[ { "number": 1, "collection": "吴军作品合集", "book": "文明之光", "chapter": "引子 / 文明史视角", "title": "长周期视角", "interview_question": "为什么产品经理需要长周期视角？", "short_answer": "长周期视角帮助我们跳出单点功能和短期指标，从文明、制度、技术基础设施和社会协作方式看产品机会。", "key_points": [ "技术变迁通常嵌在更长的制度、商业和文化演化里", "短期爆发背后往往有长期积累", "面试中要能把一个功能放到行业阶段和基础设施成熟度里解释" ], "follow_ups": [ "长周期视角会不会让判断太宏观？", "如何把历史视角落到产品决策？", "AI 产品现在处于哪个历史阶段？" ], "project_mapping": [ "分析 AI Agent 时区分模型能力、工具生态、算力成本、用户习惯四条演进曲线", "复盘一个项目时说明它解决的是阶段性问题还是长期结构问题" ], "aipm_transfer": "AI PM 要能判断哪些机会只是热潮，哪些是由基础设施变化带来的长期方向。", "tags": [ "long-term-view", "civilization", "strategy" ], "priority": "P0" }, { "number": 2, "collection": "吴军作品合集", "book": "文明之光", "chapter": "科学之路 / 理性时代", "title": "科学方法与产品验证", "interview_question": "科学方法对产品验证有什么启发？", "short_answer": "科学方法强调假设、证据、可证伪和持续修正。产品验证也应从观点之争转为实验之争，用可观察指标校正直觉。", "key_points": [ "先提出清晰假设，再设计可验证路径", "证据强度比叙事漂亮更重要", "失败实验不是失败，而是缩小不确定性" ], "follow_ups": [ "如何设计一个可证伪的 PRD 假设？", "什么样的用户反馈不能算验证？", "产品实验和科学实验有什么区别？" ], "project_mapping": [ "为 AI 面试工具定义假设：结构化卡片能提高复习效率", "用留存、采纳率、复述质量验证", "记录失败样例而不是只展示成功案例" ], "aipm_transfer": "AI PM 做模型和功能评估时，要坚持假设驱动，而不是被 demo 的流畅感带走。", "tags": [ "scientific-method", "product-validation", "experiment" ], "priority": "P0" }, { "number": 3, "collection": "吴军作品合集", "book": "文明之光", "chapter": "交通、通信、计算、互联网", "title": "距离被技术缩短", "interview_question": "交通和通信技术如何改变商业模式？", "short_answer": "当信息、商品或人的流动成本下降，新的协作范围和商业模式就会出现。技术价值常常体现在缩短距离、降低交易成本和扩大市场半径。", "key_points": [ "铁路、电报、电话、计算机、互联网本质上都在重塑连接效率", "连接效率提升会改变组织规模和平台形态", "产品机会常出现在旧流程的距离成本被重新压缩时" ], "follow_ups": [ "AI 正在缩短什么距离？", "交易成本降低后谁会受益？", "如何识别连接效率带来的平台机会？" ], "project_mapping": [ "把 AI Agent 理解为缩短“意图到执行”的距离", "把求职素材库理解为缩短“阅读到面试表达”的距离" ], "aipm_transfer": "AI PM 要寻找那些因为理解、生成、执行成本下降而被重新设计的流程。", "tags": [ "transaction-cost", "connection", "platform" ], "priority": "P0" }, { "number": 4, "collection": "吴军作品合集", "book": "文明之光", "chapter": "计算的时代", "title": "计算成为基础设施", "interview_question": "计算技术从工具变成基础设施意味着什么？", "short_answer": "当计算无处不在，它不再只是效率工具，而成为产品、组织和社会运行的底座。产品经理要从功能思维升级到基础设施思维。", "key_points": [ "基础设施的价值在于稳定、低成本、可组合", "计算普及会让更多行业软件化", "新一代基础设施会催生新的产品范式" ], "follow_ups": [ "AI 会成为类似计算的基础设施吗？", "基础设施产品和应用产品有什么不同？", "如何判断一个技术是否已基础设施化？" ], "project_mapping": [ "把 LLM 能力拆成推理、生成、检索、工具调用等基础能力，再思考上层应用", "为内部团队建设可复用 AI 能力层" ], "aipm_transfer": "AI PM 要区分“做一个 AI 功能”和“沉淀一层 AI 能力基础设施”。", "tags": [ "computing", "infrastructure", "platform" ], "priority": "P0" }, { "number": 5, "collection": "吴军作品合集", "book": "文明之光", "chapter": "华尔街 / 泡沫 / 增长极限", "title": "繁荣中的风险", "interview_question": "技术浪潮里为什么容易出现泡沫？", "short_answer": "新技术会创造真实机会，也会制造叙事溢价。泡沫常来自长期方向正确但短期商业化、成本结构和用户价值尚未兑现。", "key_points": [ "方向正确不代表所有公司都能活下来", "资本、媒体和用户预期会放大短期波动", "风险管理要看单位经济、真实需求和竞争壁垒" ], "follow_ups": [ "AI 泡沫和真实机会怎么区分？", "产品经理如何在热潮中保持判断？", "什么指标能证明不是概念驱动？" ], "project_mapping": [ "评估 AI 产品时检查复用率、付费意愿、交付成本和替代方案", "避免用融资热度或模型名称替代用户价值" ], "aipm_transfer": "AI PM 面试中要展示对技术乐观、对商业验证谨慎的双重能力。", "tags": [ "bubble", "risk", "unit-economics" ], "priority": "P0" }, { "number": 6, "collection": "吴军作品合集", "book": "文明之光", "chapter": "硅谷 / 互联网时代", "title": "创新生态的复利", "interview_question": "为什么创新常常发生在生态而不是单个公司里？", "short_answer": "持续创新依赖人才、资本、市场、大学、失败容忍和开放网络的组合。