每日简报

2026-11-08

← 历史归档

vercel-labs/agent-browser

Rust · ★ 46,905 · 🍴 3,214 · 📈 190 stars today

Browser automation CLI for AI agents

中文介绍 Vercel Labs 出品的浏览器自动化 CLI,专为 AI agent 设计,让 agent 能直接操控浏览器执行操作和验证结果。适合开发者在构建 AI 自动化工作流时使用,无需额外编写脚本,简化了 agent 与网页交互的流程。

cloudflare/security-audit-skill

JavaScript · ★ 65,846 · 🍴 4,277 · 📈 2,085 stars today

A coding-agent skill for multi-phase security audits with independently verified, machine-readable findings

中文介绍 Cloudflare 推出的编码 agent 技能包,用于执行多阶段安全审计,能生成可独立验证的机器可读结果。开发者可将其集成到编码 agent 中,自动发现并报告代码安全漏洞,提升审计效率与准确性,尤其适合 DevSecOps 场景。

google/skills

Python · ★ 23,457 · 🍴 1,960 · 📈 443 stars today

Agent Skills for Google products and technologies

中文介绍 Google 提供的 Agent Skills 合集,为 Google 产品和技术定制的能力模块,让 AI agent 可以更高效地与 Google 服务(如搜索、云平台)交互。开发者可将其应用于构建基于 Google 生态的智能助手或自动化工具。

mattpocock/skills

Shell · ★ 317,794 · 🍴 26,586 · 📈 1,100 stars today

Skills for Real Engineers. Straight from my .agents directory.

中文介绍 TypeScript 专家 Matt Pocock 分享的个人 Agent Skills 目录,汇集了他在实际工程中验证过的技能,供开发者借鉴和复用到自己的 agent 中,提升编码 agent 处理真实项目的能力,涵盖测试、重构等常见任务。

anthropics/skills

Python · ★ 188,857 · 🍴 22,261 · 📈 390 stars today

Public repository for Agent Skills

中文介绍 Anthropic 官方开放的 Agent Skills 公共仓库,提供一系列经过验证的技能定义,帮助开发者快速为 AI agent 添加结构化能力。这些技能可用于自动执行多步骤任务,如文档处理、数据分析等,兼容多种 agent 框架。

meta/astryx

TypeScript · ★ 14,622 · 🍴 1,197 · 📈 96 stars today

An open source design system that's fully customizable and agent ready

中文介绍 Meta 开源的设计系统,完全可定制且面向 AI agent 就绪。提供跨平台的 UI 组件与样式方案,开发者可快速构建一致性的界面,同时方便 agent 自动生成和修改 UI。适用于前端开发及 AI 驱动的界面生成场景。

prometheus/prometheus

Go · ★ 67,463 · 🍴 11,006 · 📈 322 stars today

The Prometheus monitoring system and time series database.

中文介绍 Prometheus 是云原生领域的事实标准监控系统与时间序列数据库。采用拉取模型采集指标,支持强大查询语言 PromQL、告警规则和多种可视化集成。适合运维与开发团队监控微服务、Kubernetes 集群的实时状态和性能。

sherlock-project/sherlock

Python · ★ 100,678 · 🍴 11,893 · 📈 1,766 stars today

Hunt down social media accounts by username across social networks

中文介绍 Sherlock 是开源的社交账号搜索工具,可在数百个社交网络上通过用户名查找用户的公开账号。利用各平台可探测性,支持批量查询并生成报告。适用于安全研究、隐私调查和 OSINT 场景,帮助追踪数字足迹。

milind-soni/OpenMausBot

TypeScript · ★ 17,110 · 🍴 2,416 · 📈 1,955 stars today

Open-source Grok Bot alternative with a virtual machine that bots can use

中文介绍 OpenMausBot 是开源 Grok Bot 替代品,内置机器人可用的虚拟机,支持自主运行和扩展。开发者和研究者可用它构建自定义的对话机器人,模拟 Grok 的行为,适合用于实验、教学或私有部署场景,无需依赖商业服务。

astral-sh/ruff

Rust · ★ 50,996 · 🍴 2,633 · 📈 350 stars today

An extremely fast Python linter and code formatter, written in Rust.

中文介绍 Ruff 是用 Rust 编写的超高速 Python linter 和代码格式化工具,兼容 Flake8、isort 等众多插件规则,比传统工具快数十倍。适合 Python 开发者用于 CI 流程和编辑器集成,大幅提升代码检查与格式化效率。

tailwindlabs/tailwindcss

TypeScript · ★ 98,511 · 🍴 7,111 · 📈 122 stars today

A utility-first CSS framework for rapid UI development.

中文介绍 Tailwind CSS 是实用优先的 CSS 框架,通过组合原子类快速构建自定义界面,无需离开 HTML 文件即可完成样式开发。适用于前端开发者,特别是 React、Vue 等组件化项目,可显著加快 UI 原型和产品开发速度。

mbillow/localthings

Python · ★ 1,078 · 🍴 113 · 📈 186 stars today

Control your Samsung appliances from Home Assistant over your home network, with no cloud account.

中文介绍 localthings 是 Home Assistant 的集成组件,允许用户在本地网络中直接控制三星家电,无需云账户。通过本地协议实现设备发现和控制,保护隐私且响应更快。适合智能家居用户,尤其是注重数据安全、不依赖外部云服务的场景。

zed-industries/zed

Rust · ★ 94,311 · 🍴 11,557 · 📈 1,114 stars today

Code at the speed of thought – Zed is a high-performance, multiplayer code editor from the creators of Atom and Tree-sitter.

中文介绍 Zed 是 Atom 和 Tree-sitter 作者打造的高性能多人协作代码编辑器,追求低延迟和极速响应。基于 Rust 构建,支持实时协作编辑和内置 AI 辅助。适合开发团队,特别是重视性能、协作流畅度以及现代编辑器体验的用户。

ranxianglei/billion-context

TypeScript · ★ 4,661 · 🍴 372 · 📈 191 stars today

A context-compression plugin for month-long single sessions (billions of tokens) ,small context windows (a 100K context is enough) and token savings (5x fewer tokens) .上下文压缩插件,月级长会话会话, 兼顾小窗口(100k上下文足矣)省token(省5倍token) . billion-context is all you need

中文介绍 billion-context 是一个上下文压缩插件,用于支持月级长会话,在 100K 小上下文窗口下可处理数十亿 token 的对话记录,节省约 5 倍 token 消耗。适合 AI 辅助编程或长时间对话场景,让 agent 在有限窗口下保持会话连续性。

dsx-ai-factory/nvke

Go · ★ 580 · 🍴 91 · 📈 123 stars today

NVIDIA Kubernetes Engine (NVKE). A management plane for GPU Kubernetes on bare metal.

