Core Viewpoint - JetBrains has decided to discontinue the development of its IDE Fleet, which has been in public preview since its launch in 2021, and will focus on a new development environment called Air aimed at agentic development [2][6]. Group 1: JetBrains and Fleet - JetBrains has a comprehensive suite of IDE products primarily based on the IntelliJ core platform, which has been in use since 2001 [4]. - Fleet was intended to be a lightweight, collaborative IDE to compete with Microsoft's Visual Studio Code (VS Code), which has gained popularity for its features [4][5]. - Despite some initial interest, most developers remained loyal to the IntelliJ series due to its robust plugin ecosystem and Fleet's prolonged public testing status [5]. Group 2: Discontinuation of Fleet - JetBrains announced that Fleet will no longer be available for download starting December 22, 2025, as maintaining two IDE product lines was causing user confusion and internal resource dilution [6]. - The company acknowledged that it failed to replace IntelliJ IDEA with Fleet or narrow its focus to a clear, differentiated niche [6]. - Although Fleet is being discontinued, its components will be integrated into other JetBrains IDEs, and the new product Air is an evolution of the Fleet platform [6]. Group 3: Introduction of Air - Air is designed to focus on a new workflow that leverages AI capabilities, allowing developers to delegate significant tasks to agents, which contrasts with traditional IDE workflows [7][8]. - The agentic workflow involves structured task definitions and asynchronous execution, which necessitates a different tool experience than traditional IDEs [8]. - Air is currently in public testing and will support multiple operating systems and cloud execution, enhancing its functionality beyond what Fleet offered [8]. Group 4: Developer Reactions and Market Position - Some developers expressed disappointment over the discontinuation of Fleet, believing it had the potential to compete effectively with VS Code and other emerging tools [10]. - The shift from Fleet to Air reflects a recurring pattern in JetBrains' strategy to adapt to evolving software development paradigms, particularly in the AI programming tool space [11]. - There are concerns about the necessity of creating a new tool rather than enhancing existing IDEs with AI features, raising questions about developer migration to Air [11].

在人工智能技术深度渗透各行业的2025年，AI大模型应用开发正经历从"单一问答"到"自主任务执行"的范式跃迁。以MCP（Model Context Protocol）协议为 核心的智能体开发模式，凭借其标准化工具调用能力与跨生态兼容性，成为企业级AI应用落地的关键基础设施。本文将从技术原理、生态构建、实战案例 三个维度，解析这一新范式的核心价值与实践路径。 一、MCP协议：AI工具调用的"万能插座" MCP协议由Anthropic于2024年11月推出，旨在解决AI模型与外部工具交互时的碎片化问题。其核心设计理念类似于USB-C接口——通过统一标准，让AI模 型能够像人类一样调用数据库、API、文件系统等外部资源。例如，某金融智能体可通过MCP协议直接连接Wind行情接口，实时获取股票价格数据；医疗智 能体则能调用医学知识图谱，生成符合诊疗规范的建议。 技术架构上，MCP采用客户端-服务器模式： 随着MCP协议的普及，AI智能体正从专业领域走向大众市场。2025年，低代码开发平台（如活字格）已集成MCP工具市场，开发者可通过拖拽方式快速构 建智能体应用。例如，某教育机构开发的"智能作业批改助手"，教师上传学生 ...

作为核心开发平台，ADP近期升级迭代：增强型 RAG 检索能力已支持主流数据库接入，工作流与 Agent 引擎提供更完整的协同配置方式，能够处理复杂任务；平台还完善了从设置、调试、评测、发布 到运营的应用全生命周期能力；同时持续拓展模型与插件生态，目前插件广场已经支持140 余种插件， 并沉淀了 70 余个应用模板与 90 余个提示词模板，极大降低了企业落地门槛。 产品力的提升，带动了丰富的行业落地场景，在企业知识问答、客服、专家知识助手等场景，腾讯云的 智能体方案已实现从0到1的打磨并在多行业复制；在媒体领域，通过多模态大模型、语音识别及 RAG 的结合，已能对视频内容做理解、摘录与检索，提升生产与治理效率，这套方案目前已经在多家权威媒 体落地，腾讯云ADP在省级媒体机构中的覆盖率，已经超过一半。 此外，腾讯云也大力推动优图 Agent、GraphRAG、YouTu-Embedding 以及ADP-Chat-Client 等开源工具 的开源；推出系统化的 ADP 课程，传递平台能力与最佳实践。面向国际市场，ADP 平台已正式在腾讯 云国际站发布，同时还发布了独立的产品落地网站，全球用户可以通过adp.ten ...

