[{"data":1,"prerenderedAt":109},["ShallowReactive",2],{"nav":3,"footer-links":64,"site-settings-header":91,"footer-social":98,"site-settings-footer":100,"news-6":102},{"data":4},[5,12,18,39,43,59],{"id":6,"label":7,"url":8,"parentId":9,"sortOrder":6,"isExternal":10,"children":11},1,"Home","\u002F",null,false,[],{"id":13,"label":14,"url":15,"parentId":9,"sortOrder":16,"isExternal":10,"children":17},14,"AI Products","\u002Fproducts",2,[],{"id":16,"label":19,"url":20,"parentId":9,"sortOrder":21,"isExternal":10,"children":22},"About Us","\u002Fwhat-we-do",3,[23,26,30,34],{"id":24,"label":25,"url":20,"parentId":16,"sortOrder":6,"isExternal":10},7,"What We Do",{"id":27,"label":28,"url":29,"parentId":16,"sortOrder":16,"isExternal":10},8,"Our Values","\u002Four-values",{"id":31,"label":32,"url":33,"parentId":16,"sortOrder":21,"isExternal":10},9,"Leadership","\u002Fleadership",{"id":35,"label":36,"url":37,"parentId":16,"sortOrder":38,"isExternal":10},10,"Locations","\u002Flocations",4,{"id":21,"label":40,"url":41,"parentId":9,"sortOrder":38,"isExternal":10,"children":42},"Perspectives","\u002Fperspectives",[],{"id":38,"label":44,"url":45,"parentId":9,"sortOrder":46,"isExternal":10,"children":47},"Careers","\u002Fworking-here",5,[48,51,55],{"id":49,"label":50,"url":45,"parentId":38,"sortOrder":6,"isExternal":10},11,"Working Here",{"id":52,"label":53,"url":54,"parentId":38,"sortOrder":16,"isExternal":10},12,"Students & Early-Career","\u002Fstudents-early-career",{"id":56,"label":57,"url":58,"parentId":38,"sortOrder":21,"isExternal":10},13,"Career Paths","\u002Fcareer-paths",{"id":60,"label":61,"url":62,"parentId":9,"sortOrder":60,"isExternal":10,"children":63},6,"Browse Open Roles","\u002Fcareers",[],{"data":65},[66,69,72,75,77,79,82,84,86,89],{"id":16,"section":67,"label":68,"url":20,"sortOrder":6},"about","我们的业务",{"id":60,"section":70,"label":71,"url":45,"sortOrder":6},"careers","在这里工作",{"id":31,"section":73,"label":74,"url":41,"sortOrder":6},"resources","新闻与观点",{"id":21,"section":67,"label":76,"url":29,"sortOrder":16},"我们的价值观",{"id":24,"section":70,"label":78,"url":54,"sortOrder":16},"学生与早期职业",{"id":35,"section":73,"label":80,"url":81,"sortOrder":16},"常见问题","\u002Ffaq",{"id":38,"section":67,"label":83,"url":33,"sortOrder":21},"领导团队",{"id":27,"section":70,"label":85,"url":58,"sortOrder":21},"职业路径",{"id":49,"section":73,"label":87,"url":88,"sortOrder":21},"一般咨询","mailto:dlxxlc@163.com",{"id":46,"section":67,"label":90,"url":37,"sortOrder":38},"全球办公室",{"data":92},{"disclaimer_1":93,"disclaimer_2":93,"stats_heading":94,"stats_description":95,"stats_footnote":96,"logo_icon":97},"","MAD ｜ 迈埃迪","迈埃迪人工智能 ","All statistics as of April 1, 2026","\u002Fuploads\u002Fgeneral\u002F922e3bbe-b397-4472-9663-53698a1ff11d.webp",{"data":99},[],{"data":101},{"disclaimer_1":93,"disclaimer_2":93,"stats_heading":94,"stats_description":95,"stats_footnote":96,"logo_icon":97},{"data":103},{"id":60,"title":104,"eyebrow":93,"excerpt":93,"content":105,"imageUrl":106,"publishedAt":107,"sortOrder":108},"MAD：AI 应用公司的生存框架","\u003Csection>红杉资本，也就是 \u003Cspan class=\"\">Sequoia Capital\u003C\u002Fspan>，是硅谷最有代表性的风险投资机构之一，投过 Apple、Google、Airbnb、Stripe、WhatsApp 等一批公司。\u003C\u002Fsection>\n\u003Csection>这次内容来自红杉 2026 年 AI Ascent 大会的一场主题演讲。