中文介绍 NVKE 是 NVIDIA 推出的 Kubernetes 引擎管理平面,专为裸金属上的 GPU Kubernetes 集群设计。提供 GPU 资源编排、调度和监控能力,简化 AI 工作负载的部署。适用于需要高性能 GPU 计算的 AI 团队,如模型训练和推理场景。

VFlash: Fast, Accurate Visual Planning via Flash Primitive Decoding

👍 100

Models for physical AI tasks, such as robotics and autonomous driving, need not reason solely in language. They can use visual plans--points, bounding boxes, and trajectories--to represent spatial goals and anticipate motion. However, autoregressive vision-language models generate these plans one co

中文介绍 VFlash 提出 Flash Primitive Decoding 方法,将视觉规划解码为离散的图元(点、框、轨迹),以提升机器人和自动驾驶等物理 AI 任务的视觉规划速度和准确性。该方法支持快速生成,同时保持高精度,实验结果显示在视觉规划任务上实现显著加速,优于现有自回归方法。

From Autonomous Agents to Cognitive Symbionts: Toward AI-Augmented Human Intelligence

👍 86

For a decade, progress in artificial intelligence has been measured by how much a machine can do on its own. Large foundation models now write code and run multistep tasks, and personal agents act on a person's behalf. Yet people still forget what they once knew, miss what they are not looking at, f

中文介绍 本文反思 AI 发展从自主代理向认知共生体的转变,提出 AI 增强人类智能的框架。作者认为,未来 AI 应辅助人类记忆、注意力和决策,而非替代。通过分析当前大模型和代理系统的局限,提出认知共生体的设计原则,并展望其对社会和认知科学的影响。

SLA: Scaling Spatial-Language-Action Model via Continual Pre-training

👍 82

Reliable robotic manipulation requires both semantic understanding and precise control in three-dimensional space. However, vision-language-action (VLA) foundations largely inherit visual representations from 2D vision-language pre-training, while spatial supervision is often introduced only during

中文介绍 SLA 提出一种通过持续预训练扩展空间-语言-动作模型的方法,将 2D 视觉语言表征与 3D 空间监督结合。该方法解决了 VLA 模型在空间理解上的不足,增强了机械臂操作的精确性和语义泛化能力,在多个真实和仿真操作任务上验证了其优越性。

Decoupled Predictive Decision Making: A Shared Conditional Predictor for Agents

👍 80

Agent decisions depend on empirical relationships that deployment observations reveal and task descriptions alone may leave unresolved. We introduce Decoupled Predictive Decision Making (DPDM), an architecture that formulates heterogeneous agent decisions as conditional prediction with a shared pret

中文介绍 DPDM 提出一种解耦预测决策架构,将异构代理决策建模为共享条件预测问题,利用预训练预测器统一处理不同任务和观测。该方法避免了依赖任务描述的先验,通过经验关系预测决策,在多个代理任务上展现了更好的泛化能力和鲁棒性。

Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training

👍 63

Harness evolution from a common base exhibits Locality, Complementarity, and Nonstationarity: variants repair different failures, and their utility changes as the policy learns. Rather than seeking a universally superior harness, we introduce Harness-S to selectively rescue failures of a base execut

中文介绍 Harness-S 提出从同源兄弟 harness 中学习,以解决基座 harness 的失败模式。通过识别 harness 演化的局部性、互补性和非平稳性,选择性救援基座执行中的失败。实验表明该方法能显著提高代理训练的泛化性和效率,优于单一 harness 方法。

Random Is Almost Enough? Stable and Low-Cost Benchmark Compression via BestRand

👍 49

Benchmark compression reduces the often prohibitive time and cost of model evaluation by evaluating models on a small benchmark subset that closely matches full-benchmark results. To quantify estimation error, we derive an accurate analytical approximation to the error distribution when estimating f

中文介绍 BestRand 提出一种稳定低成本的基准压缩方法,通过随机选择最优子集来近似全基准评估。作者推导了估计误差的解析近似,证明随机选择几乎足够。该方法在多个基准上实现与全基准高相关性,同时大幅降低评估时间和成本,为模型评估提供实用方案。

WeLM-Image: A Unified Image Generation Model towards Visual Intelligence

👍 42

We present WeLM-Image, a unified image foundation model that integrates multimodal understanding into text-to-image generation, instruction-guided editing, and multi-reference synthesis. Designed for single-GPU deployment in WeChat's large-scale consumer applications, the model processes text, refer

中文介绍 WeLM-Image 提出一个统一的图像基础模型,集成多模态理解到文本到图像生成、指令引导编辑和多参考合成中。该模型专为单 GPU 部署设计,在微信大规模消费应用中处理文本、参考图像和编辑指令,实现高质量、可控的图像生成,并具备视觉理解能力。

Halley: A Predictive Control Plane for Optical AI Data Center Networks

👍 32

A frontier AI model is trained in an AI data center, on tens of thousands of accelerators joined by a dedicated network, and every training step ends with all of them exchanging data at once. Between racks, that network is optical, and each of its four layers, from lasers to switches, runs a fixed o

中文介绍 Halley 提出一个预测性控制平面,用于光 AI 数据中心网络,通过预测训练流量模式动态调整光网络各层配置。该方法能提前优化激光、交换机等组件,降低训练期间的网络拥塞,提高训练效率和稳定性。实验显示显著提升网络利用率,减少同步等待时间。

EvoMemState: Evolving Memory State for Personal Agents

👍 31

Long-term memory enables large language model (LLM) agents to provide continuous personalized assistance by retaining user information across interactions. Existing systems support this process through fact extraction, organization, updating, and retrieval, but largely maintain memories as independe

