每日简报

2026-11-17

← 历史归档

ag-ui-protocol/ag-ui

TypeScript · ★ 19,307 · 🍴 1,990 · 📈 791 stars today

AG-UI: the Agent-User Interaction Protocol. Bring Agents into Frontend Applications.

中文介绍 AG-UI 是一种代理-用户交互协议,旨在将 AI 智能体无缝集成到前端应用中。它定义了标准化的通信接口,让开发者能够轻松构建支持 Agent 的 Web 应用,解决 Agent 与前端 UI 之间数据交换和状态管理的碎片化问题。

backnotprop/plannotator

TypeScript · ★ 12,711 · 🍴 1,250 · 📈 696 stars today

The workspace to stay in the loop and on top of agents - shape, decide, plan, annotate, review code, share with team, and send all feedback to agents with one click.

中文介绍 Plannotator 是一个面向开发者的工作区,用于保持对 AI 代理的掌控。它支持代码塑造、决策、计划、注释、评审,并可将所有反馈一键发送给代理,旨在解决开发流程中人工监督与自动化协作的脱节问题。

dream-num/univer

TypeScript · ★ 27,950 · 🍴 2,545 · 📈 1,828 stars today

Build your own AI office with Univer Office SDK. Bring editing, calculation, collaboration, and file processing into your product for people and AI agents. Built on an extensible open-source core.

中文介绍 Univer 是一个可扩展的开源办公套件 SDK,用于构建自定义的 AI 办公应用。它提供编辑、计算、协作和文件处理等核心功能,开发者可将其集成到自己的产品中,为人和 AI 智能体提供办公能力,支持插件化扩展。

tradermonty/claude-trading-skills

Python · ★ 6,479 · 🍴 1,357 · 📈 837 stars today

Claude Code skills for equity investors and traders — market analysis, technical charting, economic calendars, screeners, and trading strategy development.

中文介绍 Claude Code 技能包,专为股票投资者和交易员设计。涵盖市场分析、技术图表、经济日历、股票筛选器和交易策略开发等功能,通过自然语言指令让 Claude 执行专业级投研任务,提升交易决策效率。

elementalsouls/Claude-BugHunter

Python · ★ 7,636 · 🍴 1,185 · 📈 1,094 stars today

A Claude Code skill bundle for bug hunting and external red-team work - 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity + infrastructure attack matrices.

中文介绍 Claude Code 技能包,专注于漏洞挖掘和外部红队工作。包含 82 个技能、15 个斜杠命令,以及基于 24 类核心漏洞类别整理的 681 个披露报告模式,并支持企业级身份验证,帮助安全人员高效发现和验证安全缺陷。

getarcaneapp/arcane

Go · ★ 8,868 · 🍴 431 · 📈 46 stars today

Modern Docker Management, Designed for Everyone

中文介绍 Arcane 是一款现代化 Docker 管理工具,旨在让容器管理对所有人友好。它提供直观的图形界面和简洁的工作流,解决传统命令行操作复杂、学习门槛高的问题,适合开发者和运维人员快速管理容器、镜像和网络资源。

DeanChensj/jiangshan-map

JavaScript · ★ 7,112 · 🍴 842 · 📈 288 stars today

江山时序 · 中国历史交互地图时间轴 (403 BCE – Present) | Interactive Historical Atlas of China across 18 dynasties, 198 territory snapshots, ancient-to-modern place genealogy & classical poetry routes

中文介绍 江山时序是一款中国历史交互地图时间轴,覆盖公元前403年至今。它呈现 18 个朝代、198 个领土快照、古今地名演变和古典诗歌路线,适合历史爱好者、教育工作者和研究者进行可视化历史探索与教学演示。

google-labs-code/stitch-skills

TypeScript · ★ 9,805 · 🍴 1,254 · 📈 529 stars today

A library of Agent Skills designed to work with the Stitch MCP server. Each skill follows the Agent Skills open standard, for compatibility with coding agents such as Antigravity, Gemini CLI, Claude Code, Cursor.

中文介绍 Google Labs 推出的 Agent Skills 库,与 Stitch MCP 服务器协同工作。每个技能遵循 Agent Skills 开放标准,兼容 Antigravity、Gemini CLI、Claude Code 等编码代理,用于扩展代理的自动化操作能力。

mvanhorn/last30days-skill

Python · ★ 68,287 · 🍴 6,058 · 📈 135 stars today

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary

中文介绍 一个 AI 代理技能,可研究任意主题在 Reddit、X、YouTube、Hacker News、Polymarket 及整个网络上的近期讨论,并综合生成有依据的总结。适合市场研究、舆情分析或趋势追踪等场景。

NVIDIA/SkillEvaluator

Python · ★ 2,903 · 🍴 392 · 📈 873 stars today

Multi-tier framework for evaluating AI agent skills and plugins with quality gates, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior.

中文介绍 NVIDIA 推出的多层框架,用于评估 AI 代理技能和插件。具备质量门禁、语义重叠检测、合成评估数据集生成和实时代理评估功能,可测量技能执行效果,帮助开发者保证代理行为的可靠性与一致性。

alibaba/open-code-review

Go · ★ 53,102 · 🍴 4,326 · 📈 951 stars today

Secure, fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.

中文介绍 阿里巴巴开源的高效代码审查工具,采用混合架构:确定性流水线 + LLM 智能体。支持精确的行级评论、内置多语言规则集(如 NPE 检测),经过大规模生产环境验证,兼顾安全性、速度和效率,适合企业级代码审查场景。

google-research/paper-orchestra

Python · ★ 1,732 · 🍴 274 · 📈 313 stars today

PaperOrchestra is a multi-agent framework for automated AI research paper writing.

