整理 | 华卫 刚刚，OpenAI 正式对外推出了 AI 编码神器 Codex，其目前向 ChatGPT Plus 用户开放。据 悉，Codex 在限定时段内提供宽松的使用额度，但在需求高峰期间，可能会对 Plus 用户设置速 率限制，以确保其能广泛可用。 并且，现在 Codex 可以在任务执行过程中访问互联网了，用户可用其安装基础依赖项、运行需 要外部资源的测试、升级或安装构建新功能所需的软件包等。互联网访问功能将向 ChatGPT Plus、Pro 和 Team 用户开放，之后也将支持企业版用户。不过，该功能日常默认处于关闭状 态，在创建新环境或编辑现有环境时可随时启用。 "我们不再处于集成开发环境（IDE）时代了。我们进入了自主开发时代。"有网友激动地表 示，"这是一个能自行搜索、安装、编码、测试、修复错误并发布的 AI。"此前 OpenAI 的 CEO Sam Altman 曾坚称，到今年年底，AI 模型将能够超越初级软件工程师。 值得一提的是，在发布推出 Codex 前，OpenAI 首先宣布了用 Rust 重写 AI 命令行编码工具 Codex CLI 的消息，称此举可提升性能和安全性并避免对 N ...

Group 1 - OpenAI has rewritten its AI command-line programming tool Codex CLI in Rust to enhance performance and security while eliminating dependency on Node.js, which may frustrate some users [1][2] - Codex CLI is an experimental programming agent that can run in the ChatGPT web browser environment and locally, featuring a chat-based user interface that operates in both interactive and silent modes [1] - The source code of Codex CLI was originally in TypeScript and required Node.js, but the team has now completed the rewrite in Rust [1][5] Group 2 - The decision to rewrite in Rust is driven by four key reasons: zero dependency installation, sandboxing requirements, performance optimization, and the ability to reuse existing Rust-based Model Context Protocol (MCP) implementations [2][4] - The current tool requires Node.js version 22 or higher, which may pose a barrier for some users [4] - Rust's lack of a runtime garbage collection mechanism leads to lower memory requirements, contributing to performance improvements [5] Group 3 - The Codex project currently has a composition of 46.7% Rust, making it the leading language, followed by TypeScript at 44.7% [5] - Codex CLI will still support code extensions in other languages such as JavaScript and Python, although specific details are yet to be disclosed [5] - Vue creator Evan You has also praised Rust for its efficiency, noting that the new version of Vite, using Rust-based Rolldown, has reduced production build times by 3 to 16 times and memory usage by up to 100 times [6]

"我们不再处于集成开发环境（IDE）时代了。我们进入了自主开发时代。"有网友激动地表示，"这是一个能自行搜索、安装、编码、测试、修复错误 并发布的 AI。"此前 OpenAI 的 CEO Sam Altman 曾坚称，到今年年底，AI 模型将能够超越初级软件工程师。 刚刚，OpenAI 正式对外推出了 AI 编码神器 Codex，其目前向 ChatGPT Plus 用户开放。据悉，Codex 在限定时段内提供宽松的使用额度，但在需求高 峰期间，可能会对 Plus 用户设置速率限制，以确保其能广泛可用。 并且，现在 Codex 可以在任务执行过程中访问互联网了，用户可用其安装基础依赖项、运行需要外部资源的测试、升级或安装构建新功能所需的软件 包等。互联网访问功能将向 ChatGPT Plus、Pro 和 Team 用户开放，之后也将支持企业版用户。不过，该功能日常默认处于关闭状态，在创建新环境或 编辑现有环境时可随时启用。 值得一提的是，在发布推出 Codex 前，OpenAI 首先宣布了用 Rust 重写 AI 命令行编码工具 Codex CLI 的消息，称此举可提升性能和安全性并避免对 Node.js 的依 ...

图片来源： Unsplash 5 月 16 日 ， OpenAI 宣布推出其迄今为止最强大的 AI 编程 Codex 的研究预览版。Codex 由 codex-1 驱动，这是该公司专为软件工程任务优化的 o3 AI 推理模型版本。 OpenAI 表示， codex-1 生成的代码比 o3 "更清晰"，能更精准地遵循指令，并会迭代运行测试直至 代码通过。 Codex 运行在云端沙盒化的虚拟计算机中。 通过与 GitHub 连接， Codex 的环境可预先加载您的代码 仓库。 OpenAI 称，该 AI 编码代理撰写简单功能、修复漏洞、回答代码库相关问题及运行测试等任 务耗时从 1 分钟到 30 分钟不等。 OpenAI 表示， Codex 能同时处理多项软件工程任务，且在运行时不会限制用户访问其计算机和浏览 器。 用户打开 Codex 时所见界面： 图片来源： OpenAI Codex 即日起向 ChatGPT Pro 、企业版及团队版订阅用户开放。 OpenAI 表示初期用户将享有 " 慷慨 的使用权限 " ，但未来几周内公司将对该工具实施速率限制。 OpenAI 发言人向 TechCrunch 透露， 届时 ...

