[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fTU7YdURNhs819t1RhcQ6CiCZNBNpLunRW0b4dW2wXWs":3},{"code":4,"msg":5,"data":6},200,"操作成功",{"id":7,"title":8,"content":9,"digest":10,"source":10,"coverPath":11,"thumbsCoverPath":12,"isTop":13,"isShow":14,"baseClick":13,"clickCount":15,"createTime":16,"typeId":17,"isNewest":18,"newsInfoTypeRespVo":19,"voiceUrl":22,"voiceSize":23,"taskId":24,"releaseTime":25,"titleEn":26,"contentEn":27,"voiceUrlEn":28,"taskIdEn":29,"voiceSizeEn":30},1265,"Manus 最新对话全文：尝试 Agent 支付，公司 RRR 近 1 亿美元","\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002F65ff808a129249dcbe1e4d1d479f508f\u002FAA1L0fLZ.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">搬到新加坡的 Manus，对通用 AI Agent 的思考没有停止。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">在今天在新加坡举行的 Stripe Tour 上，Manus 联合创始人兼首席科学家季逸超（Peak）与 Stripe 亚太地区及⽇本⾸席营收官 Paul Harapin 进行了对谈。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">席间， Manus AI 披露了近期的经营数据，Manus AI 收入运行率（RRR，Revenue Run Rate）已经达到 9000 万美元，很快即将上亿。\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Manus AI 的肖弘还特意在即刻表示，Revenue Run Rate 指的是当月的 Revenue *12，并不等于 Cash Income。很多 AI 产品都会有年付选项，这部分只能算作预存款，而不能记为 Revenue。「如果我们按照这种【错误方式】披露，可以算出一个一个比 1.2 亿美金更大的数。」肖弘表示。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">除了经营数据之外，季逸超还分享了 Manus 团队是如何思考通用 Agent 的的下一步的，以及 AI Agent 和 AGI 未来到底有什么区别。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">“现在大家几乎把一切都叫做 Agent。比如一个麦克风，有人会叫它是「有环境感知的收音 Agent」。”季逸超开玩笑地讲道。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">他也给出了拓展通用 Agent 能力下一步的两条主线： 其一，用多 Agent 协作把执行规模做宽 （如在大规模调研里衍生出上百个并行的子 Agent）； 其二，给 Agent 打开更大的「工具面」 ，不把能力绑死在少数预置 API，而是像程序员一样调用开源生态、安装库、甚至在可视化后 看图自检并回改 \u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">季逸超还提到，今天的数字世界仍按「给人用」的范式建造——非 API 化网页、CAPTCHA、流程「小游戏化」带来大量摩擦， 瓶颈更像生态与制度约束，而非模型智力 。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这也是 Manus 参与 Stripe 活动的原因之一：双方正推进 在 Agent 内完成支付 ，把「研究—决策—下单\u002F结算」连成闭环，用基础设施协作来消解世界的摩擦。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">以下为对话精华，经极客公园编辑整理：\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问：给观众简单介绍一下你自己吧。你最近关于「上下文工程」的博客非常鼓舞人心，我觉得对于在座任何正在开发 AI Agent 的人来说都是必读的。我每次去和工程师们吃午饭时，他们总是在谈论这个，所以我现在只能坐到别的地方（笑）。但对于现场可能不太熟悉 Manus 的人，你能分享一下你的经历和愿景吗？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 谢谢 Paul。很高兴能来到现场。Manus 在构建一个通用 AI Agent。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">很多研究机构和公司，其实在尝试打造一个大脑——打造一个大语言模型。但我们认为，从消费者的角度来看，这其实并不好。AI 应该能真正采取行动，完成事情，所以我们构建了 Manus。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们的方式，是让 AI 可以使用人类历史上最伟大的发明之一——通用计算机。给了 AI 计算机，它能做人类能做到的所有事。Manus 能真正完成任务。比如它可以帮你做演示文稿、帮你规划一次旅行，甚至能帮你运营社交媒体——虽然我并不推荐你真的这样做。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们的用户真的很喜欢 Manus。 我们在三月份发布了 Manus，现在已经实现了大约 9000 万的收入运行率（RRR，Revenue Run Rate），很快就会突破 1 亿。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我觉得这对我们这样一个小型创业公司来说非常巨大。但更重要的是，这表明 AI Agent 已经不再只是一个研究领域的流行词，而是真正在被应用，落地生根。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我可以和大家分享一个我们构建 Manus 过程中的小故事。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们其实从 Agent coding 的应用中得到了很多灵感。比如像 Cursor 这样的 AI 编程产品，之前已经吸引了很多关注。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">作为工程师，我们自然会使用 Cursor。但让我们惊讶的是，公司里很多非工程师同事也在用 Cursor。当然，他们并不是在写软件，而是用它来做数据可视化，甚至写一些文章。他们会忽略左边的代码部分，只是跟 AI 对话来完成工作。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这让我们意识到：我们应该把这种方式泛化，赋能非程序员。这就是 AI 的一个用例。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问：我们越来越常听到人们谈论 AI Agent 和 AGI。你能不能帮我们更清楚地区分这两个概念？AI Agent 和 AGI 对你和 Manus 来说各自意味着什么？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 我们认为这是一个非常好的问题。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">现在大家几乎把一切都叫做「Agent」。比如一个麦克风，有人会说它是「有环境感知的收音 Agent\u003C\u002Fspan>\u003Cspan style=\"font-size: 18px; color: rgb(255, 255, 255); background-color: rgb(56, 56, 56);\">」\u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">但至少我们主张，Agent 应该是应用型 AI 的一个子集。我们不妨退一步，看看常见的 AI 应用类别。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">大多数人已经熟悉两类：一类是聊天机器人，比如 ChatGPT；另一类是生成式工具，比如 MidJourney 或 Sora。 在这些系统中，通常只有两个角色：用户和模型。你和模型交互，得到输出。而 Agent 的不同在于，它除了用户和模型，还引入了第三个关键元素——环境。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这个「环境」的概念会因智能体类型不同而变化，比如在设计型 Agent 里，环境可能是一个画布或一段代码；而在 Manus 这里，我们的目标是让 Agent 出现在虚拟机甚至整个互联网中。这样 Agent 就能观察环境，决定下一步该做什么，并通过行动来改变环境。这让它非常强大。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">比如在 Manus，你可以表达需求，它会打开浏览器、发布网页、帮你订一张机票。我很喜欢这个例子，因为虽然订机票听起来很简单，但这其实是 AI 在直接改变现实世界——结果不是模型的输出，而是你手里的机票。AI 真正介入了你的世界。这就是我们所说的 Agent。