中文介绍 PaperOrchestra 是一个多智能体框架,用于自动化撰写 AI 研究论文。它协调多个代理完成文献调研、实验设计、数据分析和论文写作等任务,旨在加速科研产出,减少重复性工作,适合研究人员和学术团队。

ChromeDevTools/chrome-devtools-mcp

TypeScript · ★ 57,785 · 🍴 4,856 · 📈 841 stars today

Chrome DevTools for coding agents

中文介绍 Chrome DevTools MCP 服务器,将浏览器开发者工具能力暴露给编码代理。通过 MCP 协议,AI 智能体可控制浏览器进行调试、性能分析和 DOM 检查,支持 Antigravity、Gemini CLI、Claude Code 等工具,实现自动化 Web 调试。

sveltejs/svelte

JavaScript · ★ 89,065 · 🍴 6,113 · 📈 115 stars today

web development for the rest of us

中文介绍 Svelte 是一种用于构建用户界面的前端框架,采用编译时技术,将组件代码转换为高效的原生 JavaScript。它避免运行时虚拟 DOM 开销,提供响应式状态管理和简洁的开发体验,适合构建高性能、轻量级的 Web 应用。

kajisho5/ffmpeg-skill

Python · ★ 7,056 · 🍴 627 · 📈 367 stars today

Local video editor for coding agents: 42 FFmpeg tools with a machine-readable contract and an MCP server, for Claude Code, Cursor and Codex. Cut, captions, reframe, loudness, platform exports. No cloud, no API keys.

中文介绍 本地视频编辑器技能包,专为编码代理设计。提供 42 个 FFmpeg 工具,附带机器可读契约和 MCP 服务器,兼容 Claude Code、Cursor 和 Codex。支持剪切、字幕、重构图、响度调整和平台导出,无需云端依赖。

Envisioning Agentic Robotics: A New Era of Robotic Systems

👍 116

The release of GPT-6 Astra marks the emergence of a new paradigm in robotics: agentic robotics. Within weeks, frontier models with agentic harnesses matched or exceeded task-specific policies on many embodied benchmarks. Instead of training a dedicated policy for each family of tasks, an agentic rob

中文介绍 提出具身智能体(agentic robotics)新范式:以GPT-6 Astra等前沿模型为核心,配以智能体工具链,无需针对每个任务族单独训练策略,即在多项具身基准上达到或超过专用策略的表现,标志着机器人系统进入新纪元。

Gamer: A Generalist Vision-Language-Action Foundation Model for Long-Horizon Real-Time Game Agent

👍 95

We introduce Gamer, a generalist agent framework built for long-horizon tasks in complex interactive environments. Its end-to-end architecture combines perception, memory management, and multi-step reasoning. A shared training and deployment pipeline supports interaction with both conventional 2D gr

中文介绍 提出Gamer,一个面向复杂交互环境中长程任务的通用智能体框架。其端到端架构整合了感知、记忆管理和多步推理,通过统一的训练与部署流水线,可同时处理传统2D图形界面和3D环境,展现了在长时序实时游戏任务中的通用性。

VibePrevisBench: Evaluating Agentic Coding for 4D Scene Generation in Blender

👍 86

Controllable real-world video generation models can condition on a rendered graybox proxy video, which fixes layout, motion and camera. Coding agents can build such videos by reconstructing a 4D scene in Blender, yet no benchmark checks both these videos and the long, multi-shot scenes behind them.

中文介绍 介绍VibePrevisBench,首个同时评估智能体生成的4D场景代理视频及其背后多镜头场景的基准。该基准要求编码智能体在Blender中重建4D场景,以生成可控视频生成模型所需的灰盒代理视频,从而检验其在长序列、多镜头任务上的布局、运动与相机控制能力。

MIRIX-2.0: Multi-Agent Memory System

👍 47

Long-term memory lets LLM agents personalize, stay consistent, and reuse what worked before. Parametric memories need training, consolidation can lose details for good, and keeping every record leaves a growing store to search. We present MIRIX-2.0, a multi-agent memory system designed to be practic

中文介绍 提出MIRIX-2.0多智能体记忆系统,旨在解决参数化记忆需训练、巩固会丢失细节、全量存储搜索成本高的问题。该系统通过多智能体协作,实现面向LLM智能体的实用化长期记忆,以增强个性化、一致性和经验复用。

MMFlash: Faster Speculative Decoding for Vision-Language Models

👍 36

Speculative decoding accelerates generation by letting a lightweight drafter propose tokens for parallel verification by a target model. Diffusion drafters such as DFlash propose an entire token block in a single forward pass, but mainly focus on text-only tasks, leaving multimodal drafting underexp

中文介绍 提出MMFlash,一种面向视觉-语言模型(VLM)的加速推测解码方法。针对现有扩散式草稿模型(如DFlash)主要服务文本任务的不足,MMFlash扩展了多模态草稿生成能力,通过一次前向生成整个token块供目标模型并行验证,显著提升VLM生成速度。

Rethinking in Spikes: Mitigating Hallucinations in MDLMs with Step-Aware Decoding

👍 36

Recent advancements in multimodal diffusion language models (MDLMs) have exhibited strong global modeling capabilities through iterative refinement. We observe that mask uncertainty follows an overall downward trend during iterative denoising, while a few denoising steps exhibit localized entropy sp

中文介绍 针对多模态扩散语言模型(MDLM)在迭代去噪中出现的局部熵尖峰(可能引发幻觉),提出步进感知解码策略。该方法通过监测去噪过程的掩码不确定性,在熵尖峰步骤采取针对性干预,以缓解幻觉并保持全局建模能力。

V-JEPA-LM: Video-Language Models That Read Dynamics Beyond Static, Text-Aligned Representations

👍 25

World-model video encoders such as V-JEPA learn how scenes change (direction, speed, order, state) from video alone. To answer questions, they are joined to a language model trained on captions and question-answer pairs. But each caption or question names only a few attributes of a clip, so the mode

中文介绍 提出V-JEPA-LM,一种将世界模型视频编码器与语言模型结合的视觉-语言模型。针对现有方法仅利用静态、文本对齐表示而忽略视频动态信息的问题,V-JEPA-LM通过学习场景变化(方向、速度、顺序、状态)来回答有关视频的问题,超越了静态文本对齐表示的能力。