Core Insights - ChatGPT has introduced a new feature that allows direct integration with GitHub, enabling users to analyze code repositories and technical documentation effectively [2][4]. - This feature is designed to assist developers by answering questions related to codebases and summarizing code structures, thereby saving time without replacing human input [5][4]. - OpenAI continues to enhance its programming tools, recently launching Codex CLI and upgrading the ChatGPT desktop application to support various coding software [5]. Group 1 - The new Deep Research feature can be accessed by clicking "Deep Research → GitHub," allowing users to ask questions and receive detailed reports based on source code and pull requests [2][3]. - OpenAI plans to roll out this feature to ChatGPT Plus, Pro, and Team users in the coming days, with enterprise and educational versions to follow shortly [4]. - The integration respects organizational privacy settings, ensuring users only see GitHub content they are authorized to access [5]. Group 2 - OpenAI has invested approximately $3 billion in acquiring the AI programming assistant startup Windsurf, indicating a strong commitment to developing programming-related applications [5]. - The company has also introduced fine-tuning capabilities for the o4-mini reasoning model, which will be available only to verified organizations, while the GPT-4.1 nano model's fine-tuning is accessible to all paid developers [6].

Core Insights - OpenAI has released two new models, o3 and o4-mini, which showcase advanced reasoning and multimodal capabilities, marking a significant upgrade in their product offerings [8][10][45]. - The o3 model is identified as the most advanced reasoning model with comprehensive tool use and multimodal capabilities, while o4-mini is optimized for efficient reasoning [8][10]. - The evolution of agentic capabilities in o3 allows it to perform tasks more like a human agent, enhancing its utility in various applications [14][15]. Group 1: Model Capabilities - The o3 model integrates tool use and reasoning processes seamlessly, outperforming previous models in task execution speed and effectiveness [14][10]. - OpenAI's approach to model training has shifted, focusing on creating a mini reasoning version first before scaling up, which contrasts with previous methods [9][10]. - The multimodal capabilities of o3 allow it to understand and manipulate images, enhancing its application in factual tasks [45][46]. Group 2: Agentic Evolution - The agentic capabilities of o3 enable it to perform complex tasks, such as web browsing and data analysis, with a level of efficiency comparable to human agents [14][16]. - There is a discussion on the divergence of agent product development into two technical routes: OpenAI's black-box approach versus Manus's white-box approach [15][16]. - Testing of o3 against classic use cases shows its ability to gather and analyze information effectively, although it still requires user prompts for optimal performance [16][19]. Group 3: Market Position and Pricing - OpenAI's o3 model is priced higher than its competitors, reflecting its advanced capabilities, while o4-mini is significantly cheaper, making it accessible for broader use [77][78]. - The pricing strategy indicates that all leading models are competing at a similar level, with o3 being the most expensive among them [77][79]. - The introduction of Codex CLI aims to democratize access to coding capabilities, allowing users to interact with AI models in a more integrated manner [64][68]. Group 4: User Feedback and Limitations - User feedback highlights some limitations in visual reasoning and coding capabilities of o3 and o4-mini, indicating areas for improvement [69][70]. - Specific tasks, such as counting fingers or reading clock times, have shown inconsistent results, suggesting that visual reasoning still requires refinement [70][72]. - Concerns have been raised regarding the coding capabilities of the new models, with some users finding them less effective than previous iterations [75][76]. Group 5: Future Directions - OpenAI's ongoing research into reinforcement learning (RL) suggests a focus on enhancing model performance through experience-based learning [81][85]. - The concept of "Era of Experience" emphasizes the need for agents to learn from interactions with their environment, moving beyond traditional training methods [85][88]. - Future developments may include improved planning and reasoning capabilities, allowing models to better integrate with real-world applications [89][90].