单个公司成功往往是生态复利的结果。", "key_points": [ "创新不是孤立英雄故事，而是网络效应", "人才流动和资本循环会加速知识传播", "开放协作能提高试错密度" ], "follow_ups": [ "如何判断一个地区或平台有创新生态？", "生态和平台有什么区别？", "公司如何借外部生态创新？" ], "project_mapping": [ "为 AI PM 学习路线搭建资料、项目、评测、发布和反馈生态", "产品上开放插件、模板和社区案例，让用户参与扩展" ], "aipm_transfer": "AI PM 要设计的不只是功能，还可能是让用户和开发者共同演化的生态。", "tags": [ "innovation-ecosystem", "silicon-valley", "network" ], "priority": "P1" }, { "number": 7, "collection": "吴军作品合集", "book": "智能时代", "chapter": "数据是文明基石", "title": "数据为什么是智能产品的基础？", "interview_question": "智能能力不是凭空出现的，数据把现象转化为可计算、可学习、可优化的对象。没有高质量数据，模型和产品都难以持续改进。", "short_answer": "数据连接现象、信息、知识和决策；数据质量决定模型上限；产品要设计数据产生、标注、反馈和治理闭环", "key_points": [ "数据多是否一定更好？", "什么是 AI 产品的数据飞轮？", "如何设计用户反馈数据？" ], "follow_ups": [ "面试素材库记录卡片阅读、收藏、复述和面试命中情况", "用这些反馈优化后续推荐和拆解优先级" ], "project_mapping": [ "AI PM 要把数据闭环写进产品设计，而不是上线后再想埋点。" ], "aipm_transfer": [ "data", "feedback-loop", "ai-product" ], "tags": [ "data-intelligence", "ai-era", "data", "product" ], "priority": "P0" }, { "number": 8, "collection": "吴军作品合集", "book": "智能时代", "chapter": "大数据和机器智能", "title": "把智能问题变成数据问题", "interview_question": "如何理解“把智能问题变成数据问题”？", "short_answer": "很多看似需要人工智能的任务，可以通过足够数据、统计建模和反馈优化转化为可计算问题。关键是重新定义输入、输出和评估。", "key_points": [ "先明确可观测信号，再建立预测或决策目标", "相关性在很多业务任务中比完整因果更可用", "智能产品要把模糊判断拆成数据任务" ], "follow_ups": [ "所有智能问题都能数据化吗？", "数据问题和因果问题怎么区分？", "LLM 时代这句话还成立吗？" ], "project_mapping": [ "把“候选人是否适合岗位”拆成 JD 技能、项目证据、表达匹配和风险项", "用结构化评分辅助判断" ], "aipm_transfer": "AI PM 要会把业务问题翻译成模型可处理的数据结构和评价任务。", "tags": [ "machine-intelligence", "datafication", "problem-framing" ], "priority": "P0" }, { "number": 9, "collection": "吴军作品合集", "book": "智能时代", "chapter": "思维的革命", "title": "从因果到相关", "interview_question": "大数据思维为什么强调相关性？", "short_answer": "在复杂系统中，完整因果链很难获得，强相关关系可以先帮助预测和决策。但相关不等于因果，适合优化，不适合替代全部解释。", "key_points": [ "相关性适合推荐、搜索、风控、预测等任务", "因果适合干预、政策和长期战略", "产品决策要知道自己是在预测还是在解释" ], "follow_ups": [ "什么时候相关性足够？", "什么时候必须做因果验证？", "A/B 测试在这里扮演什么角色？" ], "project_mapping": [ "用行为数据预测用户下一张该复习的卡", "对产品改版效果用实验验证因果影响" ], "aipm_transfer": "AI PM 要能区分 prediction、recommendation 和 intervention，避免把相关当因果。", "tags": [ "correlation", "causality", "data-thinking" ], "priority": "P0" }, { "number": 10, "collection": "吴军作品合集", "book": "智能时代", "chapter": "大数据与商业", "title": "数据流中的商业机会", "interview_question": "为什么大数据商业的共同点在数据流中？", "short_answer": "持续的数据流让企业能感知需求、预测变化、个性化服务并快速迭代。商业价值来自对数据流的采集、解释和行动闭环。", "key_points": [ "静态数据价值有限，实时反馈才形成运营能力", "个性化依赖用户行为流", "数据流能把粗放业务变成精细化业务" ], "follow_ups": [ "数据流和数据资产有什么区别？", "怎样判断一个产品有数据飞轮？", "隐私约束下如何设计数据流？" ], "project_mapping": [ "面试准备产品从一次性文档库升级为学习过程系统：记录目标、练习、反馈和命中率" ], "aipm_transfer": "AI PM 要设计“用户行动 -> 数据回流 -> 模型/内容优化 -> 更好行动”的闭环。", "tags": [ "data-flow", "personalization", "business-model" ], "priority": "P0" }, { "number": 11, "collection": "吴军作品合集", "book": "智能时代", "chapter": "现有产业 + 机器智能", "title": "智能化产业升级公式", "interview_question": "如何用“现有产业 + 机器智能 = 新产业”分析机会？", "short_answer": "智能化不是单独创造一个 AI 行业，而是让农业、医疗、教育、招聘、内容等既有行业以更精细、更个性化、更自动化的形态重组。", "key_points": [ "先找行业中的低效决策和重复劳动", "再判断是否有数据、模型和执行通道", "最后设计人机协作而不是简单替代人" ], "follow_ups": [ "哪些行业最适合被 AI 重做？", "如何判断 AI 是增强还是替代？", "产业升级会遇到什么阻力？" ], "project_mapping": [ "把求职辅导拆成信息收集、能力诊断、材料生成、模拟面试和反馈跟踪，用 AI 分层增强" ], "aipm_transfer": "AI PM 面试可以用这个公式快速分析行业机会和产品切入点。", "tags": [ "industry-upgrade", "machine-intelligence", "opportunity" ], "priority": "P0" }, { "number": 12, "collection": "吴军作品合集", "book": "智能时代", "chapter": "智能革命和未来社会", "title": "前 2% 与技能迁移", "interview_question": "智能时代个人如何避免被浪潮淘汰？", "short_answer": "关键不是预测每个岗位消失与否，而是尽快掌握能与机器协作的新技能，成为会定义问题、使用工具、验证结果和整合资源的人。", "key_points": [ "技术革命会改变能力溢价", "重复执行能力贬值，问题定义和系统整合升值", "个人要用新工具重做自己的工作流" ], "follow_ups": [ "AI 会替代产品经理吗？", "个人如何进入前 2%？", "哪些能力更抗自动化？" ], "project_mapping": [ "把读书、项目复盘、JD 拆解、面试训练都变成 AI 协作流程", "沉淀自己的案例库和工具链" ], "aipm_transfer": "AI PM 的竞争力来自理解业务、驾驭 AI 工具和建立验证闭环的组合。", "tags": [ "career", "ai-era", "skill-migration" ], "priority": "P0" }, { "number": 13, "collection": "吴军作品合集", "book": "大学之路", "chapter": "选校框架", "title": "选大学的方法对职业选择有什么启发？", "interview_question": "选校本质上是多目标匹配：个人目标、资源环境、文化适配、长期发展和机会成本。职业选择也应使用类似框架，而不是只追排名。", "short_answer": "排名只是单一代理指标；适配度来自目标、能力、环境和路径；长期成长空间比短期名气更重要", "key_points": [ "如何把选校框架迁移到选公司？", "AI PM 该优先选平台还是岗位？", "如何避免被排名或品牌锚定？" ], "follow_ups": [ "求职时建立岗位评分卡：行业阶段、团队资源、学习密度、项目所有权、候选人匹配度" ], "project_mapping": [ "AI PM 可以用多维匹配模型评估岗位，而不是只看公司名和薪资。" ], "aipm_transfer": [ "college-choice", "career-choice", "fit" ], "tags": [ "career", "education" ], "priority": "P1" }, { "number": 14, "collection": "吴军作品合集", "book": "大学之路", "chapter": "教育资源与环境", "title": "环境为什么会影响长期成长？", "interview_question": "好的环境提供同伴、师资、项目、信息和标准，它改变人的机会集合与默认行为。个人成长不是只靠自驱，也受系统环境影响。", "short_answer": "同伴网络会提高目标和标准；资源密度影响试错机会；环境的隐性规则会塑造行为", "key_points": [ "怎么判断一个团队是否适合成长？", "远程时代环境还重要吗？", "如何补足环境劣势？" ], "follow_ups": [ "找 AI PM 岗位时看是否有真实模型项目、数据资源、工程协作和用户反馈，而不只看 title" ], "project_mapping": [ "AI PM 要主动为自己创造高质量环境：项目、导师、反馈、作品集和同行讨论。" ], "aipm_transfer": [ "environment", "growth", "career" ], "tags": [ "career", "education" ], "priority": "P1" }, { "number": 15, "collection": "吴军作品合集", "book": "大学之路", "chapter": "通识教育与专业能力", "title": "产品经理为什么需要通识能力？", "interview_question": "产品工作常在技术、商业、用户和组织之间做翻译。通识能力让人更容易理解不同领域的约束，也更容易发现跨界机会。", "short_answer": "专业能力决定做事深度；通识能力决定连接广度；AI 时代跨学科表达和判断变得更重要", "key_points": [ "通识和专业如何平衡？", "AI PM 需要懂到多技术？", "通识能力如何在面试中体现？" ], "follow_ups": [ "用技术史解释 AI 浪潮，用行为科学解释用户信任，用商业模式解释产品收费" ], "project_mapping": [ "AI PM 的优势往往来自跨域整合：模型、体验、数据、商业、合规和组织。" ], "aipm_transfer": [ "liberal-arts", "cross-domain", "product-manager" ], "tags": [ "career", "education", "product" ], "priority": "P0" }, { "number": 16, "collection": "吴军作品合集", "book": "大学之路", "chapter": "长期主义教育投资", "title": "如何看待学习投入的长期回报？", "interview_question": "教育和能力建设的回报常常滞后，但会通过认知框架、选择质量和机会网络复利释放。