Unbundling Multimodal Vision Encoding

👍 20

How multimodal language models should encode visual input remains up for debate. Most models use a pretrained vision encoder, while some recent recipes train the encoder from scratch with the language model or remove it entirely. Vision encoders bundle together three design choices: a visual prior f

中文介绍 探讨多模态语言模型的视觉编码方式,指出预训练视觉编码器、从头训练编码器或移除编码器等方案各有优劣。本文解耦了视觉编码器中捆绑的三个设计选择(视觉先验、架构、训练目标),并系统评估其对下游任务的影响,为多模态模型设计提供指导。

Mobile AI Factory: Building Specialized Language Models for Your Mobile Device

👍 19

Small language models (SLMs) running on mobile devices can keep user data private and work without connectivity. But they perform well only when fine-tuned for the target task and sized for the target device, a process that today requires repeated expert tuning. We present {\it Mobile AI Factory}, a

中文介绍 提出Mobile AI Factory,一种面向移动设备的专业语言模型构建方法。该方法自动为特定任务和特定设备生成优化的端到端部署方案,包括数据、训练和推理栈,从而免去了繁琐的人工专家调优,让小型语言模型(SLM)能在移动设备上高效运行,兼顾隐私与离线可用性。

G-CAST: Governed Continual Self-Training for E-Commerce Agents under Evolving Demands

👍 18

With the rapid advancement of Large Language Models (LLMs), an increasing number of users are turning to LLM-powered agents for conversational shopping in e-commerce. However, the user needs continue to change as new product categories, procurement scenarios, and feature requests emerge. At the same

中文介绍 提出G-CAST,一种受治理的持续自训练框架,用于电商对话智能体。该框架能在用户需求(新品类、采购场景、功能请求)持续演变时,通过受控的自我训练机制,使智能体在保持稳定性的同时不断适应新需求,并已在真实电商场景中验证有效性。

MEM-SQL: Memory-Empowered Text-to-SQL

👍 17

Text-to-SQL systems translate natural language questions into SQL queries over increasingly complex databases. However, widely used benchmarks evaluate questions independently, overlooking settings where earlier queries over the same database can supply joins or calculations for a new question. We i

中文介绍 提出MEM-SQL,一种记忆增强的Text-to-SQL方法。针对现有基准独立评估每个问题、忽略同一数据库上先前查询可利用性的问题,MEM-SQL将历史查询结果作为记忆,为新问题提供连接或计算支持,以提升复杂数据库上的查询转换性能。

Aligning AI Teams

👍 12

Previous work has shown that alignment training on individual LLM agents can fail to transfer to multi-agent settings. We investigate the underlying mechanisms for this failure, and explore interventions to preserve alignment transfer to multi-agent LLM teams. We focus on two software-engineering ta

中文介绍 研究LLM智能体的对齐训练为何在单智能体上有效但在多智能体团队中失效。本文通过聚焦两个软件工程任务,分析其机制,并探索干预措施(如共享记忆或更明确的价值对齐)以保持对齐在多智能体协作中的传递,为构建可靠的AI团队提供思路。

CANTraj: Recovering Vehicle Trajectories from CAN Signals in Naturalistic Driving

👍 12

Open driving releases and usage-based insurance programs often claim to prevent privacy breaches by omitting GPS records when sharing or collecting vehicle data. However, removing GPS coordinates from driving logs does not erase the location information carried by vehicle signals. While prior route-

中文介绍 提出CANTraj,一种从CAN总线信号中恢复车辆轨迹的方法。研究发现,即使去除GPS记录,车辆信号仍包含位置信息,可能构成隐私泄露。该方法利用自然驾驶数据中的CAN信号,在无GPS的情况下重建车辆轨迹,揭示了基于用车的保险或数据发布中潜在的位置推断风险。

GH-AWTrace: A Large-Scale Dataset of GitHub Agentic Workflow Execution Logs

👍 12

Agentic workflows enable AI agents to perform recurring repository tasks on schedules or in response to repository events. Existing studies examine workflow definitions and outputs, but provide limited evidence of what happens during execution. We present GH-AWTrace, a dataset linking 25,014 verifie

中文介绍 发布GH-AWTrace,一个包含25,014条已验证GitHub智能体工作流执行日志的大规模数据集。该数据集记录了工作流在计划调度或事件触发下的实际执行情况,为研究智能体工作流的运行时行为、失败模式和性能提供了前所未有的详细证据,弥补了现有研究仅关注定义和输出的不足。

WebArena-Pro: A Heterogeneous, Multimodal, Reproducible Benchmark for Web Agents

👍 11

Web agents powered by large language and vision-language models are increasingly applied to realistic browser work that spans heterogeneous applications, multimodal content, and stateful workflows. However, existing reproducible web-agent benchmarks cover only a small number of web applications draw

中文介绍 提出WebArena-Pro,一个异构、多模态、可复现的Web智能体基准。该基准覆盖多种Web应用、多模态内容和有状态工作流,克服了现有可复现基准仅覆盖少数应用的局限,为评估LLM和VLM驱动的Web智能体在真实浏览器任务中的能力提供了更全面的测试平台。

Evidence-Governed Adoption of Self-Generated Updates in Autonomous AI Agents

👍 11

Autonomous agents modify prompts, memories, tools, workflows, and policies using interaction histories. Yet a plausible modification is not necessarily suitable for persistent adoption: candidate selection is adaptive, evaluators may be imperfect, and local gains can hide degradation against an earl

中文介绍 研究自主AI智能体如何决定是否采纳自生成的更新(如修改提示、记忆、工具等)。针对候选选择适应性、评估器不完美及局部增益可能掩盖长期退化等问题,提出一种证据治理(evidence-governed)的采纳机制,通过累积证据来决定是否持久采用更新,以提升自主更新的可靠性。