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">简单来说，Agent 就是能代表用户与环境交互的 AI 系统。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">至于 AGI，这个词也经常被提到，很多人把它等同于超级智能。我们认为，AGI 是一种能利用 AI 模型的通用能力，在不经过特别设计的情况下完成许多任务的系统。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们认为「Agent coding」其实是通往 AGI 的一条路径。它不是一个垂直领域的能力，而是如果你把它赋予计算机，它几乎能在计算机上做任何事。所以对我们来说，AGI 的条件就是要构建足够完善的环境，让这种能力得以发挥。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问： AI 今天究竟在哪些场景中真正发挥了作用？未来会在哪些地方发挥作用？什么时候会出现 iPhone 时刻？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 就 Agent 而言，如果单看模型能力的话，现在的旗舰模型已经非常惊人了，几乎是「超人」级别的。它们可以在数学竞赛或逻辑推理上胜过我们大多数人。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">但我认为，模型仍然像是「瓶子里的大脑」，如果想真正发挥力量，就必须让它们与真实世界交互、触达现实。但不幸的是，这正是问题开始的地方。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">比如你让一个 AI 去做一些事务性的任务，它在重复性任务上确实很擅长。比如像 Deep Research 这样的产品，它只是聚合信息然后给出一个结果，它的输出只是简单地出现在那里。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">举个例子，现在几乎所有东西都是为人类设计的，不只是物理世界，甚至数字世界也是这样。比如网页工具，它们就像小游戏一样，没有提供 API 或标准接口。验证码 CAPTCHA 无处不在，处处在拦截 Agent。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">所以我认为 AI 在封闭的自包含任务中表现很好，但一旦涉及真实世界，就会遇到障碍。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">未来什么时候能出现 iPhone 时刻？我觉得这并不是技术问题，而是更像是一种制度性限制。这不是像我们这样的 Agent 创业公司能够单独解决的事情。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我认为这需要一个渐进的转变，要求整个生态系统共同进化。这也需要像 Stripe 这样的公司在基础设施层面发力。比如我们正在集成新的 stripe 的 Agentic 支付 api。大家共同努力。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问：那我们能不能具体谈谈用户在使用 Manus 时的一些典型场景？他们是如何使用的？这其中体现出了怎样的力量？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 是的，我们虽然来自当前这一代 Agent，但已经看到很多很棒的用例。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">比如说，我们刚刚搬到新加坡，需要雇佣房产中介来帮我们找住所。是真人 Agent（笑）。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">而现在这些中介已经在使用 Manus：他们会根据客户的需求，用 Manus 来分析公司所在地、员工想住的区域，并生成相应的推荐。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我觉得这很有意思，因为这属于一种「长尾需求」。一般来说，并没有专门的 AI 产品是为这种具体场景设计的，但由于 Manus 是一个通用型 Agent，它就能满足这些需求。我们认为长尾需求非常值得关注。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">从宏观角度看，它可能是长尾，但对具体用户而言，这正是他们的日常工作。这种场景特别有价值。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">这就像今天的搜索引擎格局一样。如果你只是搜索一些常见的内容，不管用 Google 还是 Bing，结果质量差不多。那为什么人们会选择其中一个？可能是因为某个搜索引擎在特定时刻给了他们更合适的结果。而如果你搜索的是非常个性化或专业化的内容，就更能体现差异。所以我们认为通用型 Agent 的优势就在这里。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">那如何让它更好呢？我们思考了很久，因为我们认为一切都绕不开编程。如果你把计算机交给 AI，那么它与环境交互的方式其实就是通过编程。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们认为可以从两方面改进。第一是规模化。但如果你能把 Agent 的能力放大一百倍会怎样呢？\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">最近 Manus 刚刚发布了一个新功能，叫做 Wide Research。它的基本思路是允许一个 Agent 再衍生出上百个 Agent 一起去完成任务。你知道的， 如果只是让 AI 帮你做一些小事，很多时候你自己也能完成。但如果任务非常庞大，你一个人根本不可能完成，比如需要做大规模的研究，这时候让上百个 Agent 并行去做，就会变得非常强大。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">其次，我们还需要让 Agent 更灵活地使用计算机使。比如，如果你只给一个 AI Agent 设置了预设工具，那么它的行动空间就被限制在这些工具里。但想象一下，如果你是一个程序员，你有整个开源社区的资源可以调用。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">比如你在 3D 打印的时候，直接修改模型的参数很困难，但是你如果能找到 GitHub 上的合适的库，直接安装就能解决你的问题了。 在 Manus，我们在优化通用性，并且提出了一个概念，叫做「工具的网络效应」。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">有一个很有意思的例子：很多用户在用 Manus 做数据可视化。你们知道，在亚洲有时候会遇到问题，比如在图表中显示中文时可能会出现字体错误。或许有些专业用户会写一些硬编码规则，比如在输出韩文时应该用哪种字体。但这种方式会让系统越来越僵化。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">我们采取的办法是给系统增加了一个很简单的能力：查看图像。结果很惊喜——因为今天的模型已经很聪明了，它们会在生成可视化图像后自己检查，并意识到错误，然后再自动修正。我们发现， 增加工具的灵活性比硬编码规则能解决更多的问题。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问：这是一个令人兴奋的时代。我真的很激动，只希望自己能再年轻到三十岁（笑）。提到医学研究的，我知道 Manus 在这方面也很强。你们观察到有些用户在使用 Manus 研究医疗吗？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 很多人已经在用 Manus 做研究，不仅限于医学研究。我们觉得这很有意思，因为现在确实有很多所谓的「深度研究」产品，它们会帮你收集大量信息并做一些分析，但最后只给你一个 markdown 文件或文档。这远远不够。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">很多时候，研究人员真正需要的，是能够直接交付给老板或团队的成果。所以我们在 Manus 上加强了研究结果的输出。例如在医学研究中，很多时候需要生成正式的报告，比如幻灯片报告之类的东西。因此我们必须优化 AI 的输出能力，以满足研究人员的需求。这是一种「工具化」的体验。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">比如现在很多用户会先用 Manus 做研究，然后直接生成一个网站。你会觉得这和传统的网站搭建方式完全不同。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">要知道，搭建一个网站本身其实不难，难的是如何确保数据的可靠性和准确性。所以我们认为，最好能在一次会话、一个共享的上下文中完成整个流程。这样，你的研究、你的见解就能无缝转化为最终成果。 这就是我们在 Manus 里所做的事情。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">问：很多国家都在讨论一个话题：在 AI 时代，人类的未来和经济影响。你怎么看待就业被取代？又会出现哪些新的工作机会？\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">答： 我们的朋友和投资人也经常问我们这个问题。当我们推出 Manus 时，最初认为如果能构建这样一个 Agent，它就能帮人们节省很多时间，让大家轻松赚钱。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">但实际上，我们发现这个愿景并没有完全实现。