Core Insights - The rapid development of AI Agents has ignited a trend of "everything can be an Agent," particularly evident in the competitive landscape of model development and application [1][2][10] - Major companies like OpenAI, Google, and Alibaba are heavily investing in the Agent space, with new products emerging that enhance user interaction and decision-making capabilities [2][7][8] - The evolution of AI applications is categorized into three phases: prompt-based interactions, workflow-based systems, and the current phase of AI Agents, which emphasize autonomous decision-making and tool usage [17][19] Group 1: Model Development - The AI sector has entered a "arms race" for model development, with significant advancements marked by the release of models like DeepSeek, o3 Pro, and Gemini 2.5 Pro [5][6][14] - The introduction of DeepSeek has demonstrated that there is no significant gap between domestic and international model technologies, prompting major players to accelerate their model strategies [6][10] - The focus has shifted from "pre-training" to "post-training" methods, utilizing reinforcement learning to enhance model performance even with limited labeled data [11][13] Group 2: Application Development - The launch of OpenAI's Operator and Deep Research has marked 2025 as the "Year of AI Agents," with a surge in applications that leverage these capabilities [7][8] - Companies are exploring various applications of AI Agents, with notable examples including Cursor and Windsurf, which have validated product-market fit in the programming domain [9][21] - The ability of Agents to use tools effectively has been a significant breakthrough, allowing for enhanced information retrieval and interaction with external systems [20][21] Group 3: Challenges and Opportunities - Despite advancements, AI Agents face challenges such as context management, memory mechanisms, and interaction with complex software systems [39][40] - The future of Agent applications may involve evolving business models, potentially shifting from subscription-based to usage-based or outcome-based payment structures [40][41] - The industry is witnessing a competitive landscape where vertical-specific Agents may offer more value due to their specialized knowledge and closer user relationships [42][46]

几个月前掀起国内自媒体沸腾高潮的 AI 初创公司 Manus 传出裁员消息。消息指 Manus 裁撤中国区业务，原本约 120 人的国内团队，除 40 余名核心技术 人员保留并迁往新加坡外，其余员工均遭裁员。 Manus 在对外回应中未否认裁员，仅表示"基于公司自身经营效率考量，我们决定对部分业务团队进行调整。公司将继续专注核心业务发展，提升整体运 营效率。" 年初 DeepSeek 掀起国产 AI 浪潮之后，Manus 一度被视作另一个国人骄傲。但这家公司可能一开始就没打算面向国内市场。虽然它最早的研发团队位于 北京和武汉两地，但除了早期邀请部分国内极客用户体验外，它几乎没有在国内平台进行什么宣发。 它的第一支宣传片首发于 Youtube、X 等海外社媒平台，创始人用着流利的英文介绍产品。那条在视频号上引发大面积转发的 Manus 宣传片也并非由官方 账号发布，而只是在团队联合创始人张涛的个人视频号"潜云思绪"上发布。它的网站最早也只有英文版本，没有中文。 Manus 虽然也在国内社交媒体如抖音、小红书、视频号、微博等开通了官方账号，但从未发布任何动态。只有视频号发布了一条与阿里通义千问达成战略 合作的消息 ...

1、行业：模型竞赛冷却，AI进入产品化与生态整合。各大平台不再仅仅卷参数和论文，而是争夺用户入口、智能体标准和终 端能力。 2、模型：大模型不再是简单被动的"回答器"，在生成回答之前，它会进行复杂的"自我对话（Self-Reflection）"和多步推理。变 得越来越像人一样：一边思考，一边行动。 3、智能体：正在根本性地改变人与软件交互和工作的方式。从"知识助理"或"内容生成器"，进化成能独立完成任务的"数字员 工"。人机界面从传统的"工具栏+菜单"，变成更自然、更直观的"自然语言+智能执行"。 4、AI编程：AI编程从辅助走向全面委派。编程不再是一种纯粹的"手艺活"或"体力活"，价值重心从具体的编码技能，转向了更 具战略性的产品设计能力、系统架构能力和决策能力。 5、商业模式：2025年，行业共识从MaaS（Model as a Service，模型即服务），转变为RaaS（Results as a Service ，结果即服 务）。AI从成本中心直接转化为利润引擎。 6、职业冲击：品味、判断、方向感，才是人无法被替代的核心能力。别怕AI会抢走你的工作。怕的是，你没有准备好，成为 一个能够和AI协作的人。 ...

