Group 1 - The events are organized for professionals from large companies, startups, and entrepreneurs in the product and technology sectors, focusing on AI-related themes [1][3][5] - The first event will take place on April 14 in Beijing, discussing the theme of AI Agents [1][5] - The second event is scheduled for April 18 in Shanghai, centered around AI content entertainment [1][5] - The third event will occur on April 21 in Shenzhen, focusing on AI hardware [1][5] Group 2 - Each event will have a limited number of participants (6-7 people) to facilitate meaningful discussions and networking [1][3] - Registration for the events is required by 8 PM the night before, with spots available on a first-come, first-served basis [1] - The organizers will group participants based on their backgrounds and interests to ensure everyone benefits from the experience [1]

Core Viewpoint - Genspark has transitioned from a traditional AI search engine to a more advanced AI Agent, Genspark Super Agent, which aims to provide comprehensive task execution capabilities rather than just information retrieval [5][9]. Group 1: Product Development and Features - Genspark Super Agent achieved a milestone of $10 million ARR just 9 days after its launch, although this figure is based on projected averages rather than actual revenue [3]. - The product is designed to autonomously think, plan tasks, take actions, and utilize tools to handle daily tasks, marking a significant evolution from its original AI search engine concept [5][6]. - The previous AI search engine attracted over 5 million users but was ultimately shut down to focus on the new AI Agent model, as traditional AI search was deemed increasingly outdated [6][8]. Group 2: Limitations of Traditional AI Search - Traditional AI search engines follow a linear, template-based response logic, which limits their ability to adapt contextually and dynamically plan task steps [9]. - Users require complete outputs rather than fragmented information, which traditional search engines struggle to provide [9]. - Genspark's shift to an AI Agent model is driven by the need for a system that can deliver comprehensive results, such as completed plans or presentations, rather than just raw data [9][10]. Group 3: Genspark Super Agent's Capabilities - Genspark Super Agent utilizes a Mixture-of-Agents framework, coordinating multiple specialized large language models (LLMs) to ensure stability and efficiency [10]. - It includes over 80 preset sub-agents and tools, enabling it to handle complex tasks like generating presentations or executing Python code [10]. - The system accesses carefully curated and verified datasets to ensure the accuracy and reliability of its outputs, reducing the spread of misinformation [10]. Group 4: Commercialization and User Experience - Genspark has introduced features for image and video generation, enhancing its functionality and offering a one-stop solution for various tasks [10]. - The commercial model includes packaged access to various models, with additional charges for executing tasks and generating media, while free users receive daily credits sufficient for moderate tasks [10][13]. - User feedback highlights the speed, model variety, and high success rate of Genspark Super Agent, with some users finding it easier to use compared to other platforms [18].

Core Insights - The article discusses the transformation of enterprise digitalization through the integration of Data and AI, emphasizing the shift from traditional SaaS and ERP systems to a new paradigm where AI can autonomously think and act, thereby enhancing productivity [1][12]. Group 1: AI for Data - Microsoft CEO Satya Nadella predicts that AI Agents will replace all SaaS, as traditional SaaS is being restructured by AI Agents [1]. - Companies must establish a digital system that encompasses perception, training, and result data to effectively harness AI's potential and convert it into tangible business value [1][3]. Group 2: High-Quality Data Assets - In the AI era, a shift in mindset is necessary, viewing data from an AI perspective and focusing on non-structured, high-quality data that AI can understand [3]. - Even structured data must be reorganized to be tokenized for AI models to comprehend and reason effectively [3][12]. Group 3: Systematic Data Consumption - Three clear paths for enabling AI Agents to consume data effectively include optimizing data structure, evolving data service methods, and integrating SaaS or Agents for comprehensive upgrades [4][9]. - The use of tools like Quick BI allows users to interact with data through natural language, generate reports, and receive personalized information streams, enhancing decision-making capabilities [5][11]. Group 4: Data Spiral Growth Effect - The establishment of a data spiral growth effect is crucial for AI-native applications, focusing on the continuous generation and utilization of data feedback [11]. - Activating dormant data assets is the first step in building this growth effect, with examples illustrating how previously overlooked data can be transformed into valuable training resources for AI [11][12]. Group 5: Agent Store Development - The company is developing an Agent Store that serves as a hub for integrating data, models, and application capabilities, facilitating the deep integration of data and AI in business scenarios [12]. - The evolution from information-based to intelligent systems marks a significant shift in how enterprises leverage data and AI, positioning them as foundational elements of a new productivity paradigm [12].