通过大量用户调研，我们发现用户在使用之后，他们反而工作得更多了。因为他们变得更高效了，他们实际上能做更多他们本来就很擅长的事情。这是第一点。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">其次，我们认为 Manus 还打开了一个全新的空间。我们一直在讨论虚拟机和云计算。我们觉得 Manus 正在扮演一种「个人云计算平台」的角色。比如云计算已经存在几十年了，但它更多是工程师的特权，只有我们能通过编程来调用云的力量。普通知识工作者无法使用。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">但现在有了像 Manus 这样的 AI Agent，人们可以用自然语言下达指令，让 AI 去执行。这等于解锁了一种全新的生产力。这就是我们所带来的。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">而最后，关于「替代」，我认为其实很难。比如房产中介，他们每天都在用 Manus 完成日常工作。但你知道，AI 永远无法替代中介面对客户时的那种沟通方式。我们是一家 AI 公司，甚至 Manus 的发布视频都是由 Manus 写的脚本，但视频里出现的还是我，因为这是关于信任的事情。而信任，是不能完全交给 AI 的。\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(187, 187, 187);\">【新闻来源】 极客公园 \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Fmanus-%E6%9C%80%E6%96%B0%E5%AF%B9%E8%AF%9D%E5%85%A8%E6%96%87-%E5%B0%9D%E8%AF%95-agent-%E6%94%AF%E4%BB%98-%E5%85%AC%E5%8F%B8-rrr-%E8%BF%91-1-%E4%BA%BF%E7%BE%8E%E5%85%83\u002Far-AA1L0fM5?ocid=msedgntphdr&cvid=0ea18e61d8c242ccabdb61117c000caa&ei=35\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\">http:\u002F\u002Fu5a.cn\u002FwLFQS\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">（本网转发此文章，旨在为读者提供更多的信息资讯，所涉内容不构成投资、消费建议。文章事实如有疑问，请与有关方核实，文章观点非本网观点，仅供读者参考。）\u003C\u002Fspan>\u003C\u002Fp>","","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002F0d389600b877404393847b6c9dcb4c06\u002FAI领域.jpg","https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002Fthumbs\u002F0d389600b877404393847b6c9dcb4c06\u002FAI领域.jpg",0,1,210,"2025-08-22 17:24",2,false,{"id":17,"name":20,"enName":21},"芯位视野","Xinwei Vision","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3Aa765d37b-fefb-4617-9346-702b35a79a7e%3A0.wav?Expires=1755861964&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=olMKj6EIPRMdr1wHum%2Fuv0sEFrU%3D",25043146,"a765d37b-fefb-4617-9346-702b35a79a7e","2025-08-22 17:12","Manus latest full dialogue: Trying Agent payment, company RRR near 100 million USD","\u003Cimg alt=\"undefined\" src=\"https:\u002F\u002Fimage.51xinwei.com\u002F2025\u002F08\u002F65ff808a129249dcbe1e4d1d479f508f\u002FAA1L0fLZ.png\" width=\"undefined\" height=\"undefined\" style=\"display: block; margin: auto;\">\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"color: rgb(255, 153, 0); font-size: 18px;\">Manus moved to Singapore, but their thinking about general AI Agent has not stopped.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">At today's Stripe Tour in Singapore, Manus co-founder and chief scientist Jie Yichao (Peak) had a conversation with Paul Harapin, Stripe's Asia-Pacific and Japan Chief Revenue Officer.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">During the conversation, Manus AI revealed its recent operating data, and Manus AI's revenue run rate (RRR) has already reached 90 million USD and will soon exceed 100 million.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Xiao Hong of Manus AI also specifically mentioned that the revenue run rate refers to the current month's revenue multiplied by 12, which is not equal to cash income. Many AI products have annual payment options, and this part can only be considered as prepayment, not revenue. \"If we disclose it in this incorrect way, we can calculate a number larger than 120 million USD.\" Xiao Hong said.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In addition to the operating data, Jie Yichao also shared how the Manus team is thinking about the next steps for general Agents, and what the difference between AI Agents and AGI will be in the future.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">\"Now almost everything is called an Agent. For example, a microphone, someone might call it a 'microphone agent with environmental perception'.\" Jie Yichao joked.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">He also gave two main directions for expanding the capabilities of general Agents: first, using multi-Agent collaboration to broaden the execution scale (such as deriving hundreds of parallel sub-Agents in large-scale research); second, opening up a larger 'tool face' for the Agent, not binding the capabilities to a few preset APIs, but instead calling the open-source ecosystem like a programmer, installing libraries, and even visually checking and correcting after seeing the image.