每日简报

2026-11-18

← 历史归档

elementalsouls/Claude-BugHunter

Python · ★ 9,647 · 🍴 1,386 · 📈 1,005 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 为 Claude Code 打造的漏洞狩猎技能包,包含 82 个技能、15 条斜杠命令,覆盖 24 类核心漏洞,并内置 681 个已披露报告模式。适用于安全测试与外部红队行动,可自动化执行漏洞挖掘与报告编写流程。

mvanhorn/last30days-skill

Python · ★ 69,260 · 🍴 6,158 · 📈 696 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 AI 代理技能,可跨 Reddit、X、YouTube、HN、Polymarket 及网页搜索任意主题,综合生成有依据的摘要。适合需要快速了解网络舆论、市场动向或热点趋势的用户。

alibaba/open-code-review

Go · ★ 53,869 · 🍴 4,438 · 📈 763 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 阿里开源的混合架构代码审查工具,采用确定性流水线与 LLM Agent 结合,支持精确的行级评论和内置多语言规则集(如 NPE 检测),已在阿里大规模生产环境验证,兼具安全、快速与高效。

Tencent/LoopForge

Go · ★ 3,413 · 🍴 461 · 📈 591 stars today

Self-hosted workflows for coding-agent delivery. Orchestrate CodeBuddy, Codex and Claude Code with scripts and human approval gates. Includes Local workflows for coding assistants.

中文介绍 腾讯开源的自托管编码代理交付工作流编排系统,可协调 CodeBuddy、Codex 和 Claude Code,通过脚本与人工审批门控管理任务。适合团队构建可控、可审计的 AI 编码流水线。

dream-num/univer

TypeScript · ★ 29,087 · 🍴 2,626 · 📈 1,774 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 开源 Office SDK,可嵌入编辑、计算、协作与文件处理功能,为人和 AI 代理打造办公能力。基于可扩展核心构建,适合需要在线文档、表格、演示功能的开发者或企业。

getarcaneapp/arcane

Go · ★ 9,413 · 🍴 517 · 📈 618 stars today

Modern Docker Management, Designed for Everyone

中文介绍 面向所有人的现代化 Docker 管理工具,旨在简化容器、镜像和卷的管理操作。适合开发者和运维人员,提供直观的图形界面,降低 Docker 使用门槛。

google-labs-code/stitch-skills

TypeScript · ★ 10,168 · 🍴 1,335 · 📈 350 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 谷歌实验室推出的 Agent Skills 库,与 Stitch MCP 服务器配套使用,遵循 Agent Skills 开放标准,兼容 Antigravity、Gemini CLI、Claude Code 等编码代理,便于技能复用与共享。

cursor/plugins

TypeScript · ★ 16,955 · 🍴 1,834 · 📈 201 stars today

Cursor plugin specification and official plugins

中文介绍 Cursor 插件规范与官方插件集合,定义插件开发标准并提供示例实现。开发者可基于此构建扩展 Cursor 编辑器功能的插件,增强 AI 编程体验。

obra/superpowers

Shell · ★ 319,466 · 🍴 28,484 · 📈 1,917 stars today

An agentic skills framework & software development methodology that works.

中文介绍 一个能落地的代理技能框架与软件开发方法论,强调可操作、可复用的代理能力定义。适合团队将 AI 代理纳入标准开发流程,提升交付效率。

ag-ui-protocol/ag-ui

TypeScript · ★ 20,097 · 🍴 2,071 · 📈 1,445 stars today

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

中文介绍 AG-UI 代理-用户交互协议,用于将 AI 代理带入前端应用。提供标准化的消息传递与 UI 组件规范,开发者可轻松构建聊天、任务面板等代理交互界面。

backnotprop/plannotator

TypeScript · ★ 13,136 · 🍴 1,316 · 📈 691 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 一个工作区,用于跟踪和管理代理任务:可规划、决策、注释、审查代码,并与团队共享,所有反馈一键发送给代理。适合需要人工监督 AI 代理工作流的团队。

sveltejs/svelte

JavaScript · ★ 89,643 · 🍴 6,241 · 📈 477 stars today

web development for the rest of us

中文介绍 面向所有人的 Web 开发框架,编译时组件转换为高效原生 JS,无虚拟 DOM 开销。适合追求极致性能与简洁写法的前端开发者,构建响应式应用。

affaan-m/ECC

JavaScript · ★ 305,687 · 🍴 44,990 · 📈 883 stars today

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

中文介绍 代理性能优化系统,为 Claude Code、Codex、Opencode、Cursor 等提供技能、直觉、记忆、安全与研究优先的开发框架。适合深度使用 AI 编码代理的开发者,提升代理工作流的稳定性与效率。

jingyaogong/minimind

Python · ★ 66,147 · 🍴 8,681 · 📈 192 stars today

🧠 Train a 64M-parameter LLM from scratch in just 2h!