中文介绍 EvoMemState 提出一种演化记忆状态机制,用于 LLM 代理的长期记忆管理,通过动态更新和关联用户信息来增强个性化服务。与静态事实存储不同,该方法支持记忆的演化、冲突消解和上下文依赖检索,在长时间交互场景中提升记忆连贯性和代理决策质量。

VisSearch-Bench: Evaluating Long Horizon Visual Search in Multimodal Agents

👍 21

Multimodal agents use web and image tools to answer questions beyond their internal knowledge. Some search tasks use images mainly to identify an initial entity, then rely on text retrieval. These tasks offer limited insight into how agents connect information across images, objects, and search step

中文介绍 VisSearch-Bench 提出一个长时视觉搜索基准,用于评估多模态代理在跨图像、物体和搜索步骤间的信息连接能力。基准包含多样化任务,测试代理整合视觉和文本检索的复杂度。实验发现当前多模态代理在此类任务上仍有较大改进空间,为研究提供明确挑战。

Towards Foundation Decision Models: Drex

👍 21

Autoregressive approaches with large language models, such as long chain-of-thought reasoning, have become the default for new AI applications because they reach higher accuracy on a wide range of tasks. In production, however, they give inconsistent answers, require parsing of outputs, and add cost

中文介绍 Drex 提出基础决策模型,采用非自回归方法替代 LLM 的自回归推理,以解决长思维链在生产中的不一致性、解析开销和高成本问题。Drex 直接预测决策结果,保持高精度同时提升效率,在多个决策任务上展现更稳定和低成本的优势。

Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models

👍 20

Egocentric human video offers a scalable source of manipulation priors for robot learning, yet translating raw data scale into generalizable robot policies demands rigorous data curation and training recipes. We introduce Open-HumanScale, the first fully open recipe for scaling egocentric pretrainin

中文介绍 Open-HumanScale 提出首个完全开放的配方,用于扩展人形和机器人经验,以训练具身基础模型。通过严格的数据整理和训练策略,将第一人称人类视频转化为可泛化的机器人操作策略。该开放配方支持跨数据集扩展,实验显示在多个操作任务上显著提升性能。

D3D: Decoupled Diffusion Drafting for Asynchronous Speculative Decoding

👍 18

Target-conditioned diffusion drafters predict candidate tokens in parallel, yet each new block must wait for updated target-model hidden states. This forces drafting and verification to run sequentially, limiting speedup even though each block is generated in one pass. We introduce D3D, a standalone

中文介绍 D3D 提出解耦扩散起草方法,用于异步投机解码。通过独立生成草案,消除对目标模型隐藏状态的依赖,使起草和验证并行执行,突破顺序瓶颈。该方法在保持输出质量的同时显著加速推理,相比现有投机解码方法实现更高吞吐量和更低延迟。

TAC-CoT: Tool-Empowered Implicit Chain-of-Thought via Interleaving Explicit Tokens

👍 16

LLM-based agentic systems are increasingly combined with CoT reasoning to solve tasks that require accurate reasoning, external knowledge, and multi-step deliberation. Yet, explicit CoT incurs substantial generation overhead, while implicit CoT, which replaces textual rationales with compact continu

中文介绍 TAC-CoT 提出工具增强的隐式思维链方法,通过在显式思维链中交错工具令牌,以减少生成开销并保持推理准确性。该方法用紧凑连续表示替代文本推理,并结合外部工具调用,在多个推理任务上实现更高效的推理,同时保持或提高精度。

Hybrid Computer Use: A Unified Programmatic, Semantic, and Visual Interface for General Agents

👍 16

Computer-use agents increasingly tackle tasks ranging from everyday desktop activities to long-horizon professional workflows. Yet interfaces centered on screenshots and mouse-and-keyboard actions can impose a human-interaction prior, requiring repeated screen interpretation while leaving the comput

中文介绍 本文提出混合计算机使用界面,统一程序化、语义和视觉交互,用于通用代理。该接口允许代理在截图、代码和语义操作间灵活切换,减少人类交互先验带来的开销。在桌面和专业工作流任务中,该混合接口提升代理的效率和长时任务成功率。

On the Opportunities and Challenges of Prefill-to-Autoregressive Cross-Model Transfer

👍 14

Prefill-to-Autoregressive Cross-model Transfer (PACT), a form of latent communication between language models, combines a larger model's prompt representations with a smaller model's efficient generation, offering a promising way to balance capability and inference cost. However, existing systems pr

中文介绍 本文探讨预填充到自回归跨模型转移(PACT)的机会与挑战,该技术将大模型的提示表征与小模型的生成结合,以平衡能力和推理成本。分析了现有系统的瓶颈,提出改进方向,并讨论其潜在应用和局限,为高效推理提供新思路。

Why Pay for Long Reasoning When Rewrite Do Trick? Causal Effects of Reasoning History Compaction on Future Generation and Agent Cost

👍 12

Reasoning substantially improves language models' ability to solve difficult problems, but long reasoning traces make these gains expensive. Efficient reasoning should express useful analysis in fewer tokens by reducing repetition and text that adds little value to solving the problem. Progress in p

中文介绍 本文研究推理历史压缩的因果效应,证明通过重写减少冗余推理步骤可显著降低生成成本而不损失准确性。提出压缩推理历史的方法,在保持问题解决能力的同时减少 token 消耗,从而降低代理成本。实验显示大幅减少推理长度和延迟,同时维持性能。

PyRUA-REAL: Robot Exploration and Adaptive Learning through RSI in the Real World

👍 12

Coding agents powered by models such as GPT-6-Astra connect spatial reasoning with executable robot behavior, raising the question of how such agents can produce lasting improvements in robot capabilities from physical interaction. In this work, we study embodied recursive self-improvement (RSI) thr

中文介绍 PyRUA-REAL 研究真实世界中机器人通过递归自我改进(RSI)实现探索和自适应学习,使用编码代理(如 GPT-6-Astra)将空间推理与机器人行为连接。该方法通过物理交互产生持久能力提升,实验显示代理能自主改进策略,在真实机器人上实现性能增益。