\u003C\u002Fstrong>\u003Cspan style=\"font-size: 18px;\">\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Jie Yichao also mentioned that the current digital world is still built according to the 'for human use' paradigm—non-API websites, CAPTCHA, and gamified processes cause a lot of friction, and the bottleneck is more about the ecosystem and institutional constraints rather than model intelligence.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This is also one of the reasons why Manus participated in the Stripe event: both parties are pushing to complete payments within the Agent, connecting the 'research-decision-purchase\u002Fsettlement' into a closed loop, and eliminating the friction in the world through infrastructure collaboration.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The following is the essence of the dialogue, edited and organized by Geke Park:\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: Introduce yourself briefly to the audience. Your recent blog on \"context engineering\" was very inspiring, I think it's essential reading for anyone developing AI Agents. Every time I have lunch with engineers, they always talk about it, so I now just sit elsewhere (laugh). But for those who may not be familiar with Manus here, can you share your experience and vision?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: Thank you, Paul. It's great to be here. Manus is building a general AI Agent.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Many research institutions and companies are actually trying to build a brain—building a large language model. But we believe, from the consumer perspective, this is actually not good. AI should be able to take real actions and accomplish things, so we built Manus.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Our approach is to let AI use one of the greatest inventions in human history—general-purpose computers. Giving AI a computer, it can do all the things humans can do. Manus can really accomplish tasks. For example, it can help you create presentations, plan a trip, or even manage your social media—although I don't recommend you actually do that.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Our users really love Manus. We launched Manus in March, and we have already achieved a revenue run rate (RRR) of about 90 million USD, and it will soon break through 100 million.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">I think this is huge for a small startup company. But more importantly, it shows that AI Agent is no longer just a buzzword in the research field, but is truly being applied and taking root.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">I can share a small story about how we built Manus.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">We actually got a lot of inspiration from the application of Agent coding. For example, AI programming products like Cursor have already attracted a lot of attention.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">As engineers, we naturally use Cursor. But what surprised us was that many non-engineer colleagues at the company also use Cursor. Of course, they are not writing software, but using it for data visualization or even writing some articles. They ignore the code part on the left and just interact with the AI to complete their work.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">This made us realize: we should generalize this approach and empower non-programmers. This is one of the use cases for AI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: We are hearing more and more about AI Agents and AGI. Can you help us clearly distinguish these two concepts? What do AI Agents and AGI mean to you and Manus?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: We think this is a great question.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Now almost everything is called \"Agent\". For example, a microphone, someone might say it is a \"microphone agent with environmental perception\"\u003C\u002Fspan>\u003Cspan style=\"font-size: 18px; color: rgb(255, 255, 255); background-color: rgb(56, 56, 56);\">.\"\u003C\u002Fspan>\u003Cspan style=\"font-size: 18px;\">.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">But at least we argue that an Agent should be a subset of application-based AI. Let's step back and look at common AI application categories.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Most people are already familiar with two types: one is chatbots, such as ChatGPT; the other is generative tools, such as MidJourney or Sora. In these systems, there are usually only two roles: user and model. You interact with the model and get output. The difference with an Agent is that, in addition to the user and the model, it introduces a third key element—the environment.