整理 | 核子可乐、褚杏娟 近期， Claude Code 迎来一次品牌危机。据报道，部分是 Anthropic 向一位试图对 Claude Code 进行逆向工程的开发者发出了下架通知，而 Claude Code 的使用许可比 Codex CLI 更为严格。 据悉，OpenAI 的 Codex CLI 源代码遵循 Apache 2.0 许可证，允许分发和商业使用。Claude Code 遵循 Anthropic 的商业许可证，限制未经公司明确许 可的修改。 据报道，Anthropic 还对 Claude Code 的源代码进行了"混淆"。Claude Code 的源代码不易获取。当一位开发人员对其进行了去混淆处理并在 GitHub 上 发布源代码时，Anthropic 提交了一份 DMCA （数字千年版权法案）投诉，要求删除该代码。 为此，开发者们认为这种做法还不如 OpenAI 发布 Codex CLI 时的开放态度。"真的觉得 Anthropic 的态度非常令人厌烦。"有网友评价。 其他几位开发人员同样分享了自己在单条提示词中取得出色编码成果的难忘经历。EverArt 创始人 Pietro Schira ...

我们在2025年Q1的大模型季报中提到，在AGI路线图上，只有智能提升是唯一主线，因此我们持续关注头部AI Lab的模型发布。上周OpenAI密集发布了o 系列最新的两个模型o3和o4-mini，开源了Codex CLI，还推出了在API中使用的GPT 4.1。本文将着重对这些新发布进行解读，尤其是o3 agentic和多模态 CoT新能力。 我们认为OpenAI在数次平淡的更新后，终于拿出了有惊艳表现的o3。融合了tool use能力后，模型表现已经覆盖了agent产品常用的use case。Agent产品开 始分化出两类路线：一类是像o3那样把tool use通过CoT内化到模型中，模型可以用写代码调用的方式执行任务；另一类是类似Manus，把工作流程外化 成人类OS中的computer use。同时OpenAI已经把agent产品作为了未来产品商业化收入占比的大头，我们有理由担心通用agent产品在大模型公司主航道上 被覆盖。 长线看，RL Scaling是进步斜率最大的方向，上周两位RL教父Richard Sutton和David Silver发布了一篇很重要的文章Era of Experience， ...

当前，AI 领域呈现出一种近乎"追星式"的热情氛围，每当有新的东西发布，便迅速引发广泛关注与高度评价，仿佛技术变革即将一触即发。同时大 家情绪也波动剧烈，从"危机论"到"爆发论"频繁切换。OpenAI 最近出的《A Practical guide to building AI agents》的指南，就是他们最近捧上天 的"神作"。它直接被捧成了"圣经"，一时间风头无两。 整理 | Tina 这份 34 页的指南被誉为"市面上最优秀的资源"，旨在为产品和工程团队提供构建 AI 智能体的实用方法，涵盖了 Agent 的定义、识别 Agent 应用 场景、设计框架、逻辑和编排模式等关键方面。 不过，以冷静理性著称的 LangChain 创始人 Harrison Chase 对 OpenAI 的这份指南中提出的一些核心观点表达了强烈异议，甚至表示该指南一 开始就让人感到"恼火"。他公开批评这份指南"具有误导性"，并罕见地进行了逐字逐句的分析。 他认为，OpenAI 在定义 Agent 时采取了一种过于僵硬的"二元对立"方法。实际上，目前大多数"Agentic 系统"都是 Workflows 和 Agents 的有机 ...

作者：cage, haozhen 我们在 2025 年 Q1 的大模型季报 中提到，在 AGI 路线图上，只有智能提升是唯一主线，因此我们持 续关注头部 AI Lab 的模型发布。上周 OpenAI 密集发布了 o 系列最新的两个模型 o3 和 o4-mini，开 源了 Codex CLI，还推出了在 API 中使用的 GPT 4.1。本文将着重对这些新发布进行解读，尤其是 o3 agentic 和多模态 CoT 新能力。 我们认为 OpenAI 在数次平淡的更新后，终于拿出了有惊艳表现的 o3。融合了 tool use 能力后，模型 表现已经覆盖了 agent 产品常用的 use case。Agent 产品开始分化出两类路线：一类是像 o3 那样把 和 o3 的发布模式一样， OpenAI 的 reasoning model 都是先训练出一个 mini reasoning 版本，再 scale 到 一个 long inference time、full tool use 能力的模型上。 而之前 GPT 模型总是先训练出最大的模型，再蒸 馏到小模型上。这个策略值得探讨其原因，我们的猜测是 RL 算法比较脆弱， ...