短期功利容易错过底层能力。", "short_answer": "底层能力比短期技巧更耐用；复利来自持续积累和迁移；学习投资要兼顾当下可用和长期资产", "key_points": [ "面试准备该学技巧还是底层知识？", "如何判断一个学习主题值得投入？", "长期主义会不会降低短期效率？" ], "follow_ups": [ "把书籍拆解成长期素材库，同时为近期面试生成 P0 卡片和答题框架" ], "project_mapping": [ "AI PM 要把学习资料产品设计成资产积累系统，而不是一次性刷题工具。" ], "aipm_transfer": [ "education-investment", "compound-growth", "learning" ], "tags": [ "career", "education" ], "priority": "P0" }, { "number": 17, "collection": "吴军作品合集", "book": "见识", "chapter": "商业的本质", "title": "如何理解商业的本质？", "interview_question": "商业的本质是创造价值、交换价值和分配价值。看一个产品不能只看功能，还要看它为谁创造了可持续价值，以及价值如何被捕获。", "short_answer": "用户价值是基础，但不等于商业价值；支付方、使用方和受益方可能不同；可持续商业需要成本结构和规模化路径", "key_points": [ "AI 产品如何证明商业价值？", "免费用户多是否等于成功？", "B 端和 C 端价值捕获有什么不同？" ], "follow_ups": [ "为面试素材库区分个人用户、求职机构、企业培训三类支付场景", "分别设计价值指标" ], "project_mapping": [ "AI PM 要能同时讲用户价值、业务指标和成本结构。" ], "aipm_transfer": [ "business", "value-creation", "monetization" ], "tags": [ "judgment", "business", "business-model" ], "priority": "P0" }, { "number": 18, "collection": "吴军作品合集", "book": "见识", "chapter": "命和运", "title": "个人发展中如何理解选择与机会？", "interview_question": "个人发展既有结构性约束，也有可主动选择的路径。好的策略是提高自己遇到机会、识别机会和承接机会的概率。", "short_answer": "不能把结果完全归因于努力，也不能把机会当借口；选择环境、行业和同行会改变机会分布；能力建设让机会来时可承接", "key_points": [ "如何在面试中讲职业转型？", "选择赛道是否比努力更重要？", "如何提高机会密度？" ], "follow_ups": [ "选择 AI PM 方向时，用行业增长、个人经验、项目证据和学习速度共同论证" ], "project_mapping": [ "AI PM 转型要把过去经验转成可迁移资产，而不是从零讲起。" ], "aipm_transfer": [ "career-strategy", "opportunity", "positioning" ], "tags": [ "judgment", "business" ], "priority": "P1" }, { "number": 19, "collection": "吴军作品合集", "book": "见识", "chapter": "人生的智慧", "title": "为什么见识会影响决策质量？", "interview_question": "见识决定一个人能看到多少变量、多少时间尺度和多少替代方案。产品经理的判断力很大程度来自见识的广度和结构化程度。", "short_answer": "见识不是信息量，而是解释框架；高质量输入会改善选择集合；跨领域案例能帮助识别相似结构", "key_points": [ "如何提升产品见识？", "读书如何转化为面试表达？", "见识和经验有什么区别？" ], "follow_ups": [ "把每本书拆成可复用决策框架，而不是收藏摘抄", "每张卡都绑定项目例子" ], "project_mapping": [ "AI PM 要持续扩展案例库，用更多结构识别更复杂的问题。" ], "aipm_transfer": [ "judgment", "mental-model", "reading" ], "tags": [ "judgment", "business" ], "priority": "P0" }, { "number": 20, "collection": "吴军作品合集", "book": "见识", "chapter": "拒绝伪工作", "title": "如何识别产品工作中的伪勤奋？", "interview_question": "伪勤奋常表现为忙于产出文档、会议和局部优化，却没有推动关键指标或核心风险降低。真正的工作要让问题更清楚、结果更可验证。", "short_answer": "产出不等于进展；忙碌可能掩盖目标不清；高质量工作会减少不确定性或创造可验证价值", "key_points": [ "PRD 写得多就好吗？", "如何判断一个 PM 的真实贡献？", "AI 会放大伪工作吗？" ], "follow_ups": [ "用 AI 生成文档后仍要验证用户问题、决策依据和上线指标", "对每个任务定义验收标准" ], "project_mapping": [ "AI PM 更要防止用更快的生成能力制造更多低价值文本。" ], "aipm_transfer": [ "productivity", "anti-pattern", "execution" ], "tags": [ "judgment", "business", "product" ], "priority": "P0" }, { "number": 21, "collection": "吴军作品合集", "book": "见识", "chapter": "边界与取舍", "title": "为什么优秀产品经理要懂取舍？", "interview_question": "资源、注意力、时间和信任都是有限的。懂取舍不是少做，而是围绕目标选择最有杠杆的事，并愿意放弃低价值选项。", "short_answer": "取舍需要明确目标函数；拒绝需求要有依据；边界清楚反而能提高交付质量", "key_points": [ "如何拒绝老板或客户的需求？", "优先级如何排序？", "AI 产品初期该砍什么？" ], "follow_ups": [ "为 AI 面试工具先做资料拆解和问答卡，不急着做社交、社区或复杂推荐" ], "project_mapping": [ "AI PM 要在无限可能中定义 MVP 边界和风险边界。" ], "aipm_transfer": [ "prioritization", "tradeoff", "scope" ], "tags": [ "judgment", "business", "product" ], "priority": "P0" }, { "number": 22, "collection": "吴军作品合集", "book": "数学之美", "chapter": "统计语言模型", "title": "统计语言模型对今天的 LLM 有什么启发？", "interview_question": "早期统计语言模型把语言问题转成概率问题，今天的 LLM 更强大，但核心仍是基于上下文预测和生成。