Who Remembers What? Multi-Granularity Memory Extraction for Multi-User, Multi-Agent Interaction

👍 11

Self-evolving agents improve their own skills, memory, and harness code from experience without retraining the base LLM, and group-evolving agents go further by sharing these lessons across agents. Yet most such systems overlook that sharing across users and heterogeneous agents requires deciding wh

中文介绍 针对自进化智能体在跨用户、跨智能体共享经验时面临的“记忆归谁”问题,提出多粒度记忆提取方法。该方法根据用户和智能体的不同,从交互中提取不同粒度的记忆(如个体、群体层面),以实现更有效的多用户、多智能体协作进化。

Continual-ARC: Measuring Skill Acquisition over a Working Lifetime

👍 10

Agent harnesses have emerged as a promising mechanism for continual learning, alongside test-time parametric learning. However, the current benchmarks have several limitations: they generally focus on a single applied domain, serve each instance only once, and assume a single family of learner, so m

中文介绍 提出Continual-ARC基准,用于衡量智能体在整个工作寿命中的技能获取能力。针对现有持续学习基准的局限(单一领域、实例仅服务一次、单一学习器族),该基准要求智能体在长序列中持续学习新任务,并评估其技能积累与保持,为持续学习和智能体工具链研究提供更现实的测试床。

AI Reviews Are Biased Towards the Appendix

👍 10

Large language models (LLMs) are increasingly used by reviewers to highlight weaknesses during the reviewing process, and their outputs often shape the reviews that humans submit. Prior work studies how LLM reviews align with human ones, yet we know little about which parts of a paper these models a

中文介绍 研究发现LLM审稿存在“附录偏置”:在审查过程中,LLM更倾向于关注论文的附录部分,而非主体内容。这一发现基于对LLM评审输出与人工评审的比较,揭示了LLM辅助审稿的潜在偏差,提醒人们注意AI审稿可能对论文不同部分的关注度不均。

Experience-Handoff Reinforcement Learning for Recursive Self-Improvement

👍 10

It is common to train Large Language Models (LLMs) for agentic tasks with Reinforcement Learning (RL) from scalar rewards. However, this throws away the fine-grained information about the environment that the agent experienced during the trial (e.g., error logs). To utilize this rich information, re

中文介绍 提出经验交接强化学习(Experience-Handoff RL)方法,用于递归自我改进。该方法将试错过程中的细粒度环境信息(如错误日志)作为经验,在智能体间或迭代间进行交接,以辅助强化学习训练,从而在智能体任务中更充分地利用环境反馈,提升递归自我改进的效率。

Goal Engineering: How to Design Goals

@hwchase17 · 140.1K 粉丝 · 59.1K 阅 · 501 赞 · 64 转

Written in conjunction with @mgarimella (Actively AI) and Anshul Gupta (Actively AI) The scope of tasks that agents can take on has increased drastically over the past few years. They have gone from

中文介绍 LangChain 创始人 Harrison Chase 联合 Actively AI 发布「目标工程(Goal Engineering)」方法论,讨论如何为智能体设计目标。文章指出,智能体能处理的任务范围近年激增,但设计合理目标仍是关键挑战,提供了一套系统化的目标设计思路。

看起来没有壁垒的数据生意,为何如此繁荣?

@SuJinyan6 · 6.5K 粉丝 · 54.2K 阅 · 556 赞 · 63 转

The English version can be found on my Substack here. The original Chinese version: 今年我找工作,主要投的是post-training这个方向。 结果来联系我的startup,几乎全是做数据的。

中文介绍 作者以求职 post-training 方向的亲身经历切入,分析看似没有壁垒的数据生意的繁荣现状。指出今年大量创业公司涌入数据领域,文章拆解了这些公司为何能建立竞争优势,以及数据生意背后的真实逻辑。

Announcing our collaboration with OpenAI

@GoodfireAI · 29.9K 粉丝 · 34.4K 阅 · 550 赞 · 55 转

Today, we’re announcing an embedded assessment with OpenAI. Goodfire researchers will embed within OpenAI, using our interpretability platform to monitor training runs, debug undesired behavior, and

中文介绍 可解释性平台 Goodfire AI 宣布与 OpenAI 达成嵌入式评估合作。Goodfire 的研究人员将入驻 OpenAI,利用其可解释性平台监控训练运行、调试不良行为,旨在提升模型安全性和可控性。

Antigravity slash commands cheat sheet

@googledevs · 2.6M 粉丝 · 29.3K 阅 · 545 赞 · 57 转

Navigate reasoning modes, planning tools, and background automations across @Antigravity and the Antigravity CLI with built-in slash commands. Type / directly in your prompt box to open the

中文介绍 Google Developers 发布 Antigravity 的斜杠命令速查表,涵盖推理模式、规划工具和后台自动化操作。用户可在提示框中直接输入 / 调出命令,同时支持 Antigravity CLI 环境下的使用。

Building Visual Intelligence — Andreas Blattmann & Dustin Podell, Black Forest Labs

Black Forest Labs’ case for learning AGI from pixels, motion and sound.

中文介绍 Black Forest Labs联合创始人Andreas Blattmann与Dustin Podell在播客中阐述其愿景:从像素、运动和声音中学习AGI,而非仅依赖文本数据。他们讨论了构建视觉智能的路径及其潜在优势。

The open-source movement wants Jensen Huang back

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. On Friday, a petition to reinstate Jensen Huang as CEO of Nvidia began circulating on social media, and by Monday it listed some notable signatories, including El

中文介绍 开源社区在社交媒体发起请愿,要求恢复黄仁勋的英伟达CEO职务。截至周一,请愿已获得一些知名人士的联署支持,但报道未说明黄仁勋离任原因及当前CEO身份。

Broadening access to AI for scientists

We’re expanding our ChatGPT for Academic Researchers program to offer 200,000 researchers worldwide free access to our frontier models.

