最近真是捅了 MCP 窝了，上周火山开了一次开发者见面会，发布了挺多东西的，主要有： RTC 硬件这个也不太好测试，主要我也不懂，而且需要硬件，这次主要试一下 DeepSearch 服务。 其实现在所谓的 Agents 服务主要的任务和内容还是基于AI 搜索信息的加工和再整理，这部分是核心，也是 非常吃技术能力的地方。 火山把这部分能力变成应用之后对于开发者来说省了很多事情，人人都能搞 DeepSearch 了。 效果怎么样 先来一个最常见的问题和测试旅游规划。 即使这种看起来简单的任务很多 AI 搜索其实做的不好，看着内容输出很多，很多都是各个景点介绍的废话。 用户其实需要的是实时性比较强的信息，比如交通怎么安排，怎么样可以顺路，一些危险的项目需要准备哪些 东西等。 正式发布了豆包深度思考模型 Doubao-1.5-thinking-pro 和全新的视觉理解模型 Doubao-1.5-vision- pro，这个咱们上周介绍过了，视觉推理非常强大， 感兴趣可以去看我的测试 。 还发布了方舟 × RTC 硬件：把端侧自动唤醒与云端大模型语音能力一次打包，让玩具、家居、穿戴等设备 一键升级为能与人自然实时对话的 ...

Investment Rating - The report maintains an "Outperform" rating for the industry [1][37]. Core Insights - The emergence of general-purpose AI agents, such as ByteDance's "Kouzi Space," marks the beginning of competition among major tech companies, with expectations for rapid advancements from Alibaba and Tencent [2][11]. - The development of Model Context Protocol (MCP) is anticipated to standardize AI applications, enhancing the capabilities of AI agents and accelerating ecosystem expansion [8][11]. - The report emphasizes the potential of AI agents to redefine human-machine collaboration, providing new avenues for AI application development [33]. Summary by Sections ByteDance: Launch of General-Purpose Agent - ByteDance's "Kouzi Space" allows users to interact with AI agents for task completion, featuring a dual mode for exploration and planning [4][11]. - The platform supports the integration of MCP, expanding the capabilities of AI agents and facilitating the addition of various plugins [8][11]. Zhipu: Full-Stack Self-Development and Open Source - Zhipu's AutoGLM "Meditation" model showcases advanced research and operational capabilities, achieving state-of-the-art performance in multiple testing environments [15][19]. - The model's open-source approach is expected to foster ecosystem growth and accelerate application scenarios [20]. Genspark: Integration of Multiple AI Models for Accurate Responses - Genspark's Super Agent utilizes a mixed-agent system, integrating over 80 tools and 10 advanced datasets to enhance task execution efficiency [28][31]. - The platform has demonstrated superior performance in benchmark tests compared to other leading AI products [24][28]. Investment Recommendations - The report suggests that the development of AI agents opens new paradigms for human-machine collaboration, with a positive outlook on the future of AI applications [2][33].

2025年"AI Agent 之年"正逐渐走向现实。 3月初 Manus 的出现，真正让用户感受到了Agent产品与对话类大模型的本质区别：用户看到 AI 能够自主操作电脑、使用浏览器、编写代码。Manus 基本定义了 Agent 类产品的基本功能和产品 形态。 然而由于早期服务器资源的限制，Manus采用的邀请码注册形式，没有快速积攒很好的用户口碑， 最终没能像DeepSeek-R1那样将 AI Agent 的概念植入用户的日常生活。 但所幸字节跳动没有让用户等待太久。4 月 18 日，字节旗下首款 toC Agent 应用"扣子空间"正式 上线内测。 本文约 4300 字，推荐阅读时长 24 分钟，欢迎关注新财富公众号。 1 扣子空间实测：不俗的性能，完美的交付能力 字节对于扣子空间的定义，是精通各项技能的"通用实习生"与各行业的"领域专家"，具备无限拓展 能力的 AI Agent。 扣子开发平台（Coze）是字节在24年1月上线的低代码AI开发平台，由字节AI Seed 部门的 Stone 团队开发负责。扣子面向个人和企业用户，具有零代码开发、开放生态、任务自治等核心特点，致 力于让所有人都能快速、低门 ...