它讨论的不是某一个模型更新，而是 AI 应用公司接下来到底怎么活。\u003C\u002Fsection>\n\u003Csection>\n\u003Csection class=\"channels_iframe_wrp custom_select_card_wrp\">\u003C\u002Fsection>\n\u003C\u002Fsection>\n\u003Csection>演讲里有一个判断很重要：AI 不是又一轮“传播革命”，而是一场“计算革命”。\u003C\u002Fsection>\n\u003Csection>互联网、云、移动，主要改变的是信息如何被分发；AI 改变的是信息如何被处理。更进一步，Agent 开始从“回答问题”变成“接任务、做事情、持续推进直到完成”。\u003C\u002Fsection>\n\u003Csection>这也是为什么演讲者说，从商业和功能角度看，AGI 可能已经来了。\u003C\u002Fsection>\n\u003Csection>不一定是科幻电影里的 AGI，而是更现实的版本：你可以把一项工作交给 Agent，它能理解目标、调用工具、处理中途失败，并最终交付结果。\u003C\u002Fsection>\n\u003Csection>这会改变软件，也会改变服务。\u003C\u002Fsection>\n\u003Csection>过去软件主要吃的是软件预算。现在 AI 开始进入法律、咨询、客服、销售、运营、财务这些服务市场。真正大的机会，不只是做一个更聪明的软件，而是把原来由人完成的服务，重新做成 AI 原生流程。\u003C\u002Fsection>\n\u003Csection>但问题也来了：\u003C\u002Fsection>\n\u003Csection>如果底层模型每天都在变，应用层公司怎么建立自己的位置？\u003C\u002Fsection>\n\u003Csection>这就引出了演讲里提到的 MAD 框架。\u003C\u002Fsection>\n\u003Csection>老外确实很喜欢凑缩写。这个 MAD 也有一点这个味道。但公平地说，它还算合理，也好记。\u003C\u002Fsection>\n\u003Csection>M 是 Moats，护城河。\u003C\u002Fsection>\n\u003Csection>A 是 \u003Cspan class=\"\">Affordance\u003C\u002Fspan>，可用性和应用性。\u003C\u002Fsection>\n\u003Csection>D 是 Diffusion，扩散。\u003C\u002Fsection>\n\u003Csection>这三个词，基本说清楚了 AI 应用公司接下来该怎么活。\u003C\u002Fsection>\n\u003Csection>\u003Cstrong>M：护城河来自客户，不只来自技术\u003C\u002Fstrong>\u003C\u002Fsection>\n\u003Csection>这点对纯技术型公司、纯技术型团队尤其重要。\u003C\u002Fsection>\n\u003Csection>很多团队容易执着于技术本身：模型更强、推理更快、成本更低。但在 AI 时代，只盯技术，很容易被下一代模型覆盖。\u003C\u002Fsection>\n\u003Csection>真正更重要的是客户、行业和业务。\u003C\u002Fsection>\n\u003Csection>客户流程、行业惯例、预算审批、合规要求、系统集成、员工培训，都不会因为模型升级就自动改变。\u003C\u002Fsection>\n\u003Csection>所以应用层公司的护城河，不是“我用了最先进的模型”，而是你是否真正理解客户，是否嵌入客户流程，是否和客户的业务融为一体。\u003C\u002Fsection>\n\u003Csection>越是技术变化快，越要回到客户。\u003C\u002Fsection>\n\u003Csection>\u003Cstrong>A：好产品要符合直觉，不只展示能力\u003C\u002Fstrong>\u003C\u002Fsection>\n\u003Csection>Affordance 这个概念，正好和《设计心理学》里讲的是同一个东西。我的理解是：可用性和应用性。\u003C\u002Fsection>\n\u003Csection>一个好产品，不应该让用户想半天才知道怎么用。它应该符合直觉，符合行业惯例，让行业用户一上手就知道下一步该做什么。\u003C\u002Fsection>\n\u003Csection>很多 AI 产品的问题，不是能力不强，而是太像技术演示。\u003C\u002Fsection>\n\u003Csection>未来技术公司要活下来，必须有产品思维：不是把模型能力裸露出来，而是把能力包装成符合行业场景、用户直觉和业务流程的产品。\u003C\u002Fsection>\n\u003Csection>用户不应该思考模型、提示词、工具调用和系统架构。用户只应该知道：我要完成什么，点哪里，结果怎么验收。\u003C\u002Fsection>\n\u003Csection>\u003Cstrong>D：模型扩散到业务，中间永远需要落地者\u003C\u002Fstrong>\u003C\u002Fsection>\n\u003Csection>这一点也接上我上一篇文章里的观点：这个世界会一直有两侧。\u003C\u002Fsection>\n\u003Csection>一侧是 AI 科技端。大模型会快速发布，能力会快速变化，工具会快速迭代。\u003C\u002Fsection>\n\u003Csection>另一侧是业务端。企业还需要部署、落地、实施、培训、流程改造、系统集成和效果评估。\u003C\u002Fsection>\n\u003Csection>模型能力扩散到真实业务，中间永远有距离。这个距离，就是应用层公司的空间。\u003C\u002Fsection>\n\u003Csection>所以，AI 应用公司的机会不是简单套壳。\u003C\u002Fsection>\n\u003Csection>真正的机会，是理解行业，理解客户，理解业务流程，然后把 AI 能力做成客户不用过度思考、上手就能用、并且能真正产生结果的产品。\u003C\u002Fsection>\n\u003Csection>MAD 这个框架说到底，就是三句话：\u003C\u002Fsection>\n\u003Csection>护城河来自客户，不只来自技术。\u003C\u002Fsection>\n\u003Csection>好产品要符合直觉，不只展示能力。\u003C\u002Fsection>\n\u003Csection>模型扩散到业务，中间永远需要落地者。\u003C\u002Fsection>\n\u003Csection>AI 不是更聪明的聊天框。\u003C\u002Fsection>\n\u003Csection>它正在变成新的工作方式。\u003C\u002Fsection>\n\u003Csection>而应用层公司的价值，就是把这种新的工作方式，真正带进客户的业务里。\u003C\u002Fsection>","\u002Fuploads\u002Fnews\u002F511d4051-44b3-42dc-bd4f-ba6ad7d7fb0b.webp","July 9, 2026",0,1790131729705]