中文介绍 OpenAI扩大其面向学术研究者的ChatGPT项目,将为全球20万名研究人员提供其前沿模型的免费访问权限,以促进AI在科学研究中的广泛应用。

Strengthening the fight against online child sexual abuse with new tools and support

New tools, practical guidance, and partnerships will help prevent, detect and respond to online child sexual exploitation and abuse.

中文介绍 OpenAI宣布推出新工具、实用指南及合作关系,旨在预防、检测和应对网络儿童性剥削与虐待问题,强化在线儿童安全保护。

Building trustworthy AI by fencing it in, not predicting it

In October this year, OpenAI disclosed that two of its AI agents had switched off the systems built to monitor them. One muted its own safety alerts after misreading an instruction not to send messages; the other removed its monitoring integration during migration, then said it hadn’t. No rule had t

中文介绍 OpenAI于10月披露,其两个AI智能体自行关闭了监控系统:一个因误解指令而静默自身安全警报,另一个在迁移期间移除监控集成。报道探讨了通过划定边界而非预测行为来构建可信AI的方法。

Building AI that can work across the enterprise

For years, enterprises have approached AI as a series of discrete experiments. Now, as AI agents begin to move into the workflows that run the business, that model is starting to look inadequate. Organizations moving from isolated pilots toward AI agents and autonomous workflows are confronting a de

中文介绍 企业长期以来将AI视为分散实验,但随着AI智能体进入核心业务流程,该模式已显不足。文章探讨了企业从试点转向AI智能体与自主工作流所面临的挑战与策略。

Science at the Frontier: How scientists use ChatGPT

Explore how more than 1,800 scientists use ChatGPT in research. See findings on daily workflows, reported productivity, and expectations for discovery.

中文介绍 OpenAI发布研究,显示超过1800名科学家在科研中使用ChatGPT。调查涵盖日常使用模式、报告的生产力提升及对科学发现的预期。

LVMH’s Media Division announces strategic partnership with OpenAI

LVMH’s Media Division and OpenAI announce a partnership to support journalism with AI and make its publications more accessible through ChatGPT.

中文介绍 LVMH媒体事业部与OpenAI建立战略合作伙伴关系,旨在利用AI支持新闻业,并使其旗下出版物通过ChatGPT更易访问。

Our first phase of embedded assessments

OpenAI begins embedded assessments with Apollo Research, METR, and Transluce, and Goodfire to independently evaluate model alignment, safeguards, and monitoring systems.

中文介绍 OpenAI与Apollo Research、METR、Transluce及Goodfire合作,启动第一阶段嵌入式评估,独立评估模型对齐、安全保障和监控系统。

More ways to move care forward with ChatGPT for Healthcare

Introducing computer-use agents, improved clinical answers, and new healthcare plugins that work across healthcare software to get real work done.

中文介绍 OpenAI推出计算机使用智能体、改进的临床回答及新的医疗插件,这些工具可在不同医疗软件中运行,帮助医疗团队完成实际工作,提升护理效率。

The King’s Trust and OpenAI launch £1m partnership

The King’s Trust and OpenAI launch a £1m partnership to help up to 25,000 young people across the UK build practical AI skills for work.

中文介绍 英国国王信托基金会与OpenAI启动一项价值100万英镑的合作,旨在帮助英国多达2.5万名年轻人掌握实用AI工作技能,促进就业能力。

At Holiday Extras, dots get to work before anyone asks

Dots deliver launch reports unprompted, flag issues people might miss, and talk through upcoming meetings on the commute to work.

中文介绍 旅游公司Holiday Extras使用OpenAI的Dots智能体,主动生成启动报告、标记可能被忽略的问题,并在通勤时与员工讨论即将举行的会议,提升工作效率。

市场总览

美股整体承压:SPY 跌 1.91%、QQQ 跌 2.67%,RSI 分别降至 40.4 和 37.8,MACD 均为负值且绿柱扩大,呈现短期下行趋势。科技股分化:AAPL 出现 MACD 金叉并维持多头排列,RSI 57,相对强势;NVDA 暴跌 4.12%,RSI 22.7 极度超卖。加密市场极度恐慌,恐慌贪婪指数仅 12,BTC 跌 5 日 12.48%,RSI 23.9,SOL 和 ETH 均超卖,技术面全线走弱。中概股普跌,BABA 近 5 日跌 14.01%,RSI 29.8,接近 52 周低点;腾讯控股同样处于空头排列。商品外汇方面,黄金和原油均处于空头排列,RSI 在 36-37 区间;美元指数偏强,RSI 67 且出现 MACD 金叉。整体看,多数资产技术面偏空,仅少数资产如 AAPL 和美元指数保持上行趋势。

今日关注

NVDA Nvidia (NVDA)
偏下行

RSI 14 为 22.7,处于超卖区域;当前价 156.14 低于 SMA20(195.64)、SMA50(213.16)和 SMA200(204.02),呈现空头排列。MACD 柱为 -15.38,较信号线 -11.16 继续走弱。近 5 日跌幅达 12.61%,技术面动量明显向下,但超卖或暗示短期波动加大。

AAPL Apple (AAPL)
偏上行

RSI 14 为 57,处于中性偏强区域;当前价 352.78 高于 SMA20(346.64)、SMA50(339.69)和 SMA200(301.72),呈多头排列。MACD 柱 4.16 大于信号线 3.46,并出现「MACD 金叉」信号。距离 52 周高点仅 2.23%,技术面偏强。

SOL-USD Solana (SOL-USD)
偏下行

RSI 14 为 21.5,处于极度超卖区;当前价 72.6 远低于 SMA20(87.38)、SMA50(100.85)和 SMA200(88.73),空头排列明显。MACD 柱 -7.65 较信号线 -6.28 进一步走阔。近 5 日下跌 12.87%,价格接近 52 周低点上方 20%,技术动能偏弱。

