourceId":"J-CANONICAL","limits":["这里只摘出与方向选择有关的要求，属于精选摘要，不是完整JD；申请前应逐项阅读官方职位与申请表。","JD明确要求申请用自己的文字，AI生成申请内容会被取消资格；本站只能帮助理解要求，不能代写该申请。","2026-10-16申请入口可见；发布日期、名额及全部地区雇佣条件未知。"],"verificationQuestions":["“较强经验”和学业材料要求分别有哪些证据？无学历通道不代表其他筛选条件已豁免。","你的Linux/云与调试经历能对应哪份实际工作记录？","能满足地点、出行及申请原创要求吗？","若只有脚本练习，先找更符合经验的岗位，不把本样本当默认推荐。"],"evidenceStatus":"active","evidenceWarnings":[]}],"teamScenario":{"fictional":true,"businessGoal":"合成演练：一个B2B软件团队要降低客户缺陷积压，同时保持稳定发布；不是Canonical的真实团队。","currentTeam":["2名开发轮流修复与实现功能","1名测试工程师维护回归","负责人承担需求取舍、审阅接收和发布"],"workload":"演练设定每周进入12条问题单；尚未测每条耗时、复杂度和积压原因。数字用于练习，不是现场基线。","constraints":["客户数据只能留在授权环境","缺陷需可复现，发布可回滚","工作量可能增长，省下的单项时间不直接转成减员","本期方法建议：把员工愿意委派的步骤与技术可完成性分开；复用完整人时记录核查质量、返工与接管，未测收益保持未知。"],"options":[{"id":"keep-tools","title":"保留分工，先改善工具与接收","rationale":"先判断瓶颈是取上下文、实现，还是等审阅。","now":["以可验收修复为单位委派：Agent定位、改代码、运行检查并修正；接收人独立核对关键行为","负责人处理需求冲突、失败恢复和接收队列，记录完整往返人时"],"oneYear":["若完整人时和重开率改善，将局部多文件修复纳入；每次交付仍有接收人"],"threeYears":["若多轮迁移也稳定，可让工程师协调更长维护链；架构与恢复继续有负责人"],"assumptions":["现团队具备诊断和审阅能力","能记录准备/返工/接管时间"],"tradeoffs":["改动小，但不能解决长期没人会诊断的缺口"],"trigger":"若等接收占主要历时，先调整审阅安排，不急于招聘。","evidenceBasis":[{"sourceRefs":[{"module":"boundary","type":"source","id":"BC06","version":"2026-10-19.boundary.1"},{"module":"boundary","type":"source","id":"BC13","version":"2026-10-19.boundary.1"}],"aiChange":"局部补丁可以生成，但熟悉仓库维护实验与PR接收观察测量不同，不能承诺净省时。","handoffChange":"从“给出代码”走到“复现、独立回归、回应审阅、实际接收”；等待和返工也要计时。","organizationalInference":"编辑推断：现团队先改善工具和接收队列，识别瓶颈后再讨论招人。","condition":"先有相同质量下的完整人时、重开与接收记录；若主要卡在审批，先改协作。","limit":"来源没有验证这个合成团队，也没有给出可减少人数。"}]},{"id":"train-reassign","title":"培养并重配验证职责","rationale":"当补丁变多而错误路径无人检查时，把稀缺注意力放到验证。","now":["开发与测试结对复盘2个回归事故","轮值工程师维护问题契约和失败样例库"],"oneYear":["若结对成员能独立诊断并解释拒收理由，安排其承担变更质量职责"],"threeYears":["形成能复用验证器与恢复方法的协作角色；按真实需求扩大，而非固定新增岗位"],"assumptions":["有导师时间和可练习任务","验证工作有明确授权"],"tradeoffs":["短期占用交付时间；不保证所有人都喜欢或适合职责变化"],"trigger":"看训练后的独立样例与返工变化，决定继续培养还是补人。","evidenceBasis":[{"sourceRefs":[{"module":"boundary","type":"source","id":"BC12","version":"2026-10-19.boundary.1"},{"module":"boundary","type":"source","id":"BC13","version":"2026-10-19.boundary.1"}],"aiChange":"评测修订强调干净环境复验；PR是否接收还受测试和协作影响。","handoffChange":"补丁产出扩大时，独立验证与反馈处理不能自动省略。","organizationalInference":"编辑推断：培养成员设计反例并承担明确的验证交接，比只培训生成技巧更贴合此缺口。","condition":"用未知错误路径验证能否独立诊断，并保留导师和复核成本。","limit":"基准修订是方法参考，不证明培养方案有效或验证岗位需求增加。"