理解概率视角有助于理解模型能力和局限。", "short_answer": "语言可以被建模为序列概率；上下文决定预测分布；流畅输出不等于事实正确", "key_points": [ "N-gram 和 LLM 有什么连续性？", "概率模型为什么会胡说？", "如何向非技术面试官解释语言模型？" ], "follow_ups": [ "解释 AI 面试回答生成时区分语言流畅、事实依据和任务适配三个层次" ], "project_mapping": [ "AI PM 不必只背大模型名词，要能讲清楚语言建模的基本直觉。" ], "aipm_transfer": [ "language-model", "probability", "llm" ], "tags": [ "math-for-ai", "nlp", "ai-product" ], "priority": "P0" }, { "number": 23, "collection": "吴军作品合集", "book": "数学之美", "chapter": "中文分词 / HMM", "title": "NLP 为什么需要把语言结构化？", "interview_question": "分词、序列标注和隐马尔可夫模型说明，很多 NLP 任务要先把自然语言转成模型能处理的结构。结构化是理解和计算的桥梁。", "short_answer": "中文没有天然空格，需要算法判断边界；序列问题常依赖上下文状态；结构化结果会影响后续搜索、分类和推荐", "key_points": [ "LLM 时代还需要分词吗？", "结构化抽取和生成式问答如何结合？", "为什么 schema 很重要？" ], "follow_ups": [ "把 JD 拆成技能、职责、行业、工具、软素质字段", "再用这些字段生成面试卡" ], "project_mapping": [ "AI PM 设计信息抽取产品时，要重视 schema 和下游任务，而不是只看生成效果。" ], "aipm_transfer": [ "nlp", "structured-extraction", "schema" ], "tags": [ "math-for-ai", "nlp" ], "priority": "P0" }, { "number": 24, "collection": "吴军作品合集", "book": "数学之美", "chapter": "信息论", "title": "信息论怎样帮助理解产品沟通？", "interview_question": "信息论关注不确定性的减少。好的产品文案、数据报告和 AI 输出，本质上都要减少用户完成任务的不确定性。", "short_answer": "信息价值来自降低不确定性；冗余有时能提高可靠性；压缩不是删短，而是保留任务所需信息", "key_points": [ "为什么长文档不一定信息量大？", "AI 摘要如何评估？", "沟通中的冗余什么时候有价值？" ], "follow_ups": [ "为面试卡设计 30 秒回答、展开要点、追问和项目映射四层信息密度" ], "project_mapping": [ "AI PM 要把输出设计为信息产品：不同层级服务不同决策时刻。" ], "aipm_transfer": [ "information-theory", "communication", "summary" ], "tags": [ "math-for-ai", "nlp", "product" ], "priority": "P0" }, { "number": 25, "collection": "吴军作品合集", "book": "数学之美", "chapter": "搜索引擎索引", "title": "搜索引擎索引对知识库有什么启发？", "interview_question": "高效检索依赖好的表示、索引和排序。知识库不是把资料堆在一起，而是让用户能在正确时刻找到正确内容。", "short_answer": "布尔检索强调精确匹配；倒排索引降低查找成本；排序把可找到变成可使用", "key_points": [ "知识库为什么需要标签和索引？", "搜索和推荐有什么区别？", "如何评估资料库检索质量？" ], "follow_ups": [ "为书籍卡片建立书名、主题、角色、优先级和面试问题字段", "支持关键词和标签检索" ], "project_mapping": [ "AI PM 做知识产品要先做好信息架构，再叠加语义搜索或 Agent。" ], "aipm_transfer": [ "search", "index", "knowledge-base" ], "tags": [ "math-for-ai", "nlp" ], "priority": "P0" }, { "number": 26, "collection": "吴军作品合集", "book": "数学之美", "chapter": "PageRank / 相关性", "title": "相关性排序为什么是搜索产品的核心？", "interview_question": "搜索不是返回包含关键词的内容，而是按用户意图、内容质量和上下文排序。相关性排序决定用户是否觉得系统聪明。", "short_answer": "关键词匹配只是召回；排序要综合文本、链接、质量和行为信号；评估要看点击、满意度和任务完成", "key_points": [ "AI 搜索如何做排序？", "RAG 的召回结果如何重排？", "相关性和权威性冲突怎么办？" ], "follow_ups": [ "面试准备站点可按目标岗位、近期学习、P0 优先级和弱项主题排序卡片" ], "project_mapping": [ "AI PM 需要理解召回与排序分层，这对 RAG、推荐和搜索都通用。" ], "aipm_transfer": [ "ranking", "relevance", "search-product" ], "tags": [ "math-for-ai", "nlp", "product" ], "priority": "P0" }, { "number": 27, "collection": "吴军作品合集", "book": "数学之美", "chapter": "最大熵 / 分类", "title": "为什么模型要在约束下保持不过度自信？", "interview_question": "最大熵思想提醒我们：在已知约束之外不要额外假设。