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">The concept of \"environment\" varies depending on the type of agent. For example, in a design-type agent, the environment may be a canvas or a piece of code; in Manus, our goal is to have the Agent appear in a virtual machine or even the entire internet. This way, the Agent can observe the environment, decide what to do next, and change the environment through actions. This makes it very powerful.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">For example, with Manus, you can express your needs, and it will open a browser, publish a website, and help you book a flight. I like this example because although booking a flight sounds simple, it's actually AI directly changing the real world—the result is not the model's output, but the ticket in your hand. AI is truly intervening in your world. This is what we call an Agent.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">In short, an Agent is an AI system that can interact with the environment on behalf of the user.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">As for AGI, this term is also frequently mentioned, and many people equate it with super intelligence. We believe AGI is a system that can utilize the general capabilities of AI models to perform many tasks without special design.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">We believe that \"Agent coding\" is actually a path to AGI. It is not a vertical domain capability, but if you give it to a computer, it can do almost anything on the computer. So for us, the conditions for AGI require building a sufficiently complete environment to allow this capability to be realized.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: In what scenarios is AI truly making a difference today? Where will it make a difference in the future? When will the iPhone moment come?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: Regarding Agents, if we look at model capabilities alone, the current top models are already amazing, almost at a \"superhuman\" level. They can outperform most of us in math competitions or logical reasoning.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">But I think the model is still like a \"brain in a bottle,\" and to truly unleash its power, it must interact with the real world and reach reality. But unfortunately, this is where the problem begins.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">For example, if you ask an AI to do some transactional tasks, it is indeed very good at repetitive tasks. For example, products like Deep Research simply aggregate information and provide results, and the output simply appears there.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Take an example: almost everything today is designed for humans, not only in the physical world, but even in the digital world. For example, web tools are like games, without providing APIs or standard interfaces. CAPTCHAs are everywhere, intercepting Agents.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">So I think AI performs well in closed, self-contained tasks, but encounters obstacles once it involves the real world.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">When will the iPhone moment come? I think this is not a technical issue, but more like a systemic constraint. It's not something that Agent startups like us can solve alone.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">I think this requires a gradual shift, requiring the entire ecosystem to evolve together. It also requires companies like Stripe to invest in infrastructure. For example, we are integrating new Stripe Agentic payment APIs. Everyone works together.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: Can we talk specifically about some typical scenarios when users use Manus? How do they use it? What kind of power does this demonstrate?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: Yes, although we come from the current generation of Agents, we have already seen many great use cases.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">For example, we just moved to Singapore, and needed to hire a real estate agent to help us find a place to live. It was a real person Agent (laugh).