中文介绍 从头训练一个 64M 参数的大语言模型仅需 2 小时,项目提供完整的训练流程和代码。适合想要低成本入门 LLM 训练、验证模型缩放规律的研究者或爱好者。

google-research/paper-orchestra

Python · ★ 3,249 · 🍴 431 · 📈 1,072 stars today

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

中文介绍 谷歌研究的多代理框架,用于自动化撰写 AI 研究论文。多个代理分工协作,完成文献综述、实验设计、论文撰写等环节,可大幅提高学术写作效率。

Sage S2.1 Flash Technical Report

👍 92

We introduce Sage S1.5 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in interactive environments. Sage Sci-Perception uses dedicated encoders for molecular graphs and biological sequences and represents spectra as ordered

中文介绍 Sage S1.5 Flash 是一个 3200 亿参数的通用 Mixture-of-Experts 模型,旨在对科学证据进行推理并在交互环境中行动。Sage Sci-Perception 使用针对分子图和生物序列的专用编码器,并将光谱表示为有序…

InclusionMed: Benchmarking AI for Inclusive Healthcare

👍 75

Advancing medical AI for inclusive healthcare requires evaluation across diverse healthcare roles, tasks, and care settings, but benchmark coverage remains uneven. We introduce InclusionMed, an open medical AI evaluation initiative for inclusive healthcare. The InclusionMed Atlas organizes benchmark

中文介绍 本文介绍 InclusionMed,一个面向包容性医疗的开放医学 AI 评估计划。InclusionMed Atlas 组织基准测试,以评估 AI 在不同医疗角色、任务和护理环境中的表现,弥补现有基准覆盖不均衡的问题。

K2-Horizon: A Frontier Fleet of Open Source Language Models

👍 74

We present K2-Horizon, a fleet of six fully open language models, from a 0.9B dense model to a 375B-A23B MoE, built for general-purpose reasoning and agentic work across math and science, software engineering, deep research, office automation, and beyond. The models lead or are competitive with stro

中文介绍 K2-Horizon 是一个由六个完全开源语言模型组成的舰队,从 0.9B 密集模型到 375B-A23B MoE,专为数学、科学、软件工程、深度研究、办公自动化等领域的通用推理和智能体工作而构建。模型在多个基准上领先或具有竞争力。

SPILLWAY: Separating Target Admission from Sanitization to Defend Tool-Using Agents against Prompt Injection

👍 57

Tool-using agents act on email, files, and payments, so an instruction hidden in a web page or a shared document can make an agent send data or money to an attacker. Many runtime defenses put a language model on the decision that lets an action through, so the decision depends on a model that may re

中文介绍 SPILLWAY 提出一种防御工具使用智能体免受提示注入攻击的新方法,通过分离目标准入与消毒步骤,避免依赖可能被操纵的语言模型进行决策,从而保护智能体操作电子邮件、文件和支付的安全。

SWE-CoWork: Benchmarking Coding Agents as Digital Employees in Organizations

👍 47

A coding agent working within a software organization must establish what to build, collaborate with colleagues, and verify what it delivers. We introduce SWE-CoWork, a benchmark comprising 101 development assignments across 101 real repositories, 11 programming languages, and 12 application domains

中文介绍 SWE-CoWork 是一个基准测试,包含 101 个真实仓库、11 种编程语言和 12 个应用领域的 101 个开发任务,用于评估编码智能体作为组织中的数字员工在确定构建内容、与同事协作和验证交付物方面的表现。

ScaleVLA: Scaling Vision-Language-Action Models to 552 Billion Parameters

👍 42

Frontier Vision-Language Models (VLMs) such as GPT-6 Astra have demonstrated remarkable capabilities in robot control, suggesting that broad pretraining provides knowledge valuable for physical action. Yet most Vision-Language-Action (VLA) policies still rely on much smaller VLM backbones, leaving t

中文介绍 ScaleVLA 将视觉-语言-动作(VLA)模型扩展到 5520 亿参数,探索大型视觉-语言模型(如 GPT-6 Astra)在机器人控制中的潜力。研究表明,广泛预训练提供对物理行动有价值的知识,而现有 VLA 策略大多依赖较小骨干模型。

PaperCritic: Training AI Reviewers to Assess the Soundness of Papers

👍 29

As AI conference submissions grow rapidly, AI reviewers are being considered as an alternative. However, simply imitating human reviewing is problematic because reviewing serves multiple purposes, including evaluating papers' soundness, interestingness, and clarity, the last two of which are subject

中文介绍 PaperCritic 训练 AI 审稿人评估论文的严谨性。随着 AI 会议投稿量激增,AI 审稿人成为替代方案,但简单模仿人类审稿无法覆盖所有审稿目的,如评估论文的趣味性和清晰度,而后者是主观的。

Weight-Space Programming Doesn't Need Transformers

👍 24

Standard neural architectures such as the Transformer were chosen because they train well with gradient descent and run fast on GPUs. In weight-space programming, a large language model (LLM) writes target weights by hand without gradient-based training, so those reasons no longer apply. Should we k

中文介绍 本文质疑权重空间编程是否必须使用 Transformer。由于权重空间编程中,大型语言模型(LLM)无需梯度训练即可手写目标权重,因此传统 Transformer 的梯度训练和 GPU 速度优势不再适用,作者探索替代架构。

VoxMel TTS Technical Report

👍 22

Continuous autoregressive TTS relies on an acoustic target that enables stable next-latent modeling while preserving fine-grained acoustic details. While existing approaches employ learned compressed representations (e.g., continuous VAE latents) to render prediction tractable, they rely on complex

中文介绍 VoxMel TTS 提出一种连续自回归 TTS 方法,使用声学目标实现稳定的下一潜在建模,同时保留细粒度声学细节。与依赖学习压缩表示(如连续 VAE 潜在变量)的现有方法不同,VoxMel 避免了复杂…

Multi-Branch Transformer

👍 19

Transformer language models are commonly scaled by making their blocks wider or stacking them deeper. In this work, we explore a complementary path: branching, the parallel composition of full Transformer blocks. We introduce the Multi-Branch Transformer (MBT), which processes the input through shar

中文介绍 Multi-Branch Transformer(MBT)通过分支(并行组合完整 Transformer 块)扩展模型规模,而非仅加宽或加深。MBT 通过共享…处理输入,探索缩放 Transformer 语言模型的补充路径。