产品和模型评估也应避免用不足证据做过度判断。", "short_answer": "约束越少，结论越应保守；模型输出应表达不确定性；过拟合常来自把噪声当规律", "key_points": [ "最大熵如何通俗解释？", "AI 产品如何展示不确定性？", "分类模型为什么会过拟合？" ], "follow_ups": [ "候选人匹配评分给出证据和置信度，缺少证据时不强行判断" ], "project_mapping": [ "AI PM 要推动模型输出从“断言式”转为“证据 + 置信 + 下一步验证”。" ], "aipm_transfer": [ "maximum-entropy", "uncertainty", "classification" ], "tags": [ "math-for-ai", "nlp" ], "priority": "P1" }, { "number": 28, "collection": "吴军作品合集", "book": "数学之美", "chapter": "布隆过滤器 / 信息指纹", "title": "工程中的概率数据结构有什么产品价值？", "interview_question": "布隆过滤器和信息指纹展示了工程设计中的取舍：用可接受的误差换取速度、空间和规模化能力。产品设计也常需要定义可接受误差。", "short_answer": "不是所有系统都追求绝对精确；误判类型和业务后果决定方案；规模化系统常在准确率、成本和延迟间权衡", "key_points": [ "什么时候可以接受误判？", "如何向业务解释概率性系统？", "AI 安全里误杀和漏放怎么取舍？" ], "follow_ups": [ "资料去重可先用指纹快速判断相似，再人工检查高价值冲突", "敏感操作则不能只靠概率判断" ], "project_mapping": [ "AI PM 要能把模型误差翻译成业务风险和产品策略。" ], "aipm_transfer": [ "probabilistic-data-structure", "tradeoff", "engineering" ], "tags": [ "math-for-ai", "nlp", "data", "product" ], "priority": "P1" }, { "number": 29, "collection": "吴军作品合集", "book": "数学之美", "chapter": "动态规划", "title": "动态规划对产品路径设计有什么启发？", "interview_question": "动态规划把复杂问题拆成子问题，并利用中间结果避免重复计算。产品策略也可以拆阶段、存状态、逐步优化。", "short_answer": "复杂目标可分解成阶段性最优；状态定义决定问题能否求解；复用中间结果能提升系统效率", "key_points": [ "动态规划如何通俗解释？", "产品路径如何拆阶段？", "Agent 任务规划和动态规划有什么相似？" ], "follow_ups": [ "把求职准备拆成定位、素材、简历、投递、面试、复盘", "每阶段产出可复用资产" ], "project_mapping": [ "AI PM 做 Agent 工作流时，要重视状态、子任务和可复用中间产物。" ], "aipm_transfer": [ "dynamic-programming", "workflow", "planning" ], "tags": [ "math-for-ai", "nlp", "product" ], "priority": "P1" }, { "number": 30, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "信息产业规律", "title": "科技公司兴衰有哪些规律？", "interview_question": "科技公司兴衰常受技术代际、商业模式、生态位置、组织惯性和资本周期共同影响。领先者失败往往不是因为不聪明，而是被既有成功路径束缚。", "short_answer": "技术代际切换会改变竞争规则；商业模式决定价值捕获；组织惯性会让巨头错过新曲线", "key_points": [ "为什么大公司会错过新机会？", "AI 时代哪些公司容易被颠覆？", "如何判断一家公司的护城河？" ], "follow_ups": [ "分析 AI 工具公司时看模型依赖、分发渠道、数据闭环、工作流嵌入和切换成本" ], "project_mapping": [ "AI PM 要能从公司战略和产业规律解释产品机会，而不只看功能。" ], "aipm_transfer": [ "tech-cycle", "company-strategy", "moat" ], "tags": [ "tech-business", "strategy", "business-model" ], "priority": "P0" }, { "number": 31, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "生态链", "title": "计算机工业生态链对 AI 产业有什么启发？", "interview_question": "每一代技术产业都会形成硬件、平台、应用、渠道和资本的生态链。AI 也在形成模型、算力、数据、工具、应用和分发的分层生态。", "short_answer": "价值不平均分布在生态每一层；平台层往往控制标准和分发；应用层需要找到具体场景和数据闭环", "key_points": [ "AI 创业该做模型层还是应用层？", "生态位如何影响利润？", "平台依赖有什么风险？" ], "follow_ups": [ "为 AI PM 岗位分析所在公司处于模型、工具、垂直应用还是平台生态位" ], "project_mapping": [ "AI PM 要理解自己产品在生态链中的位置，才能制定差异化策略。" ], "aipm_transfer": [ "ecosystem", "value-chain", "ai-industry" ], "tags": [ "tech-business", "strategy", "ai-product" ], "priority": "P0" }, { "number": 32, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "商业模式", "title": "什么是科技公司的好商业模式？", "interview_question": "好的商业模式通常具备高毛利、可规模化、强复用、低边际成本和较强用户锁定。