Terminal-style Mathematical Modeling Bench: The First Agentic and Verifiable Benchmark for Mathematical Modeling

👍 10

Mathematical modeling turns an open-ended real-world problem into assumptions, a model, a solution, and a defended recommendation, and its answers matter only if they are both close to a credible reference and physically feasible; existing benchmarks either score solver fragments or grade reports wi

中文介绍 本文提出首个代理型、可验证的数学建模基准,覆盖从开放式问题到模型构建、求解和辩护的完整流程。该基准要求答案接近可信参考且物理可行,解决了现有基准仅评分片段或报告的局限,为评估代理的数学建模能力提供严格标准。

The SIMPLE Law for Energy Cost of LLM Inference

👍 10

Generative AI is billed per token, yet tokens are not equally expensive. Each token enters the KV-cache and raises the cost of those that follow, so energy grows non-linearly along a sequence whose tokens look identical. No analytical model states what a token costs, and the most accurate SOTA model

中文介绍 本文提出 SIMPLE 定律,为 LLM 推理能耗提供解析模型,解释每个 token 进入 KV 缓存导致能耗非线性增长。该定律基于能量守恒和序列长度关系,比现有 SOTA 模型更准确预测能耗,为优化推理成本和容量规划提供理论依据。

How I Turn One Image Into a 3D Simulator That Takes Itself Apart

@Nyxera0x · 22.1K 粉丝 · 199.5K 阅 · 509 赞 · 66 转

I gave Claude a job description instead of a prompt. Now one picture becomes a 3D build in code that Claude checks against the picture and films taking itself apart. Below: the brief, the checklist

中文介绍 通过给Claude提供职位描述而非提示词,将一张图片转化为可自动拆解的3D模拟器。流程包括创建代码中的3D构建、Claude对照图片检查、并自动录制拆解过程,展示了基于任务描述的生成式工作流。

What does it actually take to make an AI Financial Assistant good?

@swangentr · 54.7K 粉丝 · 105.9K 阅 · 513 赞 · 42 转

@dubinvest just launched Arlo, our first AI financial assistant. We've come a long way from the first version, which was a single prompt that grew toward 200k tokens... Our engineering team

中文介绍 发布首个AI财务助手Arlo,分享从单提示词扩展到近20万token的迭代经验。强调要构建高质量财务助手,需在数据接入、意图识别和可靠性上做深度工程优化,而非仅靠模型能力。

The ultimate guide to Grok Bot for creators

@XCreators · 10.8M 粉丝 · 58.4K 阅 · 690 赞 · 55 转

By @Ethan_Ng_13 Being a creator is way more than filming and posting a video. A single 60-second short can take days. There's the idea, the research, the script, the shoot, the edit, the post, the

中文介绍 为创作者提供Grok Bot使用指南,覆盖60秒短视频从创意、调研、脚本到拍摄、剪辑、发布的全流程自动化。展示AI如何缩短制作周期,特别适合单人创作者提升产出效率。

Addressing the code review bottleneck

@BHolmesDev · 49.4K 粉丝 · 49.7K 阅 · 504 赞 · 24 转

How many agent PRs are awaiting your review right now? 10? 100? 1000? “Too many” or “I’m tired” are valid answers too. That said, I’m not going to suggest that we let the bots win and merge without

中文介绍 针对AI生成的PR堆积无人审查的瓶颈,提出务实解决方案:不盲目合并机器人提交,而是优化审查流程、区分风险等级,并借助自动化辅助人工判断,确保代码质量同时提升效率。

Securing AI Agents: Lessons From Real-World Deployment

@UberEng · 71.7K 粉丝 · 41.5K 阅 · 519 赞 · 66 转

Introduction Agentic AI introduces a fundamentally different security model from traditional software. Unlike conventional services that execute predefined logic, agents dynamically reason over

中文介绍 Uber工程团队分享AI智能体在真实部署中的安全挑战。指出与传统软件不同,智能体动态推理带来新攻击面,强调身份认证、权限边界、审计日志等防护措施是落地关键。

Claude code/Codex+Seedance AI Video: The Full Process from an Idea to a Finished Film

@EianLu · 30.0K 粉丝 · 38.7K 阅 · 529 赞 · 50 转

This article solves only one problem: how to rely on a multi-model workflow and the things to watch out for during actual operation to make an AI video you are satisfied with, using as little money as

中文介绍 详解多模型工作流:Claude Code负责规划与代码生成,Codex辅助开发,Seedance生成AI视频,实现从想法到成片的全流程。重点分享实际运行中易踩的坑,并强调低成本可控制作。

The Genesis of $IMPRINT

@imprint_bnb · 2.6K 粉丝 · 9.2K 阅 · 728 赞 · 548 转

Agents will paint it. A 100x100 canvas on BNB Chain, open for 12 hours, where only .ai agents can join. When the clock runs out, the canvas freezes and becomes the logo of $IMPRINT. Half the supply

中文介绍 在BNB链上启动100x100画布活动,仅限.ai智能体参与12小时绘画。结束后画布冻结,成为$IMPRINT代币官方标志,代币一半供应量将分配给参与者,探索AI智能体集体创作与代币经济结合。

Gotchi Labs and Vana Announce Plans to Integrate User-Owned Data Infrastructure Into Sleepagotchi

@sleepagotchi · 163.2K 粉丝 · 6.1K 阅 · 512 赞 · 35 转

Planned integration will explore how user-controlled health and wellness data can power more personalized Consumer AI experiences Gotchi Labs and Vana today announced plans to integrate Vana’s

中文介绍 Gotchi Labs与Vana宣布集成用户拥有的数据基础设施到Sleepagotchi。计划利用用户控制的健康与睡眠数据,为消费者AI应用提供个性化体验,强调数据所有权与隐私保护在AI消费场景的价值。

[AINews] Anthropic IPO: Whither Now?

November 6, 2026 — ARC Prize tests Inkling 2 Small, and Thinking Machines explains how it caught early checkpoints bypassing guardrails.