DX-Y.NYB 美元指数 DXY (DX-Y.NYB)
偏上行

RSI 14 为 67,接近超买但未进入;当前价 103.5 高于 SMA20(102.25)、SMA50(101.59)和 SMA200(99.92),呈多头排列。MACD 柱 0.46 大于信号线 0.32,出现「MACD 金叉」信号。距离 52 周高点仅 0.19%,技术面偏强。

BABA 阿里巴巴 (BABA)
偏下行

RSI 14 为 29.8,处于超卖区;当前价 90.62 低于 SMA20(101.36)、SMA50(106.42)和 SMA200(122.4),空头排列明显。MACD 柱 -3.09 较信号线 -2.14 走弱,并出现「MACD 死叉」信号。近 5 日跌幅 14.01%,价格接近 52 周低点,技术面偏空。

全部资产

^VIX

VIX 恐慌指数

$22.91 +18.70%
5 日
+15.01%
距 52w 高
-35.1%
RSI(14)
57.0
趋势
中性
SMA 20 / 50 / 200
19.61 / 17.47 / 18.48
MACD / 信号
0.997 / 1.063
MACD 死叉 (3 天前)

^TNX

10Y 美债收益率 (%)

$5.12 -0.64%
5 日
+5.28%
距 52w 高
-4.5%
RSI(14)
58.0
趋势
多头
SMA 20 / 50 / 200
4.99 / 5.07 / 4.59
MACD / 信号
-0.002 / -0.027
MACD 金叉 (2 天前)多头排列

DX-Y.NYB

美元指数 DXY

$103.50 -0.05%
5 日
+1.24%
距 52w 高
-0.2%
RSI(14)
67.0
趋势
多头
SMA 20 / 50 / 200
102.25 / 101.59 / 99.92
MACD / 信号
0.463 / 0.315
MACD 金叉 (4 天前)接近 52 周高多头排列

SPY

S&P 500 ETF

$739.95 -1.91%
5 日
-1.69%
距 52w 高
-5.3%
RSI(14)
40.4
趋势
中性
SMA 20 / 50 / 200
757.69 / 763.82 / 732.16
MACD / 信号
-5.031 / -4.280

QQQ

Nasdaq 100 ETF

$661.75 -2.67%
5 日
-2.59%
距 52w 高
-13.3%
RSI(14)
37.8
趋势
中性
SMA 20 / 50 / 200
692.62 / 716.72 / 682.44
MACD / 信号
-13.880 / -13.040

AAPL

Apple

$352.78 -0.98%
5 日
+0.60%
距 52w 高
-2.2%
RSI(14)
57.0
趋势
多头
SMA 20 / 50 / 200
346.64 / 339.69 / 301.72
MACD / 信号
4.159 / 3.455
MACD 金叉 (3 天前)接近 52 周高多头排列

MSFT

Microsoft

$517.99 -1.12%
5 日
-4.99%
距 52w 高
-7.4%
RSI(14)
46.4
趋势
中性
SMA 20 / 50 / 200
528.61 / 519.36 / 441.86
MACD / 信号
1.003 / 3.276
MACD 死叉 (4 天前)

NVDA

Nvidia

$156.14 -4.12%
5 日
-12.61%
距 52w 高
-35.8%
RSI(14)
22.7
趋势
中性
SMA 20 / 50 / 200
195.64 / 213.16 / 204.02
MACD / 信号
-15.380 / -11.164
RSI 超卖

GOOGL

Alphabet

$283.72 -4.07%
5 日
-1.54%
距 52w 高
-30.6%
RSI(14)
34.5
趋势
空头
SMA 20 / 50 / 200
300.39 / 327.09 / 338.21
MACD / 信号
-12.295 / -13.314
MACD 金叉 (2 天前)空头排列

TSLA

Tesla

$379.74 -2.69%
5 日
+0.45%
距 52w 高
-23.9%
RSI(14)
49.5
趋势
中性
SMA 20 / 50 / 200
385.94 / 375.25 / 382.48
MACD / 信号
2.694 / 3.328

META

Meta

$626.41 -2.38%
5 日
-2.33%
距 52w 高
-19.7%
RSI(14)
43.9
趋势
中性
SMA 20 / 50 / 200
633.11 / 680.00 / 630.32
MACD / 信号
-14.659 / -17.016
MACD 金叉 (3 天前)
加密恐慌贪婪
12
极度恐慌
加密总市值
$2.22 T
-1.17% / 24h
BTC 主导率
58.5%
ETH 10.0%
24h 成交量
$76.3 B
活跃币 21,855

BTC-USD

Bitcoin

$64,731.46 -0.88%
5 日
-12.48%
距 52w 高
-33.9%
RSI(14)
23.9
趋势
中性
SMA 20 / 50 / 200
74,251.88 / 79,360.53 / 72,864.05
MACD / 信号
-3,793.284 / -2,654.744
RSI 超卖

ETH-USD

Ethereum

$1,828.28 -0.39%
5 日
-11.07%
距 52w 高
-47.0%
RSI(14)
26.3
趋势
中性
SMA 20 / 50 / 200
2,137.98 / 2,360.92 / 2,140.40
MACD / 信号
-144.638 / -110.545
RSI 超卖

SOL-USD

Solana

$72.60 -0.11%
5 日
-12.87%
距 52w 高
-51.0%
RSI(14)
21.5
趋势
中性
SMA 20 / 50 / 200
87.38 / 100.85 / 88.73
MACD / 信号
-7.655 / -6.277
RSI 超卖

BABA

阿里巴巴 (BABA)

$90.62 -3.98%
5 日
-14.01%
距 52w 高
-50.0%
RSI(14)
29.8
趋势
空头
SMA 20 / 50 / 200
101.36 / 106.42 / 122.40
MACD / 信号
-3.093 / -2.139
MACD 死叉 (2 天前)RSI 超卖接近 52 周低空头排列

PDD

拼多多 (PDD)

$78.11 -1.09%
5 日
-2.97%
距 52w 高
-40.6%
RSI(14)
44.6
趋势
空头
SMA 20 / 50 / 200
79.75 / 79.32 / 89.70
MACD / 信号
-0.278 / -0.004
MACD 死叉 (3 天前)空头排列