\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">But now these agents are already using Manus: they use Manus to analyze the location of the company and the areas employees want to live in, and generate corresponding recommendations based on customer needs.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">I think this is very interesting, because this belongs to a \"long-tail demand.\" Generally, there is no specific AI product designed for such specific scenarios, but since Manus is a general-purpose Agent, it can meet these needs. We think long-tail demands are very worth paying attention to.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">From a macro perspective, it may be a long tail, but for specific users, it's exactly their daily work. This scenario is particularly valuable.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">It's like the current search engine landscape. If you're just searching for common content, whether you use Google or Bing, the quality of results is similar. Why do people choose one over the other? It may be because a particular search engine provided them with more suitable results at a certain moment. And if you're searching for very personalized or specialized content, the difference becomes more apparent. So we believe the advantage of a general-purpose Agent lies here.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">How can we make it better? We've thought about it a lot, because we believe everything revolves around programming. If you give AI a computer, the way it interacts with the environment is essentially through programming.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">We think there are two ways to improve. First, scaling. What if we could scale the Agent's capabilities a hundred times?\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Recently, Manus just released a new feature called Wide Research. The basic idea is to allow an Agent to spawn hundreds of Agents to complete tasks together. You know, if you just let AI help you with small tasks, often you can do it yourself. But if the task is very large, you can't do it alone, for example, if you need to do large-scale research, then having hundreds of Agents working in parallel becomes very powerful.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Secondly, we also need to make the Agent more flexible in using the computer. For example, if you only set predefined tools for an AI Agent, its action space is limited to these tools. But imagine if you were a programmer, you could access the entire open-source community's resources.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">For example, when you are 3D printing, it's difficult to directly modify the model's parameters, but if you can find the appropriate library on GitHub and install it, you can solve your problem. In Manus, we are optimizing for generality and have proposed a concept called \"network effect of tools.\"\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">There's an interesting example: many users use Manus for data visualization. You know, sometimes in Asia, there are problems, such as font errors when displaying Chinese characters in charts. Perhaps some professional users write some hard-coded rules, like which font to use when outputting Korean. But this approach makes the system increasingly rigid.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Our approach is to add a very simple ability to the system: viewing images. The result was surprising—because today's models are already very smart, they check the generated visualized image themselves, realize the error, and then automatically correct it. We found that increasing the flexibility of tools can solve more problems than hard-coding rules.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: This is an exciting era. I'm really excited, I wish I could be young again (laugh). Speaking of medical research, I know Manus is also strong in this area. Have you observed some users using Manus for medical research?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: Many people are already using Manus for research, not limited to medical research. We find this very interesting because there are indeed many so-called \"deep research\" products that help collect a lot of information and do some analysis, but finally they only give you a markdown file or document. That's far from enough.