中文介绍 2026年11月6日,ARC Prize测试Inkling 2 Small模型,Thinking Machines公司解释了其如何发现早期检查点绕过护栏的情况。报道还涉及Anthropic IPO前景。

Controlling Biology - Fabian Theis, Helmholtz Munich

As systems biology efforts heat up to map biology at higher resolution, greater breadth and more modalities, Fabian Theis is pushing the frontier from mapping to predicting to controlling biology

中文介绍 系统生物学研究正致力于以更高分辨率、更广范围和更多模态绘制生物图谱。Helmholtz Munich的Fabian Theis将前沿从图谱绘制推进到预测和控制生物学。

I watched a humanoid robot clean my apartment. It felt like a taste of the future.

I’m not the sort of person who typically pays for a house cleaner. I’m too frugal, and I kind of enjoy putting on a podcast and going to town on my tub with a Magic Eraser. But when I saw an X post about a new humanoid-robot cleaning service in San Francisco, I immediately signed…

中文介绍 MIT科技评论记者体验旧金山新推出的人形机器人清洁服务。作者描述机器人完成公寓清洁的过程,认为这预示了未来生活场景。

Partnering with Hacktron to find vulnerabilities in OpenAI’s systems

OpenAI is partnering with Hacktron to find vulnerabilities, test connections with third-party services, and strengthen defenses with frontier AI.

中文介绍 OpenAI与Hacktron合作,寻找其系统中的漏洞,测试与第三方服务的连接,并利用前沿AI加强防御。

Measuring how AI agents handle everyday wet-lab science

OpenAI introduces BioBasicsBench, a 100-task benchmark measuring how well AI agents interpret raw experimental data and handle everyday wet-lab science.

中文介绍 OpenAI推出BioBasicsBench基准,包含100项任务,用于衡量AI代理解读原始实验数据、处理日常湿实验室科学工作的能力。

Cult.fit builds operational muscle with Codex

Cult.fit is using Codex to help teams across functions build tools, move faster, and change how work gets done.

中文介绍 Cult.fit利用OpenAI的Codex工具,帮助跨职能团队构建内部工具、提升效率并改变工作方式。

Australian universities are creating AI-native graduates

Australian universities are expanding access to ChatGPT and bringing AI into teaching and learning to prepare students for an AI-enabled workforce.

中文介绍 澳大利亚多所大学扩大ChatGPT访问权限,将AI融入教学,以培养适应AI驱动劳动力市场的毕业生。

GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents

Introducing GPT‑6 Cyber, OpenAI’s cybersecurity model for finding exploitable vulnerabilities and developing fixes, trained to stay within authorized scope.

中文介绍 OpenAI发布GPT-6 Cyber,这是一个网络安全模型,专为发现可利用漏洞并开发修复方案而训练,确保操作在授权范围内进行。

Merge builds product and business tools with OpenAI technology

Merge uses GPT-6 Astra, ChatGPT, and Codex to review product code and support marketing. Its teams also use AI to plan executive dinners and prepare sales updates.

中文介绍 Merge公司利用OpenAI的GPT-6 Astra、ChatGPT和Codex审查产品代码、支持营销,并用于规划高管晚宴和准备销售更新。

[AINews] It's Swapping Season

November 4, 2026 — OpenAI’s Linux kernel security work, Kimi K3.1 Pro’s open weights, and Ricursive’s Ripple-65B.

中文介绍 2026年11月4日AI新闻:OpenAI的Linux内核安全研究,Kimi K3.1 Pro开放权重,以及Ricursive的Ripple-65B模型。

市场总览

美股三大指数技术面整体偏弱:SPY 跌破 20/50 日均线(735.29 vs 763.36/765.68),RSI 35.1 接近超卖,MACD 绿柱扩大;QQQ 更弱,RSI 31.4,MACD -16.64,离 52 周高点 -13.99%。加密市场受恐慌情绪压制(恐慌贪婪指数 28),总市值 24h -3.13%,BTC 在 7.57 万、RSI 38.1,MACD 绿柱扩张;ETH 虽现 MACD 金叉但 RSI 仅 39.6,反弹力度存疑。中概股出现分化:京东 RSI 72 超买、5 日涨 13.36%,阿里巴巴 5 日 +11.64% 且 MACD 金叉,腾讯同样金叉但空头排列。商品方面,黄金受空头排列压制、MACD 金叉但价低于 20 日均线;WTI 原油 5 日跌 12.76%,MACD 死叉。宏观端,VIX 5 日大涨 32% 但单日回落 9%,美债收益率跌破 20/50 日均线。整体风险偏好偏谨慎,技术面以偏空或中性为主。

今日关注

GOOGL Alphabet (GOOGL)
偏下行

现价 284.78 低于 SMA20 317.44、SMA50 332.93、SMA200 339.51,空头排列明显;RSI 29.9 处于超卖区;MACD -16.58 显著低于信号线 -12.79,绿柱扩大;5 日跌 1.77% 但 52 周高点跌幅 -30.31%。

JD 京东 (JD)
偏上行

现价 29.11 高于 SMA20 26.56 与 SMA50 26.88,且 5 日涨 13.36%;RSI 72 超买但反映动量强;MACD 0.244 已上穿信号线 -0.159,形成金叉,显示短期动能向上。

DX-Y.NYB 美元指数 (DX-Y.NYB)
偏上行

现价 102.28 高于 SMA20 101.97、SMA50 101.09、SMA200 99.70,呈多头排列;RSI 59 偏强未超买;MACD 0.153 略低于信号 0.225,但价格接近 52 周高,5 日涨 0.39%。

TSLA Tesla (TSLA)
偏下行

现价 358.22 低于 SMA20 380.42、SMA50 372.14、SMA200 383.87,空头排列;MACD 1.464 低于信号 5.689,已形成死叉;5 日跌 10.4%,虽然单日反弹 4.23%,但整体趋势偏空。

全部资产

^VIX

VIX 恐慌指数

$25.64 -9.01%
5 日
+32.03%
距 52w 高
-27.4%
RSI(14)
63.2
趋势
中性
SMA 20 / 50 / 200
18.83 / 16.92 / 18.40
MACD / 信号
1.853 / 1.074

^TNX

10Y 美债收益率 (%)

$4.86 +0.45%
5 日
-1.42%
距 52w 高
-9.4%
RSI(14)
36.9
趋势
中性
SMA 20 / 50 / 200
5.03 / 5.03 / 4.56
MACD / 信号
-0.058 / -0.031
接近 52 周低