技术领先必须转化为可持续价值捕获。", "short_answer": "技术优势不自动等于利润；商业模式决定增长质量；最佳模式常嵌入用户高频工作流或关键基础设施", "key_points": [ "AI 产品按 token 收费是否可持续？", "如何设计 AI SaaS 的价值计费？", "开源会不会破坏商业模式？" ], "follow_ups": [ "评估面试素材库可按个人订阅、机构培训、企业内训和知识库 API 四种模式比较" ], "project_mapping": [ "AI PM 要能把能力包装成可计费、可复用、可扩展的产品单元。" ], "aipm_transfer": [ "business-model", "saas", "monetization" ], "tags": [ "tech-business", "strategy", "business-model" ], "priority": "P0" }, { "number": 33, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "风险投资", "title": "风险投资为什么影响科技产业走向？", "interview_question": "风险投资提供的不只是钱，还包括对高风险高增长模式的筛选、加速和治理。它改变公司扩张速度，也会影响产品优先级。", "short_answer": "VC 偏好可规模化和大市场；资本会加速试错，也可能放大泡沫；融资叙事和真实产品价值需要区分", "key_points": [ "融资多是否代表产品好？", "资本如何影响 PM 工作？", "AI 创业公司如何证明增长？" ], "follow_ups": [ "看 AI 公司时同时分析融资节奏、客户质量、成本结构和留存，而不是只看估值" ], "project_mapping": [ "AI PM 面试可展示你理解资本、增长和产品验证之间的张力。" ], "aipm_transfer": [ "venture-capital", "startup", "growth" ], "tags": [ "tech-business", "strategy" ], "priority": "P1" }, { "number": 34, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "从 IBM 到微软到 Google", "title": "平台迁移如何改变赢家？", "interview_question": "从大型机、PC、互联网到云和 AI，每次平台迁移都会重排价值链。新赢家通常抓住新平台的分发、开发者或数据入口。", "short_answer": "旧平台优势可能变成新平台包袱；迁移期给挑战者窗口；赢家往往定义开发者和用户的新默认入口", "key_points": [ "AI 是否是新平台？", "新平台的入口在哪里？", "传统 SaaS 如何迁移到 AI Native？" ], "follow_ups": [ "把 AI 面试库从静态网页升级为 Agent 入口：用户以目标驱动，而不是按目录浏览" ], "project_mapping": [ "AI PM 要寻找新平台的默认交互入口，例如聊天、Agent、插件、工作流或 IDE。" ], "aipm_transfer": [ "platform-shift", "google", "microsoft" ], "tags": [ "tech-business", "strategy" ], "priority": "P0" }, { "number": 35, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "云计算 / 互联网2.0", "title": "云计算和互联网2.0对 AI 产品有什么前史意义？", "interview_question": "云计算让算力和软件服务化，互联网2.0让用户参与内容和数据生产。AI 产品正建立在这两者之上：服务化能力加用户反馈数据。", "short_answer": "云降低启动成本和规模化门槛；UGC/行为数据形成训练与优化资源；AI 产品需要云端能力和用户反馈共同驱动", "key_points": [ "AI Native 和 SaaS 有什么继承关系？", "用户生成数据如何形成飞轮？", "云成本如何影响 AI 商业模式？" ], "follow_ups": [ "把卡片生成、在线访问、用户反馈和后续个性化推荐做成持续服务，而不是一次性文件" ], "project_mapping": [ "AI PM 要把云成本、延迟、数据回流和用户协作放进产品模型。" ], "aipm_transfer": [ "cloud", "web2", "ai-native" ], "tags": [ "tech-business", "strategy", "product", "ai-product" ], "priority": "P1" }, { "number": 36, "collection": "吴军作品合集", "book": "浪潮之巅", "chapter": "下一个 Google", "title": "怎样判断“下一个巨头”可能出现在哪里？", "interview_question": "巨头通常出现在新技术平台、巨大未满足需求、可规模化商业模式和强网络效应交汇处。判断重点是结构条件，而不是复制上一代公司形态。", "short_answer": "新巨头不一定像旧巨头；入口、数据和生态控制权是关键；早期信号包括高频使用、强留存和开发者/用户自扩散", "key_points": [ "AI 时代下一个入口是什么？", "垂直 Agent 能否成长为平台？", "如何避免套用旧模板？" ], "follow_ups": [ "分析 AI 求职助手是否能从工具变成职业成长平台：资料、练习、反馈、机会匹配、社区" ], "project_mapping": [ "AI PM 要用结构性条件评估机会，而不是追逐“某某的 AI 版”。" ], "aipm_transfer": [ "next-platform", "network-effect", "strategy" ], "tags": [ "tech-business", "strategy" ], "priority": "P0" }, { "number": 37, "collection": "吴军作品合集", "book": "硅谷之谜", "chapter": "硅谷的奇迹", "title": "硅谷为什么能持续产生创新？", "interview_question": "硅谷的持续创新来自开放网络、人才密度、资本、大学、移民文化、失败容忍和工程实践的组合，而不是单一政策或地理优势。", "short_answer": "生态系统比单点资源更重要；失败容忍提高试错频率；人才和资本流动让知识快速重组", "key_points": [ "硅谷模式能复制吗？", "中国 AI 创新生态差异在哪里？", "公司内部能否构建小硅谷？" ], "follow_ups": [ "为团队建立快速实验、公开复盘、跨职能协作和外部学习机制" ], "project_mapping": [ "AI PM 要学习硅谷的系统条件：高质量试错、开放协作和快速反馈。" ], "aipm_transfer": [ "silicon-valley", "innovation", "ecosystem" ], "tags": [ "innovation", "silicon-valley" ], "priority": "P0" }, { "number": 38, "collection": "吴军作品合集", "book": "硅谷之谜", "chapter": "企业文化与情怀", "title": "文化为什么会影响科技公司的产品能力？", "interview_question": "文化决定组织如何看待风险、用户、工程质量、人才和长期目标。