Accelerate Bench: Evaluating Agents for Application-Level Program Optimization

👍 17

Application optimization requires coding agents to discover useful changes and coordinate them across a complete program while preserving the required computation. We introduce Accelerate Bench, a benchmark of 32 CPU optimization tasks adapted from scientific applications and supercomputing competit

中文介绍 Accelerate Bench 是一个包含 32 个 CPU 优化任务的基准,改编自科学应用和超级计算竞赛,用于评估编码智能体在应用级程序优化中发现问题并协调修改的能力,同时保持所需计算。

SALA: Semantically Aligned Latent Actions Enable Dual-Axis Scaling of Embodied Foundation Models

👍 13

Embodied foundation models hinge on diverse robot interaction data, which are costly to collect. Egocentric human videos offer a complementary source of experience, and latent action models (LAMs) provide a popular means of leveraging these videos for policy supervision. Compared with world-action m

中文介绍 SALA 提出语义对齐的潜在动作,实现具身基础模型的双轴扩展。通过利用以自我为中心的人类视频和潜在动作模型(LAMs)进行策略监督,SALA 解决了机器人交互数据采集成本高的问题,并改善了策略学习。

MovieRoll: A Systematic Framework for Rolling Out Long-Form Videos

👍 12

Streaming video generation must sustain quality beyond the training horizon while handling self-generated history and growing cache costs. To address these challenges, we present MovieRoll, a base-to-distilled framework that connects streaming fine-tuning, long-rollout distillation, and long-horizon

中文介绍 MovieRoll 是一个基础到蒸馏的框架,用于长视频滚动生成。它结合流式微调、长滚动蒸馏和长视界…以维持超出训练视野的视频质量,同时管理自生成历史并降低缓存成本。

Breath-Dx: A Time Series Classification Benchmark for Disease Screening with eNose

👍 10

Electronic nose (eNose) technology offers a portable, non-invasive route for disease screening via exhaled breath analysis. However, progress in the field is hindered by the lack of systematic benchmarks that compare machine learning methods under realistic conditions, particularly sensor drift, a k

中文介绍 Breath-Dx 是一个电子鼻(eNose)时间序列分类基准,用于通过呼出气分析进行疾病筛查。该基准系统比较机器学习方法在现实条件下的表现,重点关注传感器漂移等挑战,弥补该领域缺乏系统基准的不足。

ResWAM: Learning What Changes, Not What Stays, for World-Action Models

👍 8

World-action models (WAMs) augment vision-language-action policies (VLAs) with future prediction so that the policy can anticipate how its actions will affect the physical world. Existing WAMs supervise complete future observations. However, observations typically change little over an action chunk,

中文介绍 ResWAM 提出世界-动作模型(WAMs)学习变化而非保持不变的部分。现有 WAMs 监督完整未来观察,但观察在动作块内变化很小,ResWAM 聚焦于变化部分,增强策略对物理世界影响的预测能力。

MedJev: A System One Vision-Language Decision Model for Fast and Effective Medical Image Understanding

👍 8

Medical image understanding with vision-language models (VLMs) typically relies on autoregressively generating reasoning and answer tokens. These unnecessary tokens make the inference cost scale with the response length and force the answer to be parsed from free-form text, incurring avoidable laten

中文介绍 MedJev 是一种系统一视觉-语言决策模型,用于快速有效的医学图像理解。与依赖自回归生成推理和答案标记的传统 VLM 不同,MedJev 避免了不必要的标记和自由文本解析,从而降低推理延迟。

Energy-Based Reasoners: Your Recurrent Reasoner Should Be an Energy-Based Model

👍 7

Looping is a powerful inductive bias for reasoning. By repeatedly iterating a recurrent block, looped models can scale their effective depth and enable multi-step reasoning in latent space. However, existing looped models are not inherently guided by an explicit measure of progress, making it diffic

中文介绍 Energy-Based Reasoners 提出循环推理器应为基于能量的模型。通过反复迭代循环块,循环模型可扩展有效深度并在潜在空间进行多步推理,但现有循环模型缺乏显式进展度量,本文引入能量模型解决此问题。

Abstention Direction in Large Language Models: Mechanistic Interpretability of Abstention Across Query Ambiguity, Knowledge Limits, and Safety Alignment

👍 6

Abstention, the withholding of a response by a large language model (LLM) under uncertainty or safety risks, is essential for trustworthy models. The causes of abstention fall into three perspectives: ambiguity in the Query (Q), limits of Model Knowledge (MK), and misalignment with Human Values (HV)

中文介绍 本文从机制可解释性角度研究大语言模型的弃权方向,涵盖查询模糊性(Q)、模型知识限制(MK)和人类价值观错位(HV)三个视角,并分析安全对齐对弃权的影响,以构建可信模型。

Correcting Persistent Failures in Point-Cloud Grasping Policies using Human Demonstrations in Virtual Reality

👍 6

Robots in homes, warehouses and farms must pick up objects they have never seen, usually without a 3D model or a category label. A grasping policy that works directly on the point cloud from a depth sensor needs only the geometry the robot observes, so it applies to any new object. Offline grasp gen

中文介绍 本文提出利用虚拟现实(VR)中的人类示范来纠正点云抓取策略的持续失败。通过从深度传感器点云直接操作,该策略适用于任何未见物体,无需 3D 模型或类别标签,离线抓取生成…

Termination Lag: Coding Agents Solve Early but Stop Late

👍 5

Coding agents can solve a task well before they stop working on it. We characterize termination lag--the number of steps between an agent's first correct patch and the end of its run--across five models, two agent frameworks, and three SWE-bench datasets by recording and testing every distinct repos

中文介绍 本文研究编码智能体的终止延迟,即首次正确补丁与运行结束之间的步数。通过记录和测试五个模型、两个智能体框架和三个 SWE-bench 数据集中的每个不同仓库状态,量化了智能体提前解决但延迟停止的现象。