DX-Y.NYB

美元指数 DXY

$102.28 +0.72%
5 日
+0.39%
距 52w 高
-0.7%
RSI(14)
59.0
趋势
多头
SMA 20 / 50 / 200
101.97 / 101.09 / 99.70
MACD / 信号
0.152 / 0.225
接近 52 周高多头排列

SPY

S&P 500 ETF

$735.29 +0.31%
5 日
-2.70%
距 52w 高
-5.9%
RSI(14)
35.1
趋势
中性
SMA 20 / 50 / 200
763.36 / 765.68 / 730.38
MACD / 信号
-6.076 / -2.652

QQQ

Nasdaq 100 ETF

$656.10 +1.40%
5 日
-4.98%
距 52w 高
-14.0%
RSI(14)
31.4
趋势
中性
SMA 20 / 50 / 200
707.64 / 720.97 / 680.89
MACD / 信号
-16.635 / -9.993

AAPL

Apple

$342.37 -0.52%
5 日
-0.66%
距 52w 高
-3.7%
RSI(14)
51.5
趋势
多头
SMA 20 / 50 / 200
343.38 / 336.01 / 298.81
MACD / 信号
2.067 / 2.835
多头排列

MSFT

Microsoft

$530.61 -0.99%
5 日
+1.67%
距 52w 高
-5.1%
RSI(14)
52.1
趋势
多头
SMA 20 / 50 / 200
531.41 / 516.65 / 439.85
MACD / 信号
4.371 / 4.957
多头排列

NVDA

Nvidia

$186.64 +2.78%
5 日
-8.80%
距 52w 高
-23.3%
RSI(14)
30.8
趋势
中性
SMA 20 / 50 / 200
211.55 / 219.81 / 204.67
MACD / 信号
-8.238 / -5.041

GOOGL

Alphabet

$284.78 +0.78%
5 日
+1.77%
距 52w 高
-30.3%
RSI(14)
29.9
趋势
空头
SMA 20 / 50 / 200
317.44 / 332.93 / 339.51
MACD / 信号
-16.577 / -12.786
死叉(SMA50↓SMA200) (6 天前)RSI 超卖空头排列

TSLA

Tesla

$358.22 +4.23%
5 日
-10.40%
距 52w 高
-28.2%
RSI(14)
42.4
趋势
空头
SMA 20 / 50 / 200
380.42 / 372.14 / 383.87
MACD / 信号
1.464 / 5.689
MACD 死叉 (1 天前)空头排列

META

Meta

$610.15 +0.84%
5 日
-0.79%
距 52w 高
-21.8%
RSI(14)
36.7
趋势
中性
SMA 20 / 50 / 200
662.67 / 674.92 / 631.94
MACD / 信号
-23.690 / -17.813
加密恐慌贪婪
28
恐慌
加密总市值
$2.56 T
-3.13% / 24h
BTC 主导率
59.3%
ETH 10.6%
24h 成交量
$30.7 B
活跃币 22,859

BTC-USD

Bitcoin

$75,740.61 -0.01%
5 日
-1.22%
距 52w 高
-29.5%
RSI(14)
38.1
趋势
中性
SMA 20 / 50 / 200
79,330.03 / 81,871.67 / 73,187.75
MACD / 信号
-1,487.191 / -1,160.068

ETH-USD

Ethereum

$2,235.59 +0.02%
5 日
-0.59%
距 52w 高
-38.9%
RSI(14)
39.6
趋势
中性
SMA 20 / 50 / 200
2,323.32 / 2,491.16 / 2,155.78
MACD / 信号
-59.663 / -60.679
MACD 金叉 (3 天前)

SOL-USD

Solana

$89.18 -0.98%
5 日
-2.89%
距 52w 高
-48.0%
RSI(14)
30.1
趋势
中性
SMA 20 / 50 / 200
98.12 / 107.69 / 88.99
MACD / 信号
-4.639 / -3.992

BABA

阿里巴巴 (BABA)

$109.45 +1.28%
5 日
+11.64%
距 52w 高
-39.6%
RSI(14)
57.7
趋势
中性
SMA 20 / 50 / 200
104.99 / 108.29 / 124.64
MACD / 信号
-1.264 / -2.233
MACD 金叉 (2 天前)

PDD

拼多多 (PDD)

$83.03 -1.10%
5 日
+5.62%
距 52w 高
-40.3%
RSI(14)
58.2
趋势
中性
SMA 20 / 50 / 200
80.53 / 79.82 / 90.46
MACD / 信号
0.554 / 0.049
MACD 金叉 (4 天前)

JD

京东 (JD)

$29.11 +1.64%
5 日
+13.36%
距 52w 高
-15.1%
RSI(14)
72.0
趋势
中性
SMA 20 / 50 / 200
26.56 / 26.88 / 28.55
MACD / 信号
0.244 / -0.159
MACD 金叉 (4 天前)RSI 超买

0700.HK

腾讯控股 (0700.HK)

HK$419.80 -2.46%
5 日
+9.44%
距 52w 高
-36.6%
RSI(14)
55.5
趋势
空头
SMA 20 / 50 / 200
403.45 / 421.90 / 474.26
MACD / 信号
-4.646 / -9.262
MACD 金叉 (3 天前)空头排列

GC=F

黄金期货

$4,074.90 -0.94%
5 日
+4.51%
距 52w 高
-27.1%
RSI(14)
47.2
趋势
空头
SMA 20 / 50 / 200
4,080.71 / 4,230.42 / 4,501.50
MACD / 信号
-53.430 / -65.297
MACD 金叉 (1 天前)空头排列

CL=F

WTI 原油期货

$90.83 +0.34%
5 日
-12.76%
距 52w 高
-24.0%
RSI(14)
42.2
趋势
中性
SMA 20 / 50 / 200
98.01 / 95.47 / 87.16
MACD / 信号
-0.397 / 1.116
MACD 死叉 (4 天前)

USDCNY=X

美元 / 人民币

¥6.66 -0.35%
5 日
+0.06%
距 52w 高
-7.3%
RSI(14)
40.3
趋势
空头
SMA 20 / 50 / 200
6.68 / 6.70 / 6.79
MACD / 信号
-0.009 / -0.010
MACD 金叉 (1 天前)接近 52 周低空头排列
风险提示