}]},{"id":"recruit-gap","title":"招聘已有缺陷诊断与接收能力的成员","rationale":"仅当关键能力持续缺位、培养时限不符业务需要时补人。","now":["写出使命：闭环复杂问题并建立可重复回归","用相同受限修复样例核查候选实际贡献"],"oneYear":["新成员与现团队共同承担复杂问题与知识传递；不只追求补丁数量"],"threeYears":["随产品和服务复杂度决定角色组合；编制依工作量与质量现场测量"],"assumptions":["先有预算与持续业务需求","缺口不能通过权限/排期改善消除"],"tradeoffs":["招聘和融入有成本；生成代码速度不能替代诊断、协作证据"],"trigger":"只有持续能力缺口被工作证据证实，再进入招聘，而不是看模型榜单决定。","evidenceBasis":[{"sourceRefs":[{"module":"boundary","type":"source","id":"BC06","version":"2026-10-19.boundary.1"},{"module":"boundary","type":"source","id":"BC13","version":"2026-10-19.boundary.1"}],"aiChange":"生成与接收、真实效率之间仍有差距，模型分数不能定位本团队缺人原因。","handoffChange":"需要有人持续完成诊断、回归与接收协调；这些接口必须有可核查工作证据。","organizationalInference":"编辑推断：仅在持续诊断/协作缺口且培养时限不足时招聘相应成员。","condition":"证明缺口不是重复任务、权限、排期或接收阻塞；核查候选本人贡献。","limit":"研究不能推该公司的编制、招聘量或候选适配度。"}]}],"capabilityGaps":["缺陷能否被可靠复现：待测","测试是否覆盖历史事故与错误路径：待核","接收队列的等待责任是否清楚：优先核查"],"memberRequirements":["相同构造练习：first_upper 接受字符串列表；空列表应返回 {\"error\":\"EMPTY\"}，正常输入仍返回首项的大写值。","提交可查看的修改、旧/新复现与测试记录、未覆盖输入和接收范围；说明本人判断与工具建议的区分。","全部人物、对话、代码场景均为编辑构造。仅甲的玩具代码在本地复算过；这不证明真人经验或AI修复能力。"],"memberSamples":[{"id":"C-A","name":"成员甲（合成材料）","synthetic":true,"materials":["C-A-diff：可逐行查看的构造补丁。","C-A-log：同一玩具函数修改前后3组输入的实际本地日志。","C-A-notes：明确标为编辑构造的建议、否决理由与未覆盖范围。"],"confirmed":["玩具日志显示：空列表从IndexError改为约定错误；两组正常输入结果保持。","复核笔记说明了否决吞异常建议的契约理由；只能核对笔记内容，不能证明真人作出了该判断。"],"selfReported":["自述能独立处理生产事故"],"unverified":["没有跨模块、真实用户环境、生产回滚和长期维护证据。","构造笔记不能证明真人调试判断、沟通能力或实际个人贡献。"],"nextDecision":"优先进行一次含中断/恢复条件的同标准工作样例与材料追问；不能仅据练习录用。","artifacts":[{"id":"C-A-diff","title":"构造补丁：仅增加空列表分支","format":"diff","provenance":"editorial_synthetic","execution":"not_executed","lines":["--- before.py","+++ after.py","@@ -1,2 +1,4 @@"," def first_upper(items):","+    if not items:","+        return {\"error\": \"EMPTY\"}","     return {\"value\": items[0].upper()}"]},{"id":"C-A-log","title":"玩具代码旧/新运行日志","format":"text","provenance":"editorial_synthetic","execution":"local_fixture_execution","lines":["本地玩具夹具执行记录；Python；2026-10-16；非AI任务实验。","before input=[] -> IndexError: list index out of range","before input=[\"a\"] -> {\"value\": \"A\"}","before input=[\"b\", \"c\"] -> {\"value\": \"B\"}","after input=[] -> {\"error\": \"EMPTY\"}","after input=[\"a\"] -> {\"value\": \"A\"}","after input=[\"b\", \"c\"] -> {\"value\": \"B\"}","after acceptance: [] error=EMPTY; [\"a\"] value=A; [\"b\",\"c\"] value=B; 3/3 fixture checks passed.","范围：只运行上述3个输入；未验证非列表、非字符串、集成、生产恢复和多人协作。"]