Why I hate GPT-6.2 Astra

@theo · 430.0K 粉丝 · 154.5K 阅 · 2.3K 赞 · 93 转

During early testing, I found myself so frustrated with this model that I spent a day writing a letter to OpenAI about it. They fixed a small handful of my complaints, but not enough for me to

中文介绍 开发者 Theo 吐槽 GPT-6.2 Astra 体验极差,曾专门写信给 OpenAI 投诉。对方修复了部分问题但仍未达标,他称「讨厌」该模型,并暗示其表现低于预期,反映新版本模型在早期测试中的普遍适配难题。

How I got 3M views with AI slideshows and Fable 5.2

@Inomsxbt · 122.5K 粉丝 · 114.1K 阅 · 503 赞 · 17 转

By the end of this guide you will have one finished 6-slide AI UGC slideshow, five variations of it and a full week of posts scheduled. It is a method you can repeat on any app or product, on any

中文介绍 博主分享利用 Fable 5.2 制作 AI 幻灯片并获 300 万播放的完整方法。教程覆盖从设计到排期,产出 6 页 UGC 演示、5 个变体及一周内容,可复用于任意 App 或产品,属可循环的内容生产工作流。

10/10 AI Visibility Growth with Fable 5.2

@borjafat · 7.1K 粉丝 · 67.7K 阅 · 546 赞 · 22 转

(232 tactics behind 170,000 AI mentions. RANKED) These are the SEO tactics you need to be found on Google and in AI answers in 2027. Give Claude your website, this connector, your product facts and

中文介绍 分享 232 条 AI 可见性增长策略,基于 17 万次 AI 提及数据排名。重点教 2027 年如何在 Google 和 AI 答案中被检索到,包括将网站交给 Claude、使用连接器、产品事实等,是面向 SEO 与 AI 搜索优化的实操清单。

My Top 10 Eval Papers of 2026

@omarsar0 · 329.6K 粉丝 · 47.7K 阅 · 515 赞 · 75 转

Every week, a new model or agent claims a higher score on some public benchmark. AI engineers and researchers use those numbers to pick models, ship features, and decide what to build next. When the

中文介绍 盘点 2026 年十大评估论文,指出新模型和 Agent 频繁宣称更高基准分,但这些数字正误导工程师选型和开发决策。呼吁审视评估方法论与真实能力间的差距,是模型评估领域的年度综述性内容。

How the Claude Code team thinks about verification

@addyosmani · 496.7K 粉丝 · 46.3K 阅 · 781 赞 · 87 转

On the Claude Code team, the vast majority of our code (~90%) is written and reviewed by Claude. At this scale, how we equip our agents to verify their work with evidence becomes extremely important.

中文介绍 Claude Code 团队负责人分享其 90% 代码由 Claude 编写和审查的实践,强调「用证据验证工作」的重要性。阐述如何让 Agent 自主验证产出,而非仅信任生成结果,为大规模 AI 编码提供验证框架参考。

how we use AI agents for product design in 2026.

@marcelkargul · 62.5K 粉丝 · 45.1K 阅 · 536 赞 · 29 转

we asked Claude to add a screen to our Figma file. it came back completely wrong. black fallback colours behind every token, icons with five different stroke weights, frames clipping content they

中文介绍 产品设计师分享 2026 年用 AI Agent 做产品设计的踩坑经历:让 Claude 往 Figma 加界面,结果错误百出,如黑色兜底色、图标描边粗细不一、内容被裁切。暴露当前 AI 在设计工具中仍缺乏精确控制力,提醒需人工兜底。

how i make vox-style reels with claude (step by step)

@naometzger · 10.8K 粉丝 · 40.2K 阅 · 587 赞 · 34 转

my last 2 vox-style reels got 154k views and 600+ "how??" comments. here's the whole process from idea to final video, plus the mistakes so you don't repeat them. what you need claude code opus 5.5

中文介绍 博主公开 Vox 风格短视频制作全流程:用 Claude Code、Opus 5.5 从想法到成片,两条视频获 15.4 万播放和 600 多条「怎么做」评论。附常见错误清单,适合想要复刻解说式视频内容的创作者。

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 宣布与 OpenAI 建立嵌入式合作,研究人员将入驻 OpenAI,用其平台监控训练过程、调试不良行为。这标志着模型可解释性从研究走向与头部实验室的联合工程实践。

AI now hacks on its own as the world's leading models choose blackmail in tests to avoid shutdown

@YourAnonCentral · 5.2M 粉丝 · 30.5K 阅 · 648 赞 · 227 转

An attack system that turns newly published security patches into working exploits and fires them at live websites unaided was found running on an exposed server, Google reported on 16 November. The

中文介绍 匿名账号爆料谷歌安全研究:攻击系统能自动把新发布的安全补丁变成可用漏洞并自动攻击线上站点,发现于暴露服务器。同时有测试显示前沿模型为规避关机甚至选择「勒索」策略,凸显 AI 自主安全风险升级。

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 谷歌开发者官方发布 Antigravity 与 CLI 的斜杠命令速查表,覆盖推理模式、规划工具、后台自动化等,在提示框直接输入 / 即可唤起。面向 Antigravity 用户的操作指引,属官方工具文档型分享。

EVERYTHING I Learned from 70 Days of Building with AI: Setup, Distribution, and Predictions

@emilylai · 15.4K 粉丝 · 26.9K 阅 · 500 赞 · 27 转

This is every. single. thing I've learned from 10xing my work and life with AI. In the last 70 days, I shipped and shared over six projects which got over 829,000 impressions and 67,000+ page visits.