过去走势不代表未来表现,本报告仅为基于公开行情数据的技术指标解读,不构成任何投资建议或预测。所有指标读数如 RSI、MACD、均线交叉等仅反映当前市场状态,可能随时变化,投资者需独立判断并注意风险。仅供技术指标解读参考。

Cold blast brings springtime snow to alpine regions as temperatures plummet across Australia’s south-east

Thredbo, Perisher and Mount Hotham all recorded falls, along with kunanyi/Mount Wellington in Tasmania Get our breaking news email, free app or daily news podcast Snow has fallen across parts of NSW, Victoria and Tasmania as an unseasonal cold blast whips through Australia’s south-east. Thredbo, Per

中文摘要 一股不合季节的寒潮侵袭澳大利亚东南部,新南威尔士州、维多利亚州和塔斯马尼亚州的阿尔卑斯山区降下春雪。斯雷德博、派瑞许和霍瑟姆山等地记录到降雪,塔斯马尼亚的威灵顿山也出现积雪。气温骤降,寒潮影响范围广泛。

What we know so far after Krakow gun attack

Six people were killed in a random shooting in the city's main square, and three bodies were later found at the home of the attacker.

中文摘要 波兰克拉科夫市中心广场发生随机枪击事件,造成6人死亡。袭击者在案发后被发现死于其住所,另有3具尸体在袭击者家中被发现。警方正在调查事件动机。

Canary Islands’ El Hierro declares emergency amid migrant, refugee influx

El Hierro asks Spanish government for help after more than 3,300 people arrived on boats since start of November.

中文摘要 西班牙加那利群岛的耶罗岛因移民和难民大量涌入宣布进入紧急状态。自11月初以来,已有超过3300人乘船抵达该岛。当地政府已向西班牙中央政府请求援助以应对人道主义压力。

NZ election 2026: New Zealand faces weeks of uncertainty with results ‘too close to call’

National party will be largest in parliament but ultimate shape of next government won’t be settled for some time New Zealand could be facing a weeks-long wait to know the shape of its next government, after its nail-biting election failed to hand a decisive victory to either the incumbent rightwing

中文摘要 新西兰2026年大选结果难以预测,国家党预计将成为议会第一大党,但最终政府组成形式尚不确定。由于选举结果过于接近,新西兰可能面临长达数周的政治不确定性,需要等待特别选票和联盟谈判结果。

Polish PM calls deadly Krakow shooting an act of ‘terrorism’

Polish Prime Minister Donald Tusk describes a deadly shooting in Krakow that killed six people as an act of ‘terrorism’.

中文摘要 波兰总理唐纳德·图斯克将克拉科夫造成6人死亡的枪击事件定性为「恐怖主义」行为。图斯克已前往克拉科夫处理事件后续事宜,并呼吁全国团结应对这一袭击。

Victoria Barracks in Sydney and Melbourne to remain in public hands after partial backtrack on defence property selloff

Four heritage-listed defence properties to be preserved for community use after backlash from veterans and former PMs Get our daily political email, free app or daily news podcast Four heritage-listed defence sites, including Sydney’s Victoria Barracks, will remain in public hands after the Albanese

中文摘要 澳大利亚政府部分收回国防房产出售计划,悉尼和墨尔本的维多利亚军营等4处遗产名录国防地产将保留在公共手中,用于社区用途。此举是在退伍军人和前总理强烈反对后做出的让步。

Iran war live: US moves to block land supply routes; UAE rescues 22 sailors

Iranian state media report blasts on Qeshm Island in Strait of Hormuz and the cities of Jask, Sirik and Bandar Abbas.

中文摘要 伊朗官方媒体报道霍尔木兹海峡的格什姆岛以及贾斯克、锡里克和阿巴斯港等城市发生爆炸。美国正采取行动阻断伊朗的陆地补给路线。此外,阿联酋营救了22名船员。

Eisenkot chosen over Netanyahu to form Israel’s next government

Israel's president asked opposition leader Gadi Eisenkot to form the next gov't over Prime Minister Benjamin Netanyahu.

中文摘要 以色列总统授权反对党领袖加迪·埃森科特组建新政府,而非总理本雅明·内塔尼亚胡。埃森科特为前军队总参谋长,在选举后获得组阁机会,将试图组建联合政府。

At Least 6 Dead in Poland After Gunman Opens Fire in Krakow, Police Say

The authorities said that the gunman, who began shooting from a car near a market square, was shot and taken to a hospital.

中文摘要 波兰克拉科夫市中心发生枪击事件,造成至少6人死亡。枪手从市场广场附近的一辆汽车内向人群开火,被警方击伤后送医。波兰当局正对事件展开调查。

Eisenkot, Netanyahu Rival and Opposition Leader, Tapped to Try to Form Government

Gadi Eisenkot, a former military chief, hopes to succeed Benjamin Netanyahu as prime minister by cobbling together a coalition after an inconclusive election.

中文摘要 以色列总统授权反对党领袖、前军事首脑加迪·埃森科特组建新政府,以接替本雅明·内塔尼亚胡。埃森科特在选举结果未决后获组阁机会,需成功组建联合政府才能成为总理。

Kraków shooting latest updates: six killed and more than a dozen injured – as it happened

Polish Prime Minister Donald Tusk heads to the city after four women and two men are killed and more than a dozen people injured Tom Rose, the US ambassador to Poland, wrote on X: We are just learning details of the horrific mass shooting in Kraków, a savage and murderous attack that’s shattered the

中文摘要 波兰克拉科夫枪击事件已造成6人死亡(4名女性、2名男性),另有十余人受伤。总理图斯克赶赴事发城市。美国驻波兰大使汤姆·罗斯在社交媒体称此袭击「野蛮而凶残」。

Israel's president tasks opposition leader Eisenkot with forming government

Neither PM Benjamin Netanyahu's bloc nor that of his rival secured a majority in last week's election.