这只是豆包产品迭代的一个细节。据华尔街见闻观察，3月份以来，豆包明显加快了产品迭代速度，先 是升级了1月上线的"AI编程"功能，接着豆包新版深度思考功能开启测试，此后，豆包在图片和视频功 能上迎来多次更新。 作者 | 刘宝丹 编辑 | 黄昱 AI应用追击战还在持续。自从年初被DeepSeek 超过，豆包就加快了产品迭代步伐。 6月17日，华尔街见闻获悉，豆包电脑版已全量上线AI播客功能。用户点击"AI播客"，上传PDF或网页 链接后，可一键生成双人对话的播客节目。目前，豆包App也已开启小流量测试，将于近期全量上线。 不同于传统 AI 播客通常不够口语化、听感机械且缺乏互动，该能力基于豆包大模型团队推出的语音播 客模型，让文本创作到双人对话式播客仅需秒级转化，实现"低成本、高时效、强互动"的创作体验。 一位主做C端应用的AI独角兽公司内部人士对华尔街见闻表示，公司目前的重点是研究下一代模型，至 于产品迭代，很大程度上要取决于公司对下一代模型技术路径的押注，这才是决定AI产品的核心能 力。 事实上，无论是AI对话类产品还是AI助手，抑或AI浏览器，都可能只是AI杀手级产品的中间形态。对 豆包来说，无论是AI播客 ...

Core Insights - Quark has launched China's first AI model specifically designed for college entrance examination (Gaokao) application scenarios, featuring three core functions: deep search, application report, and intelligent application selection [3] - The model was fine-tuned by benchmarking against industry experts, ensuring that it meets the unique requirements of Gaokao application, which demands high accuracy and logical coherence [3][4] - The fine-tuning process involved creating a structured knowledge base covering over 2,900 universities and nearly 1,600 undergraduate programs, with a focus on data verification and cross-referencing [4] Model Fine-Tuning - The fine-tuning of the model is critical, involving targeted instruction adjustments and the collaboration of hundreds of experts to create a unique generation mechanism [4] - The team distilled thousands of past decisions from human experts to develop a reasoning chain that informs the model's decision-making process [4] - The model's design includes a mechanism to prevent "hallucinations," ensuring that the final recommendations are based on real data and historical validation [4] Agent Product Development - Quark's team served over 30 million users, with 50% being students from third-tier cities or below, indicating a significant market reach [6] - The introduction of the Agent format represents a shift in the industry, with a focus on providing expert-level advice rather than just public information [6] - The year is seen as a pivotal moment for Agent products, with major tech companies rapidly advancing their offerings in this space [6][7] Market Trends - The global AIGC technology penetration is projected to exceed 40% by 2025, with the AI Agent market expected to grow from $5.1 billion in 2024 to $47.1 billion by 2030, reflecting a compound annual growth rate of 44.8% [7]

「奔向AGI」 栏目聚焦AI大模型、AI agent、AI应用、芯片、机器人等前沿、热门的AI技术和商业创 新。 作者丨薛皓皓 编辑丨关雎 图源 丨Midjourney 近期， OpenAI 创始人山姆·阿尔特曼表示， 2025 年将是 AI Agent 大规模应用之年。" Agent 作 为'数字员工'，将为各行业显著节省时间，提升效率，创造可观的商业价值。"他说。 简单而言， AI Agent 可替人执行查资料、订机票、 做 PPT 、写代码等任务，是能自主思考，替人 干活的"数字管家"。 今年， OpenAI 一口气推出两款 AI Agent 产品： Operator 可自主操作浏览器完成复杂任务，例如 在线订餐、购物和订票； Deep Research 能在半小时之内输出金融、科学、政策等领域的深度调研 报告。它还花 30 亿美元收购 了 编程 AI Agent 公司 Windsurf 。 在这疯狂增长的蓝海市场，有投资人认为创业机会在于垂直细分领域的 AI Agent ，而非定位于宽泛 to C 场景的通用 AI Agent ，矛头直指定位于通用 AI Agent 的 Manus 。 金沙江创投创始 ...

Investment Rating - The report maintains a "Recommended" investment rating for the AI industry, highlighting the continuous evolution of domestic agents and the promotion of universal protocols for systematic applications [2]. Core Insights - The AI agent field is experiencing rapid advancements, with capabilities doubling approximately every seven months since the release of ChatGPT in 2022, leading to an exponential increase in the tasks that AI agents can complete [5]. - Major internet companies are competing to capture the AI internet era's new interaction entry point, with AI agents inheriting the core attributes of the app era and showing advantages in service delivery, network effects, and developer ecosystems [9]. Summary by Sections Section 1: Domestic Intelligent Agent Technology Innovation - The report discusses the launch of ByteDance's Coze Space, a universal agent product designed to facilitate efficient collaboration between users and AI agents, enabling the completion of complex tasks [13][14]. - Coze Space features two collaboration modes: an exploration mode for lightweight tasks and a planning mode for complex needs, allowing for real-time user input during task execution [17]. Section 2: MCP Open Protocol Continues to Expand - The introduction of the Agent2Agent (A2A) open protocol by Google aims to enable communication and secure information exchange between different AI agents, enhancing collaborative task execution across platforms [69]. - A2A facilitates interactions between client and remote agents, allowing for capability discovery, task management, and secure collaboration, thereby increasing operational efficiency [72]. - Microsoft announced plans to leverage the MCP to create an open "agentic web," where AI agents can autonomously initiate tasks and collaborate with minimal human oversight [80]. Section 3: Related Targets - The report identifies several companies involved in AI agent development, including Zhuoyi Information, which is focused on low-code integrated development environments (IDEs) that incorporate AI capabilities to enhance coding efficiency [84]. - Puyuan Information is integrating advanced large model capabilities into its low-code development platform, enhancing developer productivity through intelligent assistance [89]. - Hehe Information has launched an MCP server service for document processing, significantly improving the efficiency of intelligent document handling tasks [90].