字节的 Agent 产品来了。 4 月 18 日晚间，字节跳动扣子空间开启内测，定位通用 Agent。与其他类似产品如 manus 一样，扣子空间采用了邀请码制。 平台上，用户可以选择精通各项技能的「通用实习生」，也可以选择行业的「领域专家」，通过与 AI 的互动完成工作任务。 就在前一天，火山引擎刚刚面向企业市场推出 OS Agent 解决方案及 AI 云原生推理套件，要帮企业更快、更省地构建和部署 Agent 应用。 扣子有探索和规划两种模式，如果想让它一步到位输出，可以选择探索模式；如果想亲自把控每个步骤，可以选择规划模式。 虽然采用了邀请码制，但扣子显然不是走饥饿营销的路线。用户激活获得的邀请码后，创建并完成一个新任务即可获得 5 个邀请码，邀请码激 活后，还可获得更多邀请资格，多创建多获得多邀请。 上线的第一时间，极客公园就上手实测了扣子空间的 Agent 功能。可以看到，在执行的三个任务中，制定旅游攻略和一周穿搭的任务顺利完 成，但另一个专家助手的任务，却出现了 Python 脚本调用失败、 API 权限异常等 Bug，一个晚上都没能执行成功。 字节迈出了走向 Agent 时代的第一步，但距离完美 ...

Market Overview - The market showed strength with a decrease in trading volume, with various concept stocks experiencing significant movements, particularly in the large aircraft and cross-border payment sectors [3] - The Shanghai Composite Index, Shenzhen Component Index, and ChiNext Index closed up by 0.45%, 1.27%, and 1.59% respectively, with over 4,300 stocks rising and around 800 declining [3] Robotics Sector - The first humanoid robot marathon took place, featuring 20 teams competing over a half marathon distance, with the winning robot developed by Beijing Humanoid Robot Innovation Center and UBTECH Technology [6][7] - The performance of Yushutech's G1 humanoid robot, which fell during the race, sparked discussions about its capabilities, despite the company clarifying that it did not participate directly [8][9] - The humanoid robot sector saw a 3.5% increase overall, with stocks like Tian工 and Songyan Power experiencing significant gains [9] MCP Concept - The MCP (Multi-Context Protocol) concept, set to launch in November 2024, aims to standardize applications for large language models, simplifying AI integration and reducing development barriers [11] - The announcement of ByteDance's new AI collaborative platform, "Coze Space," which supports MCP integration, acted as a catalyst for the surge in related stocks [12] - Stocks such as Dongfang Materials and Hengfeng Information saw significant increases, with some reaching their daily limit [12] Cross-Border Payment (CIPS) - The CIPS concept saw a resurgence, with stocks like Hailian Jinhui and Xinyada hitting their daily limits after the People's Bank of China announced plans to enhance the CIPS system and expand its global network [14] - The CIPS sector closed up by 3.41%, reversing an earlier decline [14] Paper Industry - The paper sector experienced a surge following news related to "intellectual property construction," with stocks like Annie Co. and Huatai Co. hitting their daily limits [16][17] - The increase in paper stocks was linked to rising raw material costs, particularly for specialty paper pulp, which has seen price fluctuations of nearly 50% due to supply chain issues [20]

要说折腾，现在特朗普称第二，没人敢称第一。之前用对等关税，差点把美债打到崩盘。上周特朗普说 考虑撤换美联储主席，使美联储的独立性受到质疑，投资者对美国经济的信心再次受到打击，对冲基金 大举抛售美元兑几乎所有货币，导致美元指数暴跌100多点，一度跌破98关口。 | 序号 代码 t | | 名称 | . 最新 | 淵鉄 | 涨幅% | 买一个 | 年初至今涨幅% | | --- | --- | --- | --- | --- | --- | --- | --- | | 1 | USDX | 美元指数 | 98.02100 | -1.21400 | -1.22 | | -9.64 | | 2 | USDSGD | 美元兑新加坡元 | 1.30158 | -0.01007 | -0.77 | 1.30152 | -4.30 | | 3 | USDJPY | 美元兑日元 | 140.56750 | -1.61150 | -1.13 | 140.56450 | -10.38 | | ব | USDHKD | 美元兒港币 | 7.76105 | -0.00082 | -0.01 | 7.76100 | -0.08 | | ...