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Researchers often need to deliver results directly to their bosses or teams. So we have enhanced the output of research results in Manus. For example, in medical research, generating formal reports, such as slide presentations, is often needed. Therefore, we must optimize the AI's output capabilities to meet the needs of researchers. This is a \"toolized\" experience.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">For example, many users now first use Manus for research and then directly generate a website. You would find this completely different from traditional website building methods.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">You should know that building a website itself is not difficult, but the challenge is ensuring the reliability and accuracy of the data. So we think it's best to complete the entire process in one session, in a shared context. This way, your research and insights can seamlessly transform into the final outcome. This is what we are doing in Manus.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cstrong style=\"font-size: 18px;\">Q: Many countries are discussing a topic: the future of humanity and economic impact in the AI era. What do you think about job replacement? What new job opportunities will emerge?\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">A: Our friends and investors often ask us this question. When we launched Manus, we initially thought that if we could build such an Agent, it would save people a lot of time and make it easier to earn money.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">But in fact, we found that this vision wasn't fully realized. Through extensive user research, we found that users actually work more after using it. Because they become more efficient, they can actually do more of the things they are already good at. This is the first point.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Secondly, we believe Manus has opened up a completely new space. We have been discussing virtual machines and cloud computing. We think Manus is playing the role of a \"personal cloud computing platform.\" For example, cloud computing has existed for decades, but it has mainly been a privilege of engineers, only we can use programming to call the power of the cloud. Ordinary knowledge workers cannot use it.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">But now with an AI Agent like Manus, people can give instructions in natural language, and the AI will execute them. This is equivalent to unlocking a new productivity. This is what we bring.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"font-size: 18px;\">Finally, regarding \"replacement,\" I think it's actually very difficult. For example, real estate agents use Manus every day to complete their work. But you know, AI can never replace the way a real estate agent communicates with customers. We are an AI company, and even the script for the Manus launch video was written by Manus, but the person appearing in the video is still me, because it's about trust. And trust cannot be entirely handed over to AI.\u003C\u002Fspan>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cbr>\u003C\u002Fp>\u003Cp>\u003Cspan style=\"color: rgb(187, 187, 187);\">[News Source] Geke Park \u003C\u002Fspan>\u003Ca href=\"https:\u002F\u002Fwww.msn.cn\u002Fzh-cn\u002Fnews\u002Fother\u002Fmanus-%E6%9C%80%E6%96%B0%E5%AF%B9%E8%AF%9D%E5%85%A8%E6%96%87-%E5%B0%9D%E8%AF%95-agent-%E6%94%AF%E4%BB%98-%E5%85%AC%E5%8F%B8-rrr-%E8%BF%91-1-%E4%BA%BF%E7%BE%8E%E5%85%83\u002Far-AA1L0fM5?ocid=msedgntphdr&cvid=0ea18e61d8c242ccabdb61117c000caa&ei=35\" rel=\"noopener noreferrer\" target=\"_blank\" style=\"color: rgb(187, 187, 187);\">http:\u002F\u002Fu5a.cn\u002FwLFQS\u003C\u002Fa>\u003C\u002Fp>\u003Cp class=\"ql-align-justify\">\u003Cspan style=\"color: rgb(187, 187, 187);\">（This article is reposted by this site to provide readers with more information and news. The content does not constitute investment or consumption advice. If there are any questions about the facts of the article, please verify with the relevant party. The views of the article are not the views of this site, and are for reference only.）\u003C\u002Fspan>\u003C\u002Fp>","https:\u002F\u002Fxinwei-dev-test.oss-cn-shenzhen.aliyuncs.com\u002Fintelligent\u002Faudio%3A549e511b-bc06-4516-8f77-9027730e8e2a%3A0.wav?Expires=1774838488&OSSAccessKeyId=LTAI5tNvY2RkKjZw4LLWsrPK&Signature=PkPNDgVGFO6O2chE61xBVQ4mPgo%3D","549e511b-bc06-4516-8f77-9027730e8e2a",17596370]