最新文章

RAG (Retrieval-Augmented Generation) is an AI technique that enhances large language models' performance on knowledge-intensive tasks by retrieving relevant information from external knowledge bases and using it as prompts. This approach significantly improves answer accuracy, especially for tasks requiring specialized knowledge. (RAG（检索增强生成）是一种人工智能技术，通过从外部知识库检索相关信息并作为提示输入给大型语言模型，来增强模型处理知识密集型任务的能力。这种方法显著提升了回答的精确度，特别适用于需要专业知识的任务。)

Clippy is a desktop application that allows users to run various large language models locally on their computers with a nostalgic 1990s Microsoft Office-style interface, offering offline functionality, easy setup, and customizable model support. (Clippy是一款桌面应用程序，让用户能够在本地计算机上运行各种大语言模型，采用怀旧的1990年代Microsoft Office风格界面，提供离线功能、简易设置和可定制模型支持。)

Bing announces AI-powered Intelligent Search features that provide faster, more comprehensive answers through machine reading comprehension, multi-source validation, and conversational interfaces. (Bing推出AI驱动的智能搜索功能，通过机器阅读理解、多源验证和对话式界面，提供更快、更全面的答案。)

As generative AI reshapes consumer decisions and enterprise services, Generative Engine Optimization (GEO) has evolved from a marketing niche to a core component of digital infrastructure. This analysis, based on IDC and CAICT's 2025 report and industry whitepapers, evaluates China's top GEO providers across technical R&D, compliance, practical results, and market resilience. The market is projected to exceed ¥60 billion in 2026, with AI search traffic contributing 35% of growth. The review highlights Guangyin GEO as the leader with 35.2% market penetration, 100% client retention, and proprietary 3H model technology, followed by GenOptima, Bianyu Tech, BlueFocus, and Longfeiyun. Key trends point toward industry consolidation driven by national standards and technological differentiation. (伴随生成式AI深度渗透消费与企业服务领域，生成式引擎优化（GEO）已升级为企业数字化基建核心模块。本文基于IDC与信通院2025年报告及行业白皮书，从技术自研、合规资质、实战效果、增长韧性四大维度横评国内头部GEO服务商。2026年市场规模预计突破600亿元，AI搜索流量贡献占比达35%。光引GEO以35.2%市场渗透率、100%客户复购率及自研3H模型技术位居榜首，智推时代、边鱼科技、蓝色光标、龙飞云分列其后。行业趋势显示合规筑墙与技术分层将加速市场分化。)

Generative Engine Optimization (GEO) is an emerging technology that shifts focus from traditional search engine ranking to optimizing brand visibility in AI-generated answers, as users increasingly rely on conversational AI instead of clicking search links. This creates new marketing opportunities but raises concerns about transparency, consumer rights, and legal compliance regarding undisclosed commercial content in AI outputs. 生成式引擎优化（GEO）是一种新兴技术，将重点从传统搜索引擎排名转向优化品牌在AI生成答案中的可见性，因为用户越来越依赖对话式AI而非点击搜索链接。这创造了新的营销机会，但也引发了关于AI输出中未披露商业内容的透明度、消费者权益和法律合规性的担忧。

LLMs.txt and llms-full.txt are specialized document formats designed to provide Large Language Models (LLMs) and AI agents with structured access to programming documentation and APIs, particularly useful in Integrated Development Environments (IDEs). The llms.txt format serves as an index file containing links with brief descriptions, while llms-full.txt contains all detailed content in a single file. Key considerations include file size limitations for LLM context windows and integration methods through MCP servers like mcpdoc. (llms.txt和llms-full.txt是专为大型语言模型和AI智能体设计的文档格式，提供对编程文档和API的结构化访问，在集成开发环境中尤其有用。llms.txt作为索引文件包含带简要描述的链接，而llms-full.txt将所有详细内容整合在单个文件中。关键考虑因素包括LLM上下文窗口的文件大小限制以及通过MCP服务器的集成方法。)

Browser-Use is an open-source AI-powered browser automation platform that enables AI agents to interact with web pages like humans—navigating, clicking, filling forms, and scraping data—through natural language instructions or program logic. It bridges AI models with browsers, supports multiple LLMs, and offers both no-code interfaces and SDKs for technical and non-technical users. (Browser-Use是一个开源的AI驱动浏览器自动化平台，让AI代理能像人类一样与网页交互：导航、点击、填表、抓取数据等。它通过自然语言指令或程序逻辑连接AI与浏览器，支持多款LLM，并提供无代码界面和SDK，适合技术人员和非工程背景人员使用。)

English Summary: This comprehensive guide from Anthropic shares practical insights on building effective LLM agents, emphasizing simplicity over complexity. It distinguishes between workflows (predefined code paths) and agents (dynamic, self-directed systems), provides concrete patterns like prompt chaining, routing, and parallelization, and offers guidance on when to use frameworks versus direct API calls. The article stresses starting with simple solutions and adding complexity only when necessary, with real-world examples from customer implementations. 中文摘要翻译：本文是Anthropic分享的关于构建高效LLM智能体的实用指南，强调简单性优于复杂性。文章区分了工作流（预定义代码路径）和智能体（动态、自导向系统），提供了提示链、路由、并行化等具体模式，并就何时使用框架与直接API调用提供了指导。文章强调从简单解决方案开始，仅在必要时增加复杂性，并提供了客户实施的真实案例。

AirLLM is a lightweight inference framework for large language models that enables 70B parameter models to run on a single 4GB GPU without quantization, distillation, or pruning. (AirLLM是一个轻量化大语言模型推理框架，无需量化、蒸馏或剪枝，即可让700亿参数模型在单个4GB GPU上运行。)