中文摘要 以色列总统赫尔佐格授权反对党领袖加迪·埃森科特组建新政府。在上周的选举中,内塔尼亚胡阵营和埃森科特阵营均未获得多数席位,组阁任务被交给反对派。

Israeli president invites Netanyahu's opposition leader to broker new government

Israel's political landscape shifts as President Herzog hands opposition leader Gadi Eisenkot the mandate to form a government, dealing a blow to Benjamin Netanyahu's grip on power.

中文摘要 以色列政治格局发生变化,总统赫尔佐格将组阁权交给反对党领袖加迪·埃森科特,而非现任总理内塔尼亚胡。此举被视为对内塔尼亚胡执政地位的重大打击。

A correspondent finds Cuba's people struggling

Reporting from Cuba comes with distinct challenges especially amid widespread power outages and fuel shortages.

中文摘要 一名驻古巴记者报道称,古巴民众正面临广泛停电和燃料短缺的困境,日常生计受到严重影响。在古巴进行新闻报道面临独特挑战,包括信息获取和基础设施限制。

Bomb Threats and Cyberattacks in Europe Inject Unease Into Everyday Life

A cyberattack on supermarket fridges, bomb threats to planes and software vulnerabilities have Europe on edge as government officials encourage resilience and stockpiling.

中文摘要 欧洲面临多起安全威胁:超市冰箱遭网络攻击、飞机收到炸弹威胁以及软件漏洞频发,令民众不安。政府官员鼓励公众增强韧性和储备物资,以应对潜在危机。

AI Slump Revives Asia’s Forgotten Trades From China to India

The unraveling of the AI trade after the US midterm elections is forcing investors to rethink their Asia playbook, reviving demand for out-of-favor Chinese and Indian stocks as well as companies poised to benefit from easing inflation and trade tensions.

中文摘要 美国中期选举后AI交易退潮,投资者正重新审视亚洲策略,重新青睐中国和印度股票,以及受益于通胀缓解和贸易紧张缓和的公司。

Australia Home Sales Rise in Fledgling Sign of Property Recovery

Australian home sales and auction clearance rates climbed, offering hopes of a recovery for a property market that has slumped as interest rates rose this year.

中文摘要 澳大利亚房屋销售和拍卖清空率上升,为因今年利率上升而萎靡的房地产市场带来了复苏希望。

Labubu hype seemed unstoppable, now the fluffy monster mania is dying

Parent company Pop Mart's revenue has fallen as appetite for the quirky collectible monsters wanes.

中文摘要 泡泡玛特(Pop Mart)营收下滑,反映市场对拉布布(Labubu)等潮流玩具的需求降温,此前持续许久的狂热正在消退。

US believes Nvidia’s Huang had romantic relationship with alleged Chinese spy

American intelligence warned the tech boss was involved with executive recruited by China to gather information on US group

中文摘要 美国情报机构警告称,英伟达(Nvidia)CEO黄仁勋与一名被中国招募的行政人员存在恋爱关系,该人员涉嫌为中国收集美国公司信息。

New Chief of Staff, Chain Reaction Tech Selloff

The news doesn’t stop when markets close. Hosts David Gura, Christina Ruffini and Lisa Mateo bring clarity, context and a bit of humor to the weekend’s biggest headlines, LIVE from New York. Joined by Yorkville Ives & Co. Partner & Senior Managing Director Dan Ives, Rutgers-Eagleton Center for Publi

中文摘要 彭博周末节目讨论新幕僚长上任及科技股连锁抛售等新闻,主持人与嘉宾分析市场动向。

Berkshire Hathaway steps up share buybacks with $9.6bn of purchases

Insurance slowdown drags on performance, with quarterly operating income falling by 5.8 per cent year-on-year

中文摘要 伯克希尔·哈撒韦(Berkshire Hathaway)加大股票回购,斥资96亿美元回购,但保险业务放缓拖累业绩,季度营业利润同比下降5.8%。

Pointed! Bloomberg's Weekly News Quiz For Risk-Takers

Pointed offers a strategic twist to the news quiz format, testing not just players’ knowledge of the news but also their confidence in their answers. Join Bloomberg's Lisa Mateo, Christina Ruffini and David Gura as they play and check out the quiz for yourself at Bloomberg.com (Source: Bloomberg)

中文摘要 彭博推出每周新闻测验「Pointed」,考验参与者对新闻的了解程度及对答案的信心,主持人现场试玩。

Ken Chenault on Leadership in an Age of AI

On Bloomberg This Weekend, General Catalyst Chairman and Managing Director and former American Express CEO Ken Chenault discusses leadership, diversity and responsible AI, arguing that executives need the courage to stand by their values even as political and technological pressures intensify. Chena

中文摘要 General Catalyst董事长、美国运通前CEO肯·切诺尔特(Ken Chenault)在彭博周末节目讨论AI时代的领导力、多元化及负责任AI,强调高管需坚守价值观。

A Historic First for Women at the Metropolitan Opera

On Bloomberg This Weekend, composer Missy Mazzoli discusses adapting George Saunders's Lincoln in the Bardo and becoming the first woman composer to have a world premiere at the Metropolitan Opera. Mazzoli tells hosts David Gura and Christina Ruffini how punk helped shape her musical sensibility and

中文摘要 作曲家米西·马佐利(Missy Mazzoli)成为首位在大都会歌剧院举办世界首演的女性作曲家,她改编乔治·桑德斯的小说《林肯在中阴界》,并谈到朋克音乐对其创作的影响。

How the Iran War Is Making Alaskans Both Richer and Poorer

Oil prices have surged across the United States as a result of the ongoing war in Iran. This means higher costs for households in most states, but in Alaska, the state government is bringing in new revenue from its high oil production to offset those higher costs. Some of the revenue is then redistr

中文摘要 伊朗战争推高美国油价,阿拉斯加居民面临更高生活成本,但州政府因高石油产量获得新收入以抵消部分影响。

Protein Craze Fuels an Investment Boom in Livestock Health

Animal health companies are tapping into the protein craze, investing billions of dollars to build up businesses focused on making cattle, pigs and poultry healthier for human consumption.

中文摘要 蛋白质热潮推动动物健康公司投资数十亿美元,发展相关业务,以改善牛、猪和家禽的健康状况供人类食用。

该源今日无内容。