连续创业者马宇驰，在上一波人工智能大潮中创立三角兽科技，带领公司被腾讯收购，去年重回 AI赛道，参与到新一轮大模型应用创业中， 2033科技 于 2024 年 已经完成天使轮融资，由商汤科技和东方国资共同投资近亿元 人民币。 2033科技 是一家大模型 2C应用的人工智能公 司， 致力于打造 AIGC内容平台，降低用户使用门槛，辅 助用户进行 IP原创和二创， 满足用户和 IP的深度 交互 ， 用户可以在 平台上 高度自由 创造 "AI Agent+环境+情节+时间"的3D世界， 快速 将 喜欢的角色带到眼前，以 3D的形式呈现，方便分享转发给同 好，是 年轻用户的兴趣 内容平台。 经过 1年的筹备开发，目前2033科技的产品NYXverse，PC版已经登陆steam进行灰度测试，包括中国区在内全世界地区均可下载。 UGC 内容广场 NYXverse是UGC内容平台，由用户创造上传和定制自己喜欢的角色，目前平台上已经有数千个IP的Agent，支持用户自己上传VRM文件，降低用户形象创 作门槛，同时支持用户从名字到基础人设、背景故事和对话风格的高度自由定制。 用户创作塞尔达中的林克 用户创作的特朗普 用户在平台 ...

Core Viewpoint - The company, Zhejiang Wenlian (浙文互联), reported a strong financial performance for the fiscal year 2024, focusing on high-quality development and innovation in digital culture and marketing sectors [1][2][3]. Financial Performance - The company achieved an annual revenue of 7.703 billion yuan and a net profit attributable to shareholders of 157.77 million yuan [1]. - The gross profit margin improved by 2.3 percentage points compared to the previous year, reflecting effective cost management and business optimization [1]. - After accounting for asset impairment provisions, the adjusted net profit was approximately 371 million yuan [1]. Business Strategy - The company is committed to enhancing shareholder returns through stock buybacks totaling 50.08 million yuan and cash dividends of 73.80 million yuan [1]. - In the advertising and marketing sector, revenue reached 3.421 billion yuan, marking a year-on-year growth of 12.42%, with a strong presence in the automotive sector [2]. - The company is expanding its client base in non-automotive sectors, collaborating with major brands across various industries [2]. AI and Technology Development - The company is advancing its AI Agent commercialization, focusing on applications in digital marketing, culture, and other sectors, with significant R&D investments [3][4]. - The development of proprietary AI tools, such as the "Curiosity Series" AI Agents, aims to enhance content production and marketing efficiency [3]. Digital Culture and Infrastructure - The company is establishing a robust digital culture infrastructure, having delivered nearly 4,800 petaflops of computing power to clients in various industries [4]. - It is actively involved in the "Cultural + Technology" strategy, integrating cultural and technological advancements to support the development of a strong cultural province in Zhejiang [5][6]. Governance and Future Outlook - The company is optimizing its governance structure and enhancing risk management to support sustainable growth [7]. - With the upcoming "14th Five-Year Plan" concluding and the "15th Five-Year Plan" in preparation, the company is positioned to leverage policy opportunities for industry upgrades [8].

文章转载自「INDIGO 科技加速站」 Anthropic 工程师 Barry Zhang 在 AI Engineer 工作坊上的一个分享 "如何构建有效的 Agent"，其中印象最深的一个观点： Don't build agents for everything ，反过来理解就是别做什么都能干的 Agent，那是我们大模型要干的事情 构建有效 Agent 的三大要点： Barry 主要负责 Agentic System，演讲内容基于他和 Eric 合著的一篇博文，下面详细总结他们的核心观点，以及对 Agent 系统的演进和未来的思考。 Agent 系统的演进 1. 简单功能（Simple Features）： 起初是简单的任务，如摘要、分类、提取，这些在几年前看似神奇，现在已成为基础。 2. 工作流（Workflows）： 随着模型和产品成熟，开始编排多个模型调用，形成预定义的控制流，以牺牲成本和延迟换取更好性能。这被认为是 Agent 系统的前身。 3. Agent： 当前阶段，模型能力更强，领域特定的 Agent 开始出现。与工作流不同，Agent 可以根据环境反馈自主决定行动路径，几乎独立运 作。 4 ...

