这场对话的双方，一位是特斯拉前AI总监、OpenAI创始成员 Andrej Karpathy，他是「Software 2.0」概念的提出者，一直站在编程范式转移的最前沿； 另一位是 Claude Code 的缔造者、Anthropic 的核心人物 Boris Cherny，他正在亲手打造终结传统编程的工具。 他们的讨论不仅仅是关于工具的迭代，更像是一场关于人类技能边界的哲学思辨。 当代码不再由人类一个个字符敲击而出，我们究竟是在进化，还是在退化？ 这场对话揭示了一个残酷而兴奋的事实： 我们正处于从「命令式编程」向「声明式意图」彻底转型的奇点。 Andrej Karpathy与Claude Code负责人Boris Cherny展开了一场关于编程未来的终极对谈。面对AI接管100%代码编写的现状，Karpathy坦言人类正处于 「脑萎缩」与能力进化的十字路口。 本文深度解析了从Software 2.0到Agentic Coding的范式转移，揭示了在Opus 4.5等强力模型加持下，程序员如何从「搬砖工」进化为「指挥官」，以及不 仅要面对效率的飞跃，更要警惕「垃圾代码末日」的隐忧。 2026年的开篇，科技圈被一 ...

Core Insights - Apple has officially released Xcode 26.3, integrating AI capabilities through Claude Agent from Anthropic and Codex from OpenAI, marking a significant shift in its development environment [1][2][4] - The introduction of "Agentic Coding" allows developers to use natural language commands for various tasks, enhancing productivity and efficiency [4][5] - Despite the powerful features, user feedback is mixed, with some praising the integration while others report issues such as system hangs and poor code comparison [9][10] Group 1: AI Integration in Xcode - Xcode 26.3 introduces predictive code completion and allows AI agents to perform complex tasks autonomously, such as creating files and debugging [4][5] - The integration aims to enhance the development experience by providing seamless access to advanced AI models [7] - Apple's collaboration with OpenAI and Anthropic reflects a strategic move to catch up with competitors like Microsoft, which has already integrated AI into its development tools [10][11] Group 2: Market Position and Developer Ecosystem - The rise of AI-centric coding tools like Cursor and Windsurf has prompted Apple to adapt its IDE to avoid losing developer interest [10][11] - The trend indicates a shift from plugin-based AI capabilities to native integration within productivity software, suggesting a closing window for independent AI tools [11] - The evolving landscape may challenge developers to adapt their skills while also providing opportunities for innovation and efficiency in coding practices [11][12]

Core Insights - Apple has announced the integration of Claude and Codex into Xcode 26.3, enabling "Agentic Coding" for developers on Apple platforms [1][4]. Group 1: Integration of AI Agents - Xcode 26.3 now supports both Anthropic's Claude Agent and OpenAI's Codex, marking the first native integration of these models for comprehensive app development [5]. - The upgrade allows developers to assign tasks directly to agents, which autonomously handle task breakdown, code modification, interface debugging, and documentation retrieval [3][4]. Group 2: Enhanced Development Capabilities - The new agents can understand project structures, execute complex tasks based on developer goals, and perform visual validation through Xcode Previews, improving UI code generation quality [5][8]. - Xcode generates a retrievable "node" for each change made by the agent, ensuring development safety [7]. Group 3: Model Context Protocol (MCP) - Xcode 26.3 introduces the Model Context Protocol (MCP), allowing compatibility with various external agent systems beyond Claude and Codex [10]. - This standard enables agents to explore project structures, modify files, preview views, and interact with Xcode in command-line environments, paving the way for broader AI system integration [10]. Group 4: Future Prospects - Apple emphasizes its collaboration with Anthropic and OpenAI to optimize token usage and tool invocation, enhancing agent efficiency and controllability [11]. - The integration of AI agents into Xcode positions it as a leading platform for AI agent deployment, with potential for future multi-agent collaborative development [11].

Hi, I’m Ken, I lead the Xcode team. With the release of Xcode 26.3%, we’re super excited to introduce agentic coding in Xcode, powered by tools using the Model Context protocol. Xcode and coding agents can now work together to handle complex, multi-step tasks on your behalf.Let me show you. I’m working on the Landmarks sample app, which lets users explore interesting places around the world – like national parks and mountain ranges. Now I'm going to add a feature to show the current weather at a landmark.Fi ...

OpenAI launches new MacOS app for agentic coding https://t.co/cRFwYari98 ...

【华泰计算机】Agenticcoding的加速崛起对软件/算力的影响 12025年是Agent元年，2026年将是Agent爆发的年份，体现在两方面：一是Agentic Coding（智能体编码）的迭代速 度会大幅加快；二是国内外大厂会激烈争夺个人Agent助手的超级入口，均会成为下一轮token加速的重要推手。 这种爆发，不仅会带来算力需求数量级的提升，也会带来存储和CPU新的结构性需求。 【华泰计算机】Agenticcoding的加速崛起对软件/算力的影响 12025年是Agent元年，2026年将是Agent爆发的年份，体现在两方面：一是Agentic Coding（智能体编码）的迭代速 度会大幅加快；二是国内外大厂会激烈争夺个人Agent助手的超级入口，均会成为下一轮token加速的重要推手。 这种爆发，不仅会带来算力需求数量级的提升，也会带来存储和CPU新的结构性需求。 当前算力通胀的链条正从存储、先进封装、CPU，向云服务和数据中心传导。 2它们有两个显著特点：1）200K tokens的超长上下文；cowork要加知识库来永久性的存储用户资料，这些都指向 了存储需求的增加，且呈现出从HBM逐步外 ...