计算机行业 投资评级 推荐 维持 国产Agent不断演进，通用协议推进系统性应用 ——AI行业专题报告 证券研究报告 2025年6月9日 分析师：王朗 邮箱：wanglang2@gyzq.com.cn SAC执业资格证书编码：S0020525020001 目录 • 第一部分：国产智能体技术创新，C端产品百花齐放 请务必阅读正文之后的免责条款部分 2 • 第二部分：MCP开放协议持续扩圈，跨平台协作提升Agent商业价值 • 第三部分：相关标的 • 风险提示 1 国产智能体技术创新，C端产品百花齐放 AI Agent领域的新摩尔定律，显示智能 体的能力急剧快速提升。 2022年ChatGPT发布以来，大模型的能 力持续快速提升，带动AI智能体的能力 提升，进而促进人们竞逐研发更强大的 AI模型。 根据AI研究网站AI Digest发布的研究， AI智能体能够完成的任务时长呈现指数 增长的趋势，其中，任务长度指的是专 业人士完成这些任务需要的时间，从不 到30秒到超过8小时不等。目前，智能体 可以自主完成人类需要一小时才能完成 的编程任务，顶尖的AI系统可以完成的 任务长度正在呈指数级增长——每7个月 翻一番。 ...

近日，笔者在与IBM大中华区技术销售总经理、首席技术官翟峰的对话中，他也提出了同样的观点。在此基础上，翟峰还认为，AI的试验阶段已经结束，企 业核心竞争力将取决于定制化AI应用和可量化的业务成果方面。 在前不久举办的IBM Think大会上，IBM董事长兼首席执行官（CEO）阿温德·克里希纳（Arvind Krishna）曾指出，当前的AI需要聚焦四个层面，分别是：智 能体、数据、集成、基础设施。 在翟峰看来，企业想要通过AI获得收益的过程中，首先需要面对三个问题：第一，高质量的数据有没有？第二，有没有在用？第三，有没有发挥作用？"企 业级AI落地最重要的因素是数据，这是核心生产力，没有数据一切都是空谈，"翟峰进一步指出，"现在大家在谈的大模型，可能有很多互联网数据，但是企 业最核心的数据有没有整合出来？这是最关键的因素。" 另一方面，在AI技术风潮步入智能体时代的背景下，AI技术如何与传统自动化技术融合，成为接下来AI技术发展的关键点。 以制造业为例，在生产制造过程中，原本工厂就已经应用上了诸如自动化流水线、自动化机器臂等自动化设备，在AI时代，这些传统的自动化设备如何与 AI智能体融合，产生更大的力量是个 ...

智能体进一步改变了人机协作的范式：用户不需要主动操作工具，而是简单地发出指令，让智能体去完成一系列复杂的任务。 01 智能体的"魔力"：下一个交互入口 开始讨论前，不妨先花点时间了解下"智能体"的概念。 随着Deepseek、Manus火遍全网，科技圈的热点过半和大模型、智能体有关。 4月22日的酷开春季发布会上，由影音、健康、生活、设备、创作、教育六大智能体构成的超级智能体正式亮相； 3天后的百度AI开发者大会，李彦宏连发通用超级智能体心响APP、内容操作系统沧舟OS等多款AI应用； 5月中旬举办的红杉AI峰会，意料之中地将"智能体"列为核心议题，坦言AI有着"10倍于云计算"的市场潜力； 以及接踵而至的谷歌I/O 2025、微软 Build 2025等开发者大会，无不提到了智能体，涵盖编程、医疗、金融等多个行业……无论是微软、谷歌、OpenAI等 海外巨头，阿里、腾讯、百度、酷开等国内企业，还是红杉代表的资本机构，无不开始大力推进智能体。 与之对应的问题是：到底什么是智能体，为什么"大厂"都在竞逐智能体，又将会带来哪些改变？ 智能体是英文AI Agent，其中Agent的含义中有"代理人"的意思，让智能体 ...