AI Agent 领域也存在 scaling law，甚至还在加速。 2022 年 ChatGPT 刚发布时能够实现的代码任务差不多等同于人类耗时 30s 的任务，到今天， AI Agent 已经能够自主完成需要人类花费一个小时的 coding 任务。"任务长度"是一个相当直观地测量 AI Agent 能力变化的标准。 编译：haozhen 编辑：Siqi AI 独立研究机构 META 的数据分析发现，Agent 能够完成的任务长度正以指数级增长，大约每 7 个 月翻一倍，预计 2029 年 Agent 能够完成时长为 1 个工作月的任务。 有意思的是，最近这一趋势甚至还在加速，2024-2025 年 Agent 能完成的任务长度约每 4 个月翻一 倍，如果这种更快的趋势持续下去，Agent 可能在 2027 年就能完成长达一个月的任务。 本文是对 META、Forethought 和 AI Digest 研究对于 agent scaling law 的整理编译。AI 研究人员们认 为，AI scaling law 的终局是 AI agent 自主开发 AI agent，到了那个时候我们就会进入软件智能爆炸时 ...

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 ...

西风 发自 凹非寺 量子位 | 公众号 QbitAI 不是辅助设计宣传海报or制定营销策略，新姿势是： 帮忙质检 ，不仅包括产品质量，还包括每个店面当前实时的运行情况。 比如库迪咖啡，门店数量众多巡检成本高，为了给顾客提供更好的店面环境、产品质量、人工服务，就用上了AI智能检测。 刚刚，在 阿里 云AI势能大 会 上， 阿里云智能集团资深副总裁、公共云事业部总裁 刘 伟 光 介绍了AI大模型的社会价值在企业市场释放的 一系列最新成果和新趋势。 在上述质检任务中，AI大小模型协同，视觉专家小模型负责业务目标的理解，通义千问VL大模型负责通用场景理解，还有阿里云提供的异步 工程链路提高吞吐量， AI质检整体准确率达95%，事件准确率达80% 。 为加速AI落地最后一公里，在大会现场，阿里云宣布 百炼上线业界首个全生命周期MCP服务 。 业界首个全生命周期MCP服务 MCP已被公认为大模型连接软件应用的标准协议。 AI大模型在咖啡店怎么落地？ 比如说，直接在百炼平台上选择通义千问大模型和高德MCP服务，就能快速搭建一个具备城市旅游美食规划的Agent应用。 这个Agent不仅能完成基础的地图信息查询任务，还可根据用 ...

编译：Xeriano 编辑：Cage 浏览器的使用者正在逐渐从人类用户转移到 AI Agent ，Agent 与互联网环境互动的底层设施也因此 正在变得越来越重要。传统浏览器无法满足 AI Agent 自动化抓取、交互和实时数据处理的需求。 Browserbase 的创始人 Paul Klein 早在 23 年底就敏锐地洞察到 AI Agent 亟需一个全新的交互载体 ——一个"为 AI 而生"的云端浏览器。这个浏览器不仅要解决现有工具的性能和部署问题，更核心的 是要利用 LLM 和 VLM 赋予浏览器理解和适应网页变化的能力，让 AI Agent 能用更接近自然语言的 方式与之交互，稳定地完成任务。 Browserbase 是一家成立一年多的 headless browser 服务提供商，以云服务的形式为 AI Agent 公司提 供 scalable、高可用性的浏览器服务。近期，Browserbase 又推出了 StageHand，一种利用 LLM 使得 开发者可以用自然语言与网页进行交互的框架，进一步拓展了其在 headless browser 领域的影响。 本文基于创始人早期备忘录进行了编译，详细阐述 ...