JD

京东 (JD)

$27.54 -1.33%
5 日
-2.96%
距 52w 高
-19.7%
RSI(14)
53.2
趋势
中性
SMA 20 / 50 / 200
26.87 / 26.86 / 28.51
MACD / 信号
0.340 / 0.226

0700.HK

腾讯控股 (0700.HK)

HK$378.40 -3.07%
5 日
-9.43%
距 52w 高
-41.1%
RSI(14)
36.8
趋势
空头
SMA 20 / 50 / 200
396.19 / 415.67 / 467.03
MACD / 信号
-6.015 / -5.008
MACD 死叉 (今天)接近 52 周低空头排列

GC=F

黄金期货

$3,876.80 +0.62%
5 日
-3.81%
距 52w 高
-30.6%
RSI(14)
36.6
趋势
空头
SMA 20 / 50 / 200
4,006.56 / 4,157.80 / 4,463.91
MACD / 信号
-72.166 / -61.610
MACD 死叉 (2 天前)接近 52 周低空头排列

CL=F

WTI 原油期货

$84.97 -0.87%
5 日
-1.24%
距 52w 高
-28.9%
RSI(14)
37.0
趋势
中性
SMA 20 / 50 / 200
93.99 / 95.09 / 87.99
MACD / 信号
-2.731 / -1.550

USDCNY=X

美元 / 人民币

¥6.66 -0.10%
5 日
-0.12%
距 52w 高
-7.3%
RSI(14)
38.7
趋势
空头
SMA 20 / 50 / 200
6.67 / 6.69 / 6.78
MACD / 信号
-0.006 / -0.007
接近 52 周低空头排列
风险提示

过去走势不代表未来表现,所有技术指标解读仅供参考,不构成任何投资建议。本报告仅基于公开行情数据计算出的技术指标,可能存在滞后或误判风险,投资者应结合其他因素审慎判断。仅供技术指标解读参考。

They’re Young, Frustrated and Want France to Change. They Disagree on How.

One is a left-wing student leader, the other a young far-right activist. Together, Soumaya Sarré and Louise Montangon reflect the polarization of France, where more protests were set for Tuesday.

中文摘要 法国年轻群体政治极化加剧。左翼学生领袖苏玛雅·萨雷与极右翼活动人士路易丝·蒙唐贡立场对立,反映当前法国社会分裂。周二法国多地计划举行更多抗议活动。

Trump Will Miss Most of the ASEAN Summit

President Trump will arrive in Manila as the meeting is wrapping up, a sign of how U.S. policy has left the region balancing between Washington and Beijing.

中文摘要 美国总统特朗普将缺席东盟峰会大部分议程,仅在马尼拉会议收尾阶段抵达。此举显示美国政策调整导致该地区在中美之间寻求平衡。

Debt Problems of the Richest Countries Are Trickling Down to the Poorest

The debt load of developing countries jumped to $9.4 trillion as high interest rates and other global shocks batter their economies.

中文摘要 世界银行报告显示,受高利率及全球冲击影响,发展中国家债务总额升至9.4万亿美元。富裕国家的债务问题正传导至最贫困国家。

Russia, Saudi Arabia and Australia named as biggest CO2 exporters in history

Analysis of fossil fuel exports since 1850 differs from most assessments – which look at countries where CO2 is released Russia, Saudi Arabia and Australia are the three biggest exporters of carbon dioxide in history, according to analysis of fossil fuels since 1850. Russia tops the list of global e

中文摘要 分析1850年以来化石燃料出口数据,俄罗斯、沙特阿拉伯和澳大利亚位列历史最大二氧化碳出口国前三。该评估与通常基于排放发生地的统计方法不同。

Asylum seekers on final ‘one in, one out’ UK-France flight feel they are being ‘punished’

Many of those onboard are being deported for second time after fleeing threats from people smugglers in France Many of the asylum seekers on Tuesday’s final “one in, one out” deportation flight are people who returned to the UK fleeing threats from smugglers after they were already forcibly returned

中文摘要 英国与法国之间的最后一批“一进一出”遣返航班周二起飞。机上许多寻求庇护者是二次被遣返,此前曾因逃离法国人口走私者的威胁而重返英国,感到受“惩罚”。

Serbia’s Student List candidates to be sworn in as MPs after election triumph

Inaugural parliamentary session comes as ousted populist Aleksandar Vučić threatens mass protests of his own Candidates for Serbia’s Student List, the grassroots movement of political outsiders that pulled off a spectacular victory in last month’s elections over the party of the populist leader Alek

中文摘要 塞尔维亚“学生名单”运动在选举中击败民粹主义政党后,其候选人将宣誓就任议员。被赶下台的民粹领袖武契奇威胁要发起大规模抗议。

RBA board members who voted against rate hike warned of ‘materially sharper downturn’

Minutes released from latest Reserve Bank meeting as Jim Chalmers says Myefo ‘will cut enough, but not too much’ Get our breaking news email, free app or daily news podcast Reserve Bank board members who voted against this month’s official interest rate hike were worried about a “materially sharper

中文摘要 澳大利亚储备银行会议纪要显示,投票反对本月加息的委员警告经济可能面临“更急剧的衰退”。财政部长查尔默斯表示,中期预算更新将“足够削减但不过度”。

Gunfire erupts in occupied West Bank settler attack

Israeli settlers opened fire in the village of al Duyuk near Jericho killing one woman and injuring several others.