Core Viewpoint - The report emphasizes the upcoming earnings season, suggesting to accumulate quality leading stocks at the bottom as earnings reports are set to be disclosed soon [2] Group 1: Earnings Reports and Market Trends - As of April 19, 2024, 133 companies have disclosed their 2024 annual reports and 12 companies have released their Q1 2025 reports, accounting for 40% and 4% respectively, with 8 working days remaining in the earnings season [2] - The computer sector has experienced continuous adjustments due to earnings disclosures and international developments, but the overall performance trend and industrial changes remain positive [2] - Historical data shows that since 2010, the probability of the computer index rising in April, May, and June is 20%, 53%, and 60% respectively, indicating a potential upward trend [2] Group 2: Industry Developments - The Model Context Protocol (MCP) is accelerating the application of AI, facilitating seamless interaction between AI models and external tools/data, which aids in the rapid development of Agent applications [3] - Major cloud service providers like Alibaba Cloud, Tencent Cloud, and Baidu Map have launched MCP-related services, enhancing the AI ecosystem [3] - The eCall industry is poised for growth, driven by increased importance of intelligent driving safety and rising domestic market demand, alongside potential opportunities in overseas markets due to geopolitical changes [3] Group 3: Industry Research - A retrospective study on the Xinchuang industry highlights its growth certainty, suggesting that the sector is expected to achieve excess growth in the future [4]

Core Insights - The introduction of "Payment MCP Server" by Alipay aims to facilitate the integration of payment capabilities into AI applications, addressing a significant barrier to AI commercialization [1][4][8] - MCP (Model Context Protocol) is likened to a universal interface for AI applications, streamlining the connection between AI models and external systems [2][4] - The service allows AI developers to implement payment functionalities using natural language, significantly reducing development time and complexity [6][7][8] Group 1: Payment MCP Server Launch - Alipay launched the "Payment MCP Server" on April 15, enabling AI developers to easily integrate payment capabilities into their applications [1][4] - The service is designed to bridge the gap in the AI ecosystem, facilitating smoother commercial transactions for AI services [4][5] - The initial rollout includes support for both mobile and web payment scenarios, catering to a wide range of AI applications [6][7] Group 2: Benefits for AI Developers - The Payment MCP Server reduces the complexity of payment integration, allowing developers to focus on core functionalities rather than payment processing [8][9] - It supports full payment management, including payment initiation, status queries, and refunds, enhancing user experience [7][8] - The service is expected to increase the commercial viability of AI applications by enabling a complete payment loop, thus improving data collection and service optimization [9][10] Group 3: Industry Implications - The introduction of this service is seen as a milestone for AI commercialization, potentially reshaping the payment landscape by integrating AI with various sectors like finance, retail, and healthcare [5][10] - Industry experts believe that this initiative will encourage other payment institutions to innovate and develop similar services, fostering a more robust AI payment ecosystem [10][11] - The ongoing evolution of AI protocols like A2A and MCP is expected to enhance communication between AI agents, further driving the development of AI applications [4][10]

以下文章来源于CSDN ，作者万有引力 CSDN . 成就一亿技术人 作者 | 万有引力 出品 | CSDN（ID：CSDNnews） 当 Manus 以其惊艳的自主任务执行能力点燃 AI Agent 领域的热潮时，其"一码难求"的现 状也让众多开发者望而却步。几乎在同时，一个名为 OpenManus 的开源项目以"火箭 般"的速度问世，不仅成功复刻了核心功能，更以完全开放的姿态，在短短不到一个月的时 间内于 GitHub 吸引了超过四万 Star 数的关注（截止本文发布，项目 Star 数已经达到 42.2k）。 OpenManus 项目 Star 数 这一现象背后，站着一群充满活力的 00 后程序员。他们利用下班后的短短三小时，凭借对 技术的热爱与开源精神，迅速将一个想法变成了现实。这种惊人的执行力与纯粹的"Just for Fun"动机，引发了业界的广泛讨论：这一代年轻开发者是如何学习、成长并拥抱前沿技 术的？他们与 AI 工具的深度协作达到了何种程度？支撑他们快速行动的技术积累和开源理 念又是什么？OpenManus 的诞生仅仅是复刻吗？其技术内核与未来方向又将如何演进？ 梁新兵 ： 向劲宇 ： Op ...