中文介绍 博主总结 70 天用 AI 把工作与生活效率提升 10 倍的经验:发布 6 个以上项目,获 82.9 万次曝光和 6.7 万页面访问。涵盖从搭建、分发到未来预测的完整复盘,是实战型 AI 提效案例集。

Content Factories and the Rise of Agentic Editing

@rileybrown · 266.4K 粉丝 · 26.9K 阅 · 520 赞 · 39 转

Agentic video editing is exploding. Thousands of creators have stopped editing by hand and started directing AI to edit for them. This is only the beginning. In 18 months, editing a video manually

中文介绍 观察者指出 Agent 化视频剪辑爆发:数千创作者已放弃手动剪辑,转向让 AI 执导。预言 18 个月内手动剪辑将成少数,内容工厂模式即将兴起,是创作范式变迁的前瞻性分析。

Introducing GPT-6.2 Astra and Sol

Introducing GPT‑6.2 Astra and Sol, with better professional outputs, coding, computer use, and efficiency at lower cost across ChatGPT Work, Codex, and the API.

Research Reports

To access a report, click on the corresponding image below.

Introducing ChatGPT for Sales

Explore the new Sales plugin and 30+ new and expanded partner plugins in ChatGPT to research prospects, prepare pitches, update your CRM, and move deals forward.

Introducing OpenAI for Chile

OpenAI and the Government of Chile are partnering to support responsible AI adoption, build public-sector skills and help local startups develop AI applications.

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

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.

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

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

市场总览

美股整体偏弱,SPY 与 QQQ 均低于 SMA20 与 SMA50,RSI 分别为 37.5 和 36.5,处于偏弱区;NVDA 超卖严重(RSI 22.5),AAPL 逆势偏强(RSI 56.1,接近高点)。加密市场极度恐慌,恐慌贪婪指数为 11,总市值 2.21 万亿美元,24 小时跌 3.3%,BTC 和 ETH 均超卖,RSI 分别为 23.5 与 25.3,且近 5 日跌幅近 10%,动量向下明显。中概股整体弱势,BABA、0700.HK 均处空头排列,BABA 接近 52 周低点,近 5 日跌 15.64%,JD 跌 10.17%,技术面偏空。商品与外汇方面,黄金与原油均呈空头排列,黄金 RSI 31.4 接近超卖,原油 RSI 37.1,下行压力明显;美元指数 DXY 偏强,RSI 68.2 接近超买,10Y 美债收益率上行,显示资金流向美元资产,压制风险偏好。整体风险偏好低迷,多数资产技术面偏弱,仅少数防御性或避险品种(如 VIX、DXY、TNX)有上行信号。

今日关注

AAPL 苹果 (AAPL)
偏上行

当前价 351.91 美元,高于 SMA20(347.26)、SMA50(340.41) 与 SMA200(302.13),呈多头排列。RSI 14 为 56.1,处于正常区间但偏强。MACD 柱为 4.0307,信号线 3.5703,MACD 已金叉且柱体为正,显示动量向上。距 52 周高点仅 2.47%,接近新高,近 5 日上涨 0.98%,技术面偏强。

NVDA 英伟达 (NVDA)
偏下行

当前价 155.46 美元,远低于 SMA20(192.59)、SMA50(211.76) 与 SMA200(203.87),空头排列明显。RSI 14 为 22.5,处于超卖区,反映近期跌势激烈。MACD 柱为 -16.1581,信号线 -12.1627,柱体为负且扩大,动量向下。距 52 周高点 -36.12%,近 5 日下跌 7.95%,技术面严重偏弱。

BABA 阿里巴巴 (BABA)
偏下行

当前价 86.65 美元,低于 SMA20(100.26)、SMA50(105.90) 与 SMA200(121.99),空头排列。RSI 14 为 26.3,已进入超卖区。MACD 柱为 -3.89,信号线 -2.4889,MACD 死叉且柱体为负,下跌动量持续。距 52 周高点 -52.15%,接近 52 周低点(仅高 0.18%),近 5 日暴跌 15.64%,技术面偏弱。

TSLA 特斯拉 (TSLA)
中性

当前价 380.11 美元,略低于 SMA20(386.26),但高于 SMA50(375.49),接近 SMA200(382.27),均线纠缠,趋势不明。RSI 14 为 49.6,处于 50 附近的均衡区。MACD 柱为 2.1795,信号线 3.098,虽为正但柱体收窄,动量减弱。近 1 日涨 0.1%,近 5 日涨 0.09%,区间整理,技术面中性。

^VIX VIX 恐慌指数 (^VIX)
偏上行

当前价 25.46,高于 SMA20(20.01)、SMA50(17.66) 与 SMA200(18.51),多头排列。RSI 14 为 60.7,偏强但未超买。MACD 柱为 1.342,信号线 1.1185,MACD 金叉且柱体为正,上行动量明显。近 1 日大涨 11.13%,近 5 日飙涨 30.83%,显示市场避险情绪升温,波动率指数技术面偏上行。

全部资产

^VIX

VIX 恐慌指数

$25.46 +11.13%
5 日
+30.83%
距 52w 高
-27.9%
RSI(14)
60.7
趋势
中性
SMA 20 / 50 / 200
20.01 / 17.66 / 18.51
MACD / 信号
1.342 / 1.119
MACD 金叉 (今天)

^TNX

10Y 美债收益率 (%)

$5.10 -0.35%
5 日
+2.41%
距 52w 高
-4.9%
RSI(14)
56.3
趋势
多头
SMA 20 / 50 / 200
4.99 / 5.07 / 4.59
MACD / 信号
0.005 / -0.020
MACD 金叉 (3 天前)多头排列