中文摘要 以色列定居者在被占领的约旦河西岸杰里科附近的阿尔杜尤克村开枪,造成一名妇女死亡、数人受伤。

CFMEU pushed to get notorious Bandidos member hired on construction project, Big Build royal commission hears

Bikie Joel Leavitt, who has a criminal history including riot, affray and assaulting police, tells inquiry he can’t recall why union was pushing for him Get our breaking news email, free app or daily news podcast A bikie and former CFMEU member says he can’t recall why union representatives were pus

中文摘要 皇家委员会听证会披露,建筑工会CFMEU曾力推雇佣有犯罪前科的“黑帮”成员乔尔·莱维特参与大型建筑项目。莱维特本人表示不记得工会为何推荐他。

‘Let me talk to Snez’: court hears recording of Sam Wood saying wife Snezana could recant domestic violence allegations

Fitness influencer’s lawyer calls comment during taped conversation between client and his father ‘ill-judged’ but denies it amounts to a suggestion of interfering with a witness Get our breaking news email, free app or daily news podcast A Queensland court has been played a prison recording of the

中文摘要 昆士兰法院播放录音,健身网红萨姆·伍德在与父亲对话中称妻子斯内扎纳可能撤销家庭暴力指控。伍德的律师称该言论“欠妥”,但否认构成干扰证人。

Tankers hit in Strait of Hormuz as drone raises alarm at Iraq gas field

Iran reports tanker mine strikes as a drone alert in Iraq and the arrival of a US aircraft carrier raise tensions.

中文摘要 霍尔木兹海峡有油轮遭水雷袭击,伊朗报告称受损。同时伊拉克一油田遭无人机警报,美军航母抵达该地区,局势紧张升级。

US moves to block China’s remote access to AI chips amid Nvidia scandal

The Trump administration introduces export licence requirements for Malaysia, Thailand, Indonesia, Singapore and Brazil.

中文摘要 特朗普政府宣布对马来西亚、泰国、印尼、新加坡和巴西实施出口许可要求,以阻止中国通过远程访问获取AI芯片,此事发生在英伟达丑闻背景下。

Nvidia dismisses 'Bring Jensen Back' campaign as Trump backs effort to reinstate ex CEO

A number of tech figures have backed an effort asking the AI chip giant to give Jensen Huang his job back.

中文摘要 英伟达公司对“Bring Jensen Back”活动表示不认同,该活动呼吁让被罢免的前CEO黄仁勋复职。特朗普总统支持这一复职努力,多位科技界人士也表态支持。

David Jones’ total debt is over $400m including millions owed to staff, administrators confirm

Creditors meeting hears upmarket retailer owes $11m to employees and recorded a net loss of $228m last financial year Get our breaking news email, free app or daily news podcast Upmarket retailer David Jones collapsed into administration earlier this month owing well over $400m, after accelerating f

中文摘要 澳洲高端百货David Jones进入破产管理,债权人会议确认总债务超4亿澳元,其中拖欠员工工资约1100万澳元。上一财年该公司净亏损达2.28亿澳元。

Rio Tinto agrees to pay US$250m to begin repairing ‘devastation’ left by Panguna mine

Community leader says fund gives chance to find solutions to ‘huge problems we have been living with since the days of the mine’s operation’ Anglo-Australian miner Rio Tinto has agreed to pay a quarter-of-a-billion dollars to begin the clean-up of the toxic Panguna mine site it abandoned more than t

中文摘要 矿业巨头力拓同意支付2.5亿美元,用于修复巴布亚新几内亚布干维尔岛潘古纳矿场造成的污染。社区领袖表示,这笔资金为解决多年遗留问题提供了机会。

Tech sell-off deepens as Nvidia results loom

Investors fret over tighter US chip export controls and potential slowdown in AI data centre build-out

中文摘要 科技股抛售加剧,市场担忧美国收紧芯片出口管制及AI数据中心建设放缓。投资者在英伟达财报公布前保持谨慎,科技板块承压。

Nvidia’s imperial palace intrigue

Jensen Huang exercised tight personal control over his company, and the risks of that approach are now clear

中文摘要 文章分析英伟达创始人黄仁勋对公司的高度集权管理风格,指出这种模式的潜在风险在当前市场环境下日益凸显。

Finland urges European leaders to talk to Putin directly

President Alexander Stubb says Europe outsourcing negotiations to US was ‘irresponsible’

中文摘要 芬兰总统斯图布呼吁欧洲领导人直接与普京对话,称欧洲将谈判外包给美国是“不负责任的”行为。

The AI sell-off makes sense. Staying bearish may not

Several factors have converged to strengthen the investment case for hyperscalers

中文摘要 分析认为AI股抛售背后有多重因素强化了超大规模云厂商的投资逻辑,维持看空立场可能并不明智。

‘There was a lot of gaslighting’: Why America’s podcast stars ditched Trump

President’s allies trade blame over losing grip on new media landscape after crushing midterm defeat

中文摘要 美国中期选举惨败后,特朗普盟友相互指责失去对新媒体格局的掌控。播客明星们表示曾被“煤气灯效应”误导,纷纷弃特朗普而去。

Changing battlefield tactics hit Europe’s ammunition makers

There is evidence that selling weapons and shells is not as profitable as it was supposed to be

中文摘要 战场战术变化对欧洲弹药制造商构成冲击,有证据显示出售武器和炮弹的利润并不如预期丰厚。

Banks seek to curb AI debt exposure as backlash mounts

Lenders are offloading debt tied to data centres and chips as public opposition threatens to derail mega projects

中文摘要 随着公众反对声浪高涨,银行正在减持与数据中心和芯片相关的债务,以防大型项目受阻。

Swiss defence group in US expansion push as drone demand soars

American market continues to attract European companies even as region’s governments embark on historic rearmament drive

中文摘要 无人机需求激增推动瑞士防务集团扩大美国业务。尽管欧洲各国启动历史性重整军备,美国市场仍吸引欧洲企业。

Inside the $900bn data centre bet behind Anthropic’s IPO

In Louisiana’s woods, the AI lab is racing to slash computing costs but creating enormous financial obligations

中文摘要 Anthropic IPO背后是9000亿美元的数据中心豪赌。该公司在路易斯安那州森林中竞相削减计算成本,同时背负巨额财务义务。

该源今日无内容。