今日早参 2026 年 1 月 27 日 易峘 首席宏观经济学家 邮箱：evayi@htsc.com 今日热点 宏观：进退两难的日元再度迎来"干预窗口" 上周五（1 月 23 日），日元在日央行发布会后以及美国交易时间出现快速升 值，截至周一（1 月 26 日）的两个交易日内累计升值 2.6%至 154.2 日元/美 元，背后主要是美国和日本释放出"利率检查"（"rate check"）信号，市场 猜测美日未来或联合干预日元汇率。历史上，日本当局在"利率检查"后较 短时间内就会启动汇率干预，若外汇干预落地，短期或推动日元升值。鉴于 日央行货币政策大幅落后于利率曲线，或已错失"优雅转身"的最佳时间窗 口：日本财政扩张预期叠加利率落后于曲线，日债收益率或易上难下，日元 资产亦或进入高波动阶段。 风险提示：日本众议院选举结果超预期；日本通胀水平超预期。 研报发布日期：2026-01-26 研究员 易峘 SAC:S0570520100005 SFC:AMH263 胡李鹏 SAC:S0570525010001 SFC:BWA860 陈玮 SAC:S0570524030003 SFC:BVH374 宏观：1 月 FOMC ...

Core Insights - The article discusses the significant growth of Anthropic's Claude Code, which has surpassed OpenAI's market share, reaching over 52% in recent months, driven by its superior coding capabilities [2] - The conversation highlights the evolution of AI in programming, emphasizing the shift towards "vibe coding" and the emergence of AI as collaborative agents in software development [3][4] Market Dynamics - Anthropic's model share was around 25% from 2024 to early 2025, but it experienced a "hockey stick" growth in the last 3 to 6 months, indicating a rapid increase in adoption [2] - The article notes that the focus on coding capabilities has made Anthropic a preferred tool among developers, leading to its penetration into various use cases beyond traditional programming [2][11] Product Development - Mike Krieger, Anthropic's Chief Product Officer, outlines the future direction of "vibe coding," emphasizing the need for AI to enhance reliability and interaction, rather than just improving model intelligence [3][4] - The development of Claude Code was a strategic decision to allow the model to operate autonomously over longer periods, reflecting a belief in the model's potential for future capabilities [8][9] User Engagement - The article mentions that the initial release of Claude Code exceeded expectations, with users applying it in diverse fields such as bioinformatics and data science, indicating its versatility beyond coding [11] - There is a recognition that many users, including non-technical individuals, are beginning to explore AI tools for various applications, suggesting a shift in user demographics [12][19] Future Vision - Looking ahead to 2026, the focus will be on creating collaborative AI agents that can handle more complex tasks and integrate seamlessly into existing workflows, moving beyond simple tool usage [20][24] - The article emphasizes the importance of ensuring high-quality outputs from AI to genuinely enhance productivity, rather than adding to the workload [19][22]

Core Insights - The emergence of AI Agents is significantly reducing labor costs in software development, potentially allowing projects that previously took weeks or months to be completed in just hours or a week [1][4] - The article discusses the transformative impact of AI tools on development processes and suggests that 2026 may mark a pivotal turning point for the industry [1] Software Delivery Costs - The initial wave of cost reduction in custom software development was driven by the rise of open-source solutions, which eliminated high licensing fees associated with proprietary databases [2] - The complexity of software engineering has increased in recent years, leading to a perception that development costs have not decreased significantly [2] Cost Savings from AI Agents - AI Agents are expected to drastically lower labor costs in software development, with the potential for a 90% reduction in costs [5][4] - Traditional development processes require a small team for tasks such as CI/CD setup, data access organization, and extensive testing, which can take a month to complete [5] - With AI Agents, these processes can be completed in a matter of hours, allowing for rapid development and reduced communication overhead [6] Release of Potential Demand - The reduction in production costs does not merely lead to lower spending but can result in increased demand for software solutions, as illustrated by the Jevons Paradox [7] - Many companies have significant untapped software needs, and lowering development costs could lead to a surge in demand for new applications [8] Importance of Domain Knowledge - Despite the advancements in AI, human oversight remains crucial to ensure quality and direction in software development [10] - Developers who master AI tools will become highly efficient in solving business problems, leveraging their domain knowledge to enhance productivity [10] Future of Software Development - The combination of business experts and skilled developers using AI tools will enable rapid iteration and development, potentially reducing the need for large teams [11] - The industry is on the brink of significant change, with the potential for software development to evolve faster than anticipated as AI technology continues to advance [12] AI in Legacy Code Management - AI Agents can simplify the understanding and maintenance of legacy code, making it easier to identify bugs and suggest fixes [13]

Please join me in envisioning a world where software has zero bugs. Not just a few bugs, but actually literally zero bugs. Okay. Okay.Just bear with me now. So for most people, let's just say people who aren't software engineers, bugs are actually just not a very big part of their life. Period.Most of the apps that we use on our phones, our social media, our news, that stuff pretty much works most of the time. The camera works most of the time. Any of those most popular apps, banking, they work really well ...