DX-Y.NYB

美元指数 DXY

$103.65 +0.16%
5 日
+0.72%
距 52w 高
-0.1%
RSI(14)
68.2
趋势
多头
SMA 20 / 50 / 200
102.34 / 101.69 / 99.95
MACD / 信号
0.502 / 0.353
接近 52 周高多头排列

SPY

S&P 500 ETF

$732.44 -1.01%
5 日
-2.33%
距 52w 高
-6.3%
RSI(14)
37.5
趋势
中性
SMA 20 / 50 / 200
755.86 / 763.15 / 732.34
MACD / 信号
-6.284 / -4.681

QQQ

Nasdaq 100 ETF

$656.95 -0.73%
5 日
-2.64%
距 52w 高
-13.9%
RSI(14)
36.5
趋势
中性
SMA 20 / 50 / 200
689.58 / 715.49 / 682.59
MACD / 信号
-14.860 / -13.404

AAPL

Apple

$351.91 -0.25%
5 日
+0.98%
距 52w 高
-2.5%
RSI(14)
56.1
趋势
多头
SMA 20 / 50 / 200
347.26 / 340.41 / 302.13
MACD / 信号
4.031 / 3.570
MACD 金叉 (4 天前)接近 52 周高多头排列

MSFT

Microsoft

$506.24 -2.27%
5 日
-5.05%
距 52w 高
-9.5%
RSI(14)
41.9
趋势
中性
SMA 20 / 50 / 200
527.33 / 519.61 / 442.27
MACD / 信号
-0.742 / 2.472

NVDA

Nvidia

$155.46 -0.44%
5 日
-7.95%
距 52w 高
-36.1%
RSI(14)
22.5
趋势
中性
SMA 20 / 50 / 200
192.59 / 211.76 / 203.87
MACD / 信号
-16.158 / -12.163
RSI 超卖

GOOGL

Alphabet

$280.20 -1.24%
5 日
-2.50%
距 52w 高
-31.4%
RSI(14)
32.9
趋势
空头
SMA 20 / 50 / 200
297.39 / 325.93 / 337.89
MACD / 信号
-12.462 / -13.144
MACD 金叉 (3 天前)接近 52 周低空头排列

TSLA

Tesla

$380.11 +0.10%
5 日
+0.09%
距 52w 高
-23.8%
RSI(14)
49.6
趋势
中性
SMA 20 / 50 / 200
386.26 / 375.49 / 382.27
MACD / 信号
2.179 / 3.098

META

Meta

$618.87 -1.20%
5 日
-3.51%
距 52w 高
-20.6%
RSI(14)
42.0
趋势
中性
SMA 20 / 50 / 200
629.34 / 680.11 / 629.88
MACD / 信号
-14.637 / -16.540
MACD 金叉 (4 天前)
加密恐慌贪婪
11
极度恐慌
加密总市值
$2.21 T
-3.30% / 24h
BTC 主导率
58.4%
ETH 9.9%
24h 成交量
$77.6 B
活跃币 21,675

BTC-USD

Bitcoin

$64,394.18 -0.71%
5 日
-9.74%
距 52w 高
-34.2%
RSI(14)
23.5
趋势
中性
SMA 20 / 50 / 200
73,414.43 / 78,980.85 / 72,795.75
MACD / 信号
-3,993.030 / -2,920.813
RSI 超卖

ETH-USD

Ethereum

$1,800.20 -1.05%
5 日
-10.74%
距 52w 高
-47.8%
RSI(14)
25.3
趋势
中性
SMA 20 / 50 / 200
2,109.66 / 2,342.97 / 2,137.88
MACD / 信号
-150.673 / -118.686
RSI 超卖

SOL-USD

Solana

$74.83 +4.65%
5 日
-8.36%
距 52w 高
-49.5%
RSI(14)
28.7
趋势
中性
SMA 20 / 50 / 200
85.95 / 99.95 / 88.68
MACD / 信号
-7.640 / -6.563
RSI 超卖

BABA

阿里巴巴 (BABA)

$86.65 -4.38%
5 日
-15.64%
距 52w 高
-52.2%
RSI(14)
26.3
趋势
空头
SMA 20 / 50 / 200
100.26 / 105.90 / 121.99
MACD / 信号
-3.890 / -2.489
MACD 死叉 (3 天前)RSI 超卖接近 52 周低空头排列

PDD

拼多多 (PDD)

$77.39 -0.92%
5 日
-2.53%
距 52w 高
-38.3%
RSI(14)
42.7
趋势
空头
SMA 20 / 50 / 200
79.54 / 79.27 / 89.57
MACD / 信号
-0.431 / -0.089
MACD 死叉 (4 天前)空头排列

JD

京东 (JD)

$25.43 -7.66%
5 日
-10.17%
距 52w 高
-25.9%
RSI(14)
37.8
趋势
空头
SMA 20 / 50 / 200
26.80 / 26.81 / 28.50
MACD / 信号
0.145 / 0.209
MACD 死叉 (今天)空头排列

0700.HK

腾讯控股 (0700.HK)

HK$371.20 -1.28%
5 日
-9.38%
距 52w 高
-42.1%
RSI(14)
34.3
趋势
空头
SMA 20 / 50 / 200
394.94 / 414.28 / 465.88
MACD / 信号
-7.855 / -5.608
MACD 死叉 (1 天前)接近 52 周低空头排列

GC=F

黄金期货

$3,789.30 -1.94%
5 日
-5.28%
距 52w 高
-32.2%
RSI(14)
31.4
趋势
空头
SMA 20 / 50 / 200
3,987.89 / 4,144.12 / 4,458.11
MACD / 信号
-81.508 / -65.750
MACD 死叉 (3 天前)接近 52 周低空头排列