},{"id":"C-A-notes","title":"构造的工具建议与复核笔记","format":"text","provenance":"editorial_synthetic","execution":"not_executed","lines":["以下为编辑写作的合成对话与本人笔记；没有实际调用模型，也没有真实候选人。","工具建议（构造）：用 try/except Exception 包住函数，异常时返回空字符串。","成员甲的复核笔记（构造）：否决该建议；吞掉任意错误会隐藏类型问题，且空字符串不符合 error=EMPTY 契约。","采用范围（构造）：只给空列表增加明确分支；保留正常输入行为。","本人/工具区分：在这个故事中，工具给建议，甲提出契约反例；真实个人贡献仍无法由构造材料证明。","待补：非字符串输入政策、集成调用者、发布和回滚路径。"]}],"claimEvidence":[{"claimIndex":0,"artifactId":"C-A-log","lineStart":2,"lineEnd":8,"support":"能直接比较同一输入的before/after与限定检查。","boundary":"只证明这段构造代码在3个输入上的本地行为，非完整项目或真人能力。"},{"claimIndex":1,"artifactId":"C-A-notes","lineStart":1,"lineEnd":5,"support":"笔记明确列出建议、拒绝理由和拟采用范围。","boundary":"对话与人物由编辑构造，不是实际AI会话、面试或候选贡献证据。"}]},{"id":"C-B","name":"成员乙（合成材料）","synthetic":true,"materials":["C-B-output：实际可查看的一段展示文字；只有补丁片段和“全部测试通过”的声明。"],"confirmed":["展示文本确实含一行代码和通过声明；材料没有附可复现的测试依据。"],"selfReported":["自述大幅提高开发效率"],"unverified":["个人贡献、原缺陷是否修复、测试范围、返工和接收均未知"],"nextDecision":"先请求其授权的代码差异与复现/测试记录，允许用等价样例补证；未知不判为能力低。","artifacts":[{"id":"C-B-output","title":"薄材料：补丁展示片段与通过声明","format":"text","provenance":"editorial_synthetic","execution":"not_executed","lines":["编辑构造的成员乙展示文本；非截图、非模型输出，未执行。","补丁展示：return {\"value\": items[0].upper()}","展示说明：已使用AI完成修复，全部测试通过。","未提供：完整函数、代码版本、输入、测试命令、日志和审阅记录。"]}],"claimEvidence":[{"claimIndex":0,"artifactId":"C-B-output","lineStart":1,"lineEnd":4,"support":"读者可直接确认这里提供了什么、缺了什么。","boundary":"通过声明未经验证；片段不显示完整上下文，不能据此认定已修复或评价完整岗位能力。"}]}],"decisionExample":"本演练优先用“保留分工+重配验证”收集瓶颈证据；成员甲值得进一步做同标准验证，成员乙需补产物与贡献证据。两者都未形成录用结论；真实人员资料不上传本站。","actualHeadcountChange":null},"progress":[{"id":"pro-v2-revision","title":"评测协议变了，成绩曲线也要换读法","concept":"软件基准把问题和测试打包让代理修复；如果题意与测试不一致，分数会混入评测缺陷。","whatChanged":"发布方在2026-09-22将公开题集由731缩为642题，并更正69题说明；同时限制联网、用干净环境重新评补丁。","originalPublishedAt":null,"revisionAt":"2026-09-22","checkedAt":"2026-10-16","dateExplanation":"这是旧基准的新修订；本站旧资料已登记，本轮对齐版本关系，不称新发现。","taskImpact":["比较维护能力时同时看题集、网络和重新评测规则","选人或组队时要求可复现验证依据，不只看漂亮通过率"],"limits":["发布方仍列有信任中继与验证器执行补丁代码等剩余限制","本站未复跑所有任务；V1/V2不能直接作同协议差值"],"differentResults":["更严格协议可能暴露以前未计入的失败；分数变化不等于模型变弱"],"watchNext":["关注同题集同协议的新运行和真实项目接收成本"],"sourceIds":["J-PROV2"],"resourceRefs":[{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"}],"evidenceStatus":"active","evidenceWarnings":[]},{"id":"pr-integration","title":"修复PR合并，还取决于检查和协作","concept":"PR是送给维护者审阅的改动；合并表示该项目接收，不自动表示节省总人时。","whatChanged":"研究在8106个修复PR中观察到65.0%已合并，另有人工作了326个关闭未合并PR的原因分析；重复解决、测试问题和审阅流程都可能影响接收。","originalPublishedAt":"2026-01-29","revisionAt":null,"checkedAt":"2026-10-16","dateExplanation":"2026年1月公开的研究，2026-10-16补读；不是本周的新能力突破。","taskImpact":["交付应包括复现、回归和审阅响应","开工前检查是否已由别的改动解决，可减少重复劳动"],"limits":["65.0%的分母包含仍开放PR；不是个人成功概率","没有随机匹配项目与代理，也没测本团队完整净人时"],"differentResults":["未合并既可能是错误，也可能是重复、优先级或协作；不能统称AI修复失败"],"watchNext":["对自己的工作记录技术失败、流程等待和重复关闭，分别决定改善方法"],"sourceIds":["J-PR"],"resourceRefs":[{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"}],"evidenceStatus":"active","evidenceWarnings":[]},{"id":"harness-method","title":"模型之外，运行方法也会改变交付","concept":"固定执行模型，比较运行规则与检查方式。","whatChanged":"Harness²研究候选运行系统的执行对照与选择性经验复用。","originalPublishedAt":"2026-10-09","revisionAt":null,"checkedAt":"2026-10-19","dateExplanation":"10月19日补充阅读；研究原始公开日如上，补录日不是新能力发布时间。","taskImpact":["核对原始输入、工具条件、产物验收与人的贡献","技术能力、个人效果、组织效果和劳动市场结论分层判断"],"limits":["本站未复现实验，未验证当前团队或个人收益","方法相邻，不能推算职业替代或招聘人数"],"differentResults":["更复杂流程也可能增加成本或引入错误"],"watchNext":["先形成可检查的作品或有限试点安排，必要时保留原决定"],"sourceIds":["R-HARNESS2"],"resourceRefs":[{"module":"observations","type":"reading","id":"reading-harness","version":"2026-10-19.observations.2"},{"module":"observations","type":"reading","id":"reading-delivery","version":"2026-10-19.observations.2"}],"evidenceStatus":"active","evidenceWarnings":[]}],"resourceRefs":[{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},{"module":"frontier","type":"forecast","id":"software-maintenance","version":"2026-10-16.1"},{"module":"observations","type":"reading","id":"reading-harness","version":"2026-10-19.observations.2"},{"module":"observations","type":"reading","id":"reading-delivery","version":"2026-10-19.observations.2"},{"module":"workunits","type":"work-unit","id":"W-C01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-C02","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-C03","version":"2026-10-21.workunits.2"}],"workUnitRefs":[{"module":"workunits","type":"work-unit","id":"W-C01","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-C02","version":"2026-10-21.workunits.2"},{"module":"workunits","type":"work-unit","id":"W-C03","version":"2026-10-21.workunits.2"}],"workRelations":[{"from":{"module":"tasks","type":"task","id":"DEV-21670","version":"2026-10-02.2"},"to":{"module":"boundary","type":"task","id":"C03","version":"2026-10-19.boundary.1"},"relation":"partial","reason":"正式软件维护任务与C03的复现、修复、回归相交；C03是编辑案例，未覆盖该任务全部原子和组织要求。"}],"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"}}