Core Concept - The article discusses the new course by Andrew Ng on Agentic AI, emphasizing the development of workflows that mimic human-like task execution through decomposition, reflection, and optimization [1][9][74]. Summary by Sections Agentic AI Overview - Agentic AI focuses on breaking down tasks into manageable steps, allowing for iterative improvement rather than generating a single output [5][14][74]. - The course reveals a systematic methodology behind Agentic AI, highlighting the importance of task decomposition and continuous optimization [9][10][74]. Core Design Patterns - The course identifies four core design patterns for developing Agentic workflows: Reflection, Tool Usage, Planning, and Multi-agent Collaboration [3][17][44]. Reflection - Reflection involves the model assessing its outputs and considering improvements, which can be enhanced by using multiple models in tandem [18][21]. - Objective evaluation standards can be established to assess outputs, improving the quality of the model's self-correction [23][27]. Tool Usage - Tool usage allows the model to autonomously decide which functions to call, enhancing efficiency compared to traditional methods where developers manually implement tools [28][34]. - The article discusses the importance of a unified protocol for tool calls, which simplifies the integration of various tools [41][43]. Planning - Planning enables the model to adjust the sequence of tool execution based on different requests, optimizing performance and resource use [46][48]. - A practical technique involves converting execution steps into JSON or code format for clearer task execution [47]. Multi-agent Collaboration - Multi-agent collaboration involves creating multiple agents with different expertise to tackle complex tasks, improving overall efficiency [51][52]. - This structured collaboration mirrors organizational structures, enhancing task division and scalability [52]. Iterative Improvement Process - The article outlines a feedback loop for building Agentic workflows, consisting of sampling, evaluation, and improvement [59][60]. - Error analysis is crucial for optimizing the system, allowing for targeted improvements based on specific performance issues [61][66]. Practical Insights - The course provides practical insights into selecting and testing different models, emphasizing the importance of iterative refinement in workflow design [68][70]. - The concept of Agentic AI represents a significant opportunity for developers to explore more complex, multi-step workflows, moving beyond traditional end-to-end agents [80].

Core Insights - Ant Group launched the Tbox super intelligent agent at the 2025 Inclusion Bund Conference, showcasing advancements in AI technology [1] Company Developments - The Tbox platform utilizes a "dynamic orchestration engine" that allows for real-time adjustments in the number of agents and their collaboration paths based on task complexity, offering greater flexibility compared to traditional serial processes [1]

Core Viewpoint - Beijing Lier has signed a strategic cooperation agreement with Shanghai SenseTime Technology Co., Ltd. and Hangzhou Xiwang Chip Technology Co., Ltd. to explore collaboration in AI computing power, industrial manufacturing, and decision-making AI vertical model development and application [1] Group 1 - The strategic cooperation aims to jointly research and explore AI computing power collaboration [1] - The partnership will focus on the development and application of vertical models for industrial manufacturing and decision-making AI [1] - The agreement includes the development and promotion of related intelligent agents [1]

Core Insights - The rise of intelligent agents has created a lucrative sector in technology, attracting a diverse range of developers, including those without programming skills [1][3] - The survival conditions for the first batch of intelligent agent developers in China are more complex compared to their counterparts in Silicon Valley [3][6] Developer Landscape - The first batch of intelligent agent developers in China can be characterized as "grassroots," with many lacking traditional programming skills but leveraging low-code platforms to create functional products [4][5] - The age range of developers is broad, with participants as young as 9 and as old as 51, indicating a unique phenomenon in the domestic market [5] Market Dynamics - The domestic low-code and no-code platforms have significantly lowered the technical barrier, allowing a wider range of individuals to participate in development [6][8] - The demand for niche solutions in China has amplified the value of creativity among grassroots developers, who focus on specific problems in various sectors [8] Revenue Challenges - Despite the flourishing development scene, the monetization of intelligent agents in China is significantly more challenging than in overseas markets, where subscription models are well-established [9][11] - Domestic developers often face low willingness to pay from end-users, leading to a long commercial startup cycle and low conversion rates from free trials [11][12] Survival Strategies - Grassroots developers have adopted pragmatic survival strategies, often integrating their agents into established ecosystems of major internet platforms to generate revenue [12][14] - Some developers focus on small, niche markets that larger companies overlook, allowing them to create stable income streams despite lower earnings [17][18] Future Directions - The development of intelligent agents has lowered the entry barrier for ordinary individuals, highlighting the vitality of grassroots developers in China's AI industry [22] - To attract more participants, low-code platforms need to enhance their commercial frameworks and support developers in monetizing their creations effectively [25][27]