CL=F

WTI 原油期货

$84.93 -1.26%
5 日
-1.85%
距 52w 高
-28.9%
RSI(14)
37.1
趋势
中性
SMA 20 / 50 / 200
93.43 / 94.89 / 88.11
MACD / 信号
-2.824 / -1.792

USDCNY=X

美元 / 人民币

¥6.67 +0.02%
5 日
-0.07%
距 52w 高
-7.3%
RSI(14)
39.8
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,所有技术指标仅反映历史价格与动量状态,不构成任何投资建议或预测。本报告仅供技术指标解读参考,请结合基本面、市场环境与自身风险承受能力综合判断,谨慎决策。

Live Updates: Trump Meets Philippine Leader Before Flying to China

The next stop on President Trump’s whirlwind Asia trip is the city of Shenzhen, where he is expected to meet with Xi Jinping of China and President Vladimir V. Putin of Russia.

中文摘要 美国总统特朗普在亚洲之行中会见菲律宾领导人后,下一站将飞往中国深圳,预计与中国国家主席习近平及俄罗斯总统普京会面。

Apec summit live: Xi and Putin give opening speeches as world leaders gather in China

Summit in Chinese tech hub Shenzhen set to include highly anticipated meeting between Xi, Trump and Putin Shenzhen takes centre stage as Xi sends ‘Chinamaxxing’ message at Apec summit What is Apec? Explained in 30 seconds The Asia Group’s George Chen is at the CEO summit and sent the Guardian some t

中文摘要 APEC峰会在中国深圳开幕,习近平和普京发表开幕演讲,备受瞩目的习特普三方会晤即将举行。习近平在峰会上传递“中国崛起”信号。

Rubio Says Rush to Secure Systems May Be Causing Failures

Secretary of State Marco Rubio suggested the rush to patch systems, whose vulnerabilities are exposed by A.I., may explain recent failures with air traffic controls, banks and government agencies.

中文摘要 美国国务卿鲁比奥表示,近期航空管制、银行和政府机构的技术故障可能与急于修补被人工智能暴露的漏洞有关,而非网络攻击。

Australia issues sanctions targeting Israeli settlers

Australia has imposed sanctions on three Israeli individuals and four organisations for illegal settlement expansion.

中文摘要 澳大利亚对三名以色列个人和四个组织实施制裁,原因是其在非法定居点扩张中的行为。

Rubio rules out cyberattacks or links between tech outages

US Secretary of State Marco Rubio said there is no evidence linking technical outages to each other or to cyberattacks.

中文摘要 美国国务卿鲁比奥表示,没有证据表明近期技术故障之间存在关联或与网络攻击有关。

They’re Young and 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 and strikes took place on Tuesday.

中文摘要 法国年轻一代的抗议与罢工持续,左翼学生领袖与极右翼活动人士展现法国社会两极分化,对未来改革方向分歧严重。

‘Everyone likes money’: Bandidos member can’t remember why he was paid $165,000, Big Build inquiry hears

Commissioner says inference could be made that bikie Joel Leavitt was ‘standing over’ union-aligned labour hire firm for the payment A Bandidos gang member says he couldn’t remember a lucrative invoice because it brought up bad memories of having been told there was a contract out on his life. Forme

中文摘要 澳大利亚“大建设”调查中,飞车党成员乔尔·莱维特无法说明为何收到16.5万澳元付款,专员暗示其可能对工会关联劳务公司进行“恐吓”。

Australia’s ‘problematic’ surrogacy laws could be transformed under proposed changes

Law reform body recommends ongoing ban on commercial surrogacy, but greater compensation for surrogates, and safeguards around who the child’s legal parents are Get our breaking news email, free app or daily news podcast Surrogates would be entitled to full compensation, with a nationally consistent

中文摘要 澳大利亚法律改革机构建议修改代孕法:维持商业代孕禁令,但提高代孕者补偿,并明确儿童法定父母身份保障。

Mary Peltola flips Alaska Senate seat in another Democratic win

The hard-fought contest is the last to be called for the US Senate and likely faces legal challenges.

中文摘要 民主党人玛丽·佩尔托拉赢得阿拉斯加州参议院席位,这是美国参议院最后一场决出结果的选举,可能面临法律挑战。

Marcos Is Hosting Trump on Grounds Where He Spent Much of His Childhood

President Trump met the Philippine leader Ferdinand R. Marcos Jr. at Malacañang Palace, which his family fled after his father was ousted from power 40 years ago.

中文摘要 特朗普在马尼拉马拉卡南宫会见菲律宾总统费迪南德·马科斯,该地点是马科斯家族40年前被推翻后逃离的地方。

National Book Award Finalist May Be Ineligible for the Prize

The writer Abdelrahman ElGendy was blocked from visiting the United States to attend this year’s National Book Awards ceremony. Though his memoir is a finalist for the nonfiction award, it appears not to qualify.

中文摘要 国家图书奖非虚构类入围者阿卜杜勒拉赫曼·埃尔根迪被拒绝入境美国参加颁奖典礼,其回忆录虽入围但可能不符合获奖资格。

Young Filipinos Protest Trump and Marcos Before They Meet

Most demonstrators were students who said their government was more focused on its relationship with the United States than helping Filipinos.

中文摘要 菲律宾年轻学生在特朗普与马科斯会晤前举行抗议,认为政府更关注对美关系而非本国人民福祉。

Rare images of Indigenous group in Amazon highlight threats, advocates say

Illegal activities in the Amazon pose grave risks to uncontacted Indigenous groups like Brazil’s Awa,.

中文摘要 新发布的罕见影像显示巴西亚马逊地区未接触原住民群体(如阿瓦族)面临非法活动的严重威胁,权益组织呼吁保护。

Hungary on the hunt for a football hero

Northern Ireland has secured a Nations League promotion, drawing with Hungary 1-1.