技术公司竞争到最后，很多时候是文化和组织能力的竞争。", "short_answer": "工程文化影响产品可靠性；开放文化影响创新速度；长期主义影响战略耐心", "key_points": [ "如何在面试中判断团队文化？", "AI 团队需要什么文化？", "文化和流程哪个更重要？" ], "follow_ups": [ "选择 AI PM 岗位时观察团队是否重视 eval、用户反馈、工程质量和透明复盘" ], "project_mapping": [ "AI PM 要推动真实反馈文化，因为 AI 产品质量高度依赖持续评测和纠错。" ], "aipm_transfer": [ "culture", "organization", "engineering-quality" ], "tags": [ "innovation", "silicon-valley", "business-model", "product" ], "priority": "P0" }, { "number": 39, "collection": "吴军作品合集", "book": "硅谷之谜", "chapter": "信息时代的科学基础", "title": "信息时代的科学基础如何影响产品思维？", "interview_question": "信息论、控制论、计算机科学和网络科学让产品经理能用系统、反馈、复杂性和概率来理解现代产品。", "short_answer": "现代产品是反馈系统；网络结构决定传播与协作；概率思维帮助处理不确定性", "key_points": [ "产品经理需要懂哪些科学基础？", "系统思维如何落地到 PRD？", "AI 产品为什么更需要反馈控制？" ], "follow_ups": [ "把 AI 面试工具建成反馈系统：生成、练习、评分、修正、再练习" ], "project_mapping": [ "AI PM 的底层能力是用系统科学语言描述产品，而不是只写功能点。" ], "aipm_transfer": [ "systems-thinking", "feedback", "information-age" ], "tags": [ "innovation", "silicon-valley", "product" ], "priority": "P0" }, { "number": 40, "collection": "吴军作品合集", "book": "硅谷之谜", "chapter": "拒绝简单归因", "title": "为什么解释成功不能只找一个原因？", "interview_question": "复杂系统的成功通常由多因素共同作用。只找一个原因会导致错误模仿，真正有价值的是识别必要条件、增强条件和偶然因素。", "short_answer": "成功案例容易被事后叙事简化；可复制的是结构，不是表象；案例分析要区分因果、相关和叙事", "key_points": [ "竞品成功如何拆解？", "为什么复制硅谷表象没用？", "面试中如何讲复杂问题？" ], "follow_ups": [ "分析 Claude Code 或某 AI 产品时拆成模型、工具、权限、分发、品牌、用户场景等多个变量" ], "project_mapping": [ "AI PM 要避免单因解释，学会做多变量产品诊断。" ], "aipm_transfer": [ "complex-system", "case-analysis", "causality" ], "tags": [ "innovation", "silicon-valley" ], "priority": "P0" }, { "number": 41, "collection": "吴军作品合集", "book": "吴军作品合集", "chapter": "全书整合", "title": "吴军式技术商业分析框架", "interview_question": "如何用吴军作品形成技术商业分析框架？", "short_answer": "可以用四层框架：技术基础设施变化、产业生态重组、商业模式价值捕获、个人与组织能力迁移。这样能把技术趋势讲成可执行判断。", "key_points": [ "先判断底层技术是否成熟", "再看生态链价值如何分布", "再看商业模式能否持续", "最后看个人和组织如何迁移能力" ], "follow_ups": [ "如何用这个框架分析 AI Agent？", "它和普通竞品分析有什么区别？", "面试时怎样讲得简洁？" ], "project_mapping": [ "用该框架写 AI PM 岗位行业分析：模型基础设施、Agent 生态、SaaS 商业化、PM 能力升级" ], "aipm_transfer": "AI PM 可以把这套框架作为技术趋势题、商业题和职业规划题的共同底座。", "tags": [ "analysis-framework", "tech-business", "ai-pm" ], "priority": "P0" }, { "number": 42, "collection": "吴军作品合集", "book": "吴军作品合集", "chapter": "全书整合", "title": "AI PM 面试素材地图", "interview_question": "这套合集最适合沉淀哪些面试素材？", "short_answer": "最适合沉淀五类素材：技术史趋势、数据智能方法、数学/NLP 基础、科技公司商业规律、个人成长与选择框架。它能补足 AI PM 的宏观判断和底层表达。", "key_points": [ "技术史帮助讲趋势", "智能时代帮助讲数据闭环", "数学之美帮助讲模型直觉", "浪潮之巅帮助讲商业模式", "见识和大学之路帮助讲职业选择" ], "follow_ups": [ "哪些卡最该先背？", "如何把这些内容讲进项目经历？", "和前几本技术书如何组合复习？" ], "project_mapping": [ "面试前按问题类型取卡：AI 趋势题用智能时代，系统设计用数学之美，商业题用浪潮之巅，成长题用见识" ], "aipm_transfer": "AI PM 面试不是只考工具使用，而是考技术、商业、用户和自我成长的综合判断。", "tags": [ "interview-map", "study-plan", "ai-pm" ], "priority": "P0" } ]