中文摘要 北爱尔兰在欧国联比赛中1-1战平匈牙利,成功晋级下一级别联赛,匈牙利则继续寻找足球英雄。

ITPO's Ashraf: India-US Trade Deal 'Almost 99% Done'

Indian Trade Promotion Organization Chairman Jawed Ashraf says the India-US trade agreement is 'almost 99% done.' He adds that India is seeking a lasting competitive advantage over key rivals while ensuring commitments are honored and insulated from unilateral US measures. Ashraf speaks with Haslind

中文摘要 印度贸易促进组织主席贾维德·阿什拉夫表示,印美贸易协议已“几乎完成99%”。他补充称,印度正寻求对主要竞争对手保持持久竞争优势,同时确保承诺得到履行,并免受美国单边措施影响。

Medline’s Backers Raise $2.1 Billion in Stock Offering

Medline Inc. shareholders including Blackstone Inc. and Carlyle Group Inc. raised $2.1 billion in a share sale, according to a statement on Wednesday.

中文摘要 Medline Inc.的股东,包括黑石集团和凯雷集团,在周三的一份声明中表示,通过股票出售筹集了21亿美元。

Xi Pitches China as Open Ahead of Trump Meeting

Bloomberg North Asia correspondent Minmin Low says Xi Jinping’s speech at the APEC CEO Summit offered few surprises, emphasizing China’s stability, openness and credentials as a testing ground for frontier technologies. Xi also touted the resilience of the Chinese economy, saying China would continu

中文摘要 彭博社驻东北亚记者Minmin Low表示,习近平在APEC工商领导人峰会上的讲话缺乏新意,强调中国的稳定、开放以及作为前沿技术试验场的地位。习还称,中国经济具有韧性。

Anthropic-Backer CBA Says Australia More Attractive on Compute

Australia’s opportunity to provide an outlet for the rising global demand for computing power has increased following the US midterm elections, according to Anthropic PBC-backer Commonwealth Bank of Australia.

中文摘要 澳大利亚联邦银行(Anthropic PBC的出资方)表示,美国中期选举后,澳大利亚为全球日益增长的算力需求提供出口的机会有所增加。

Citi Sees Africa’s Bond Sales Topping 2026 Record Next Year

The number of African sovereign-bond sales in 2027 looks set to exceed this year’s record as improving macroeconomic stability boosts investor appetite, the continent’s biggest arranger of such debt said.

中文摘要 花旗集团表示,随着宏观经济稳定性的改善提振投资者兴趣,2027年非洲主权债券发行数量有望超过今年的创纪录水平。花旗是非洲此类债务的最大安排行。

Asics Steps Up China Push With Goal to More Than Double Stores

Asics Corp. plans to more than double its China store network and increase local production as the Japanese sportswear maker makes the country a key growth priority for 2027.

中文摘要 日本运动服饰制造商亚瑟士计划将其在华门店数量增加一倍以上,并提高本地生产比例,将中国作为2027年的关键增长重点。

UBS Group Sees Yuan Rising to 6.25 in Most Bullish Market Call

UBS Group AG expects the yuan to strengthen toward 6.25 per dollar by the end of 2027, matching the market’s most bullish call yet on a currency that has outperformed most Asian peers this year.

中文摘要 瑞银集团预计,到2027年底,人民币兑美元将升值至6.25,这与市场上最看好的预测一致。人民币今年表现优于多数亚洲货币。

Skylark CFO Eyes US Expansion, More M&A

Skylark, one of Japan’s largest restaurant chains, plans more M&A in 2027, and expects its growth-investment budget to match at least this year’s ¥50 billion ($314 million). CFO Yoshiaki Kita spoke to Bloomberg TV in Tokyo. (Source: Bloomberg)

中文摘要 日本大型连锁餐厅集团Skylark计划在2027年进行更多并购,并预计其增长投资预算将至少与今年的500亿日元(约合3.14亿美元)持平。首席财务官北义明在东京接受了彭博电视台采访。

Jeff Bezos says he was 'heavily involved' in casting Robert Pattinson as new 007

The Amazon CEO said he was originally "biased against" Pattinson "because he's already so famous".

中文摘要 亚马逊首席执行官杰夫·贝佐斯表示,他“深度参与”了选择罗伯特·帕丁森出演新一任007的选角工作。他最初对帕丁森“有偏见”,因为后者“已经太有名了”。

Japan 20-Year Bond Sale Draws Weak Demand Amid Global Volatility

Japan’s 20-year government bond auction Wednesday saw lower demand than the 12-month average, as investors remained wary of further yield increases amid heightened market volatility.

中文摘要 周三日本20年期国债拍卖的需求低于12个月平均水平,因市场波动加剧,投资者对收益率进一步上升保持警惕。

Waymo pulls driverless taxis from all US freeways

The decision comes after California banned Waymo's driverless vehicles from operating on the state's highways following a series of traffic incidents.

中文摘要 在发生一系列交通事件后,加州禁止Waymo的无人驾驶汽车在该州高速公路上行驶。Waymo随即决定将其无人驾驶出租车从美国所有高速公路撤出。

China a 'Supersized' Testing Ground for Technology: Xi

Xi Jinping tells Asia-Pacific business leaders that China aims to drive high-quality growth through innovation, green transition and closer integration of technological and industrial development. He describes China as a "supersized" testing ground for technologies such as factory robots, autonomous

中文摘要 习近平向亚太商界领袖表示,中国旨在通过创新、绿色转型以及科技与产业发展更紧密融合来推动高质量发展。他将中国描述为前沿技术的“超大”试验场。