每日简报

2026-09-24

← 历史归档

anthropics/financial-services

Python · ★ 37,180 · 🍴 5,405 · 📈 664 stars today

中文介绍 Anthropic 面向金融服务业发布的开源方案集合,围绕 Claude 在银行、保险、证券等场景的落地:合规审查、客户对话、风控分析等。面向希望在金融业务中引入大模型的工程与产品团队,提供可复用的 Agent、工具链与示例代码,缩短从 PoC 到生产的路径。

google/ax

Go · ★ 9,619 · 🍴 473 · 📈 1,543 stars today

Google's open agentic orchestration runtime

中文介绍 Google 开源的 Agentic 编排运行时 AX,为多 Agent 系统提供统一的调度、状态管理与工具调用基础设施。支持复杂任务流、跨模型协作与可观测性,适合需要构建生产级 Agent 平台的工程团队,解决自研编排胶水代码成本高、扩展困难的问题。

davila7/claude-code-templates

Python · ★ 31,674 · 🍴 3,596 · 📈 389 stars today

CLI tool for configuring and monitoring Claude Code

中文介绍 面向 Claude Code 的 CLI 配置与监控工具,通过预置模板快速初始化项目、管理子命令与钩子,并提供运行状态观测。适合重度使用 Claude Code 的开发者,用来把 AI 编码工作流标准化、可复用,减少每次上手的环境搭建成本。

BuilderIO/agent-native

TypeScript · ★ 6,732 · 🍴 607 · 📈 87 stars today

A framework for building agentic apps

中文介绍 Builder.io 推出的 Agent-Native 应用开发框架,围绕大模型 Agent 构建用户交互、工具调用与状态持久化。面向希望把现有 Web 应用改造成 Agent 可驱动的前端团队,强调以 Agent 为一等公民的组件与渲染模型,适合构建对话式、任务式产品。

obra/superpowers

Shell · ★ 290,953 · 🍴 26,032 · 📈 474 stars today

An agentic skills framework & software development methodology that works.

中文介绍 obra 开源的 Agentic Skills 框架与软件开发方法论,把编码、调试、重构等技能封装为 Agent 可复用的能力单元。面向使用 Claude Code、Cursor 等 AI 编程助手的开发者,通过系统化工作流提升 Agent 的产出质量与可维护性。

dream-num/univer

TypeScript · ★ 16,714 · 🍴 1,488 · 📈 1,142 stars today

The Office Harness for AI Agents — Spreadsheets, Docs, Slides, Canvas, Relational Tables, and PDF in one runtime.

中文介绍 面向 AI Agent 的开源办公运行时 Univer,在单一运行时中提供电子表格、文档、幻灯片、画布、关系表与 PDF 能力。让 Agent 可以直接读取、生成和协作处理办公文件,适合构建 Copilot 类办公产品、自动化报告与数据分析工具。

Open-Dev-Society/OpenStock

TypeScript · ★ 19,081 · 🍴 2,315 · 📈 344 stars today

OpenStock is an open-source alternative to expensive market platforms. Track real-time prices, set personalized alerts, and explore detailed company insights — built openly, for everyone, forever free.

中文介绍 OpenStock 是开源的行情平台替代品,提供实时股价、个性化告警与公司深度信息。面向开发者、研究者与小团队,作为替代付费金融数据源的低成本方案,可自托管集成到量化策略、看盘工具或教学项目中。

agent-substrate/substrate

Go · ★ 3,671 · 🍴 425 · 📈 558 stars today

Agent Substrate: the core system

中文介绍 Agent Substrate 的核心系统,为多 Agent 架构提供统一的执行底座,包括能力注册、消息路由、记忆与权限等基础组件。面向需要搭建通用 Agent 平台的团队,避免每个项目重复造轮子,提升跨场景 Agent 的复用与治理效率。

strands-agents/harness-sdk

Python · ★ 8,044 · 🍴 1,216 · 📈 115 stars today

Build an agent harness and control it end-to-end. Open-source SDK for production AI agents in Python & TypeScript - any model, any cloud.

中文介绍 Strands 推出的 Agent Harness SDK,支持在 Python 与 TypeScript 中端到端构建与控制生产级 AI Agent,兼容任意模型与云厂商。适合企业工程团队封装 Agent 运行时、集成工具链并接入业务系统,是构建可观测、可扩展 Agent 服务的基础设施。

HKUDS/CLI-Anything

Python · ★ 50,123 · 🍴 4,609 · 📈 57 stars today

"CLI-Anything: Making ALL Software Agent-Native" -- CLI-Hub: https://clianything.cc/

中文介绍 香港大学 HKUDS 的 CLI-Anything 项目,目标是让所有软件都具备 Agent-Native 的命令行能力,把现有应用通过 CLI 封装为 Agent 可调用的工具。面向希望扩展 LLM 工具生态的开发者与社区,配合 CLI-Hub 提供标准化的工具分发与调用协议。

superdesigndev/treg

Python · ★ 2,911 · 🍴 249 · 📈 506 stars today

OpenRouter for agent tools. Join community here: https://discord.gg/6mQYYfFMAn

中文介绍 treg 定位为 Agent 工具版本的 OpenRouter,为 LLM Agent 提供跨厂商、跨实现的工具路由与发现服务。开发者无需关心具体后端,按标准协议调用搜索、代码、API 等工具,适合快速搭建具备多工具能力的 Agent 应用与平台。

pbakaus/impeccable

JavaScript · ★ 70,594 · 🍴 4,277 · 📈 304 stars today

The design language that makes your AI harness better at design.

中文介绍 pbakaus 发布的 Impeccable 设计语言包,为 AI Agent 的产出提供一致、规范的 UI 视觉与组件风格。当 Agent 生成前端界面时,能遵循该设计语言输出更专业、易读的结果,适合构建面向终端用户的 AI 生成应用与 Copilot 产品。

mvt-project/mvt

Python · ★ 14,584 · 🍴 1,388 · 📈 543 stars today

MVT (Mobile Verification Toolkit) helps with conducting forensics of mobile devices in order to find signs of a potential compromise.

中文介绍 Mobile Verification Toolkit 是开源移动设备取证工具集,用于检测设备是否被植入间谍软件、越狱或异常权限滥用。面向安全研究人员、取证团队与企业 MDM 管理员,支持 iOS 与 Android,帮助识别潜在的被入侵或篡改迹象。

DeusData/codebase-memory-mcp

C · ★ 44,773 · 🍴 3,649 · 📈 190 stars today

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

中文介绍 Codebase-Memory MCP 是一个高性能代码智能 MCP 服务端,将代码仓库索引为持久知识图谱,支持 158 种语言、毫秒级查询。让 AI 编程助手以远低于直接读代码的 Token 成本理解仓库结构,适合大型代码库的智能问答与重构。

harry7557558/spirula-studio

C++ · ★ 871 · 🍴 61 · 📈 69 stars today

Cross-vendor 3D Gaussian Splatting trainer - video to splat to mesh, Vulkan or CUDA.

中文介绍 Spirula Studio 是跨厂商的 3D Gaussian Splatting 训练工具,可从视频直接生成 Splat 场景并导出 Mesh,支持 Vulkan 与 CUDA 后端。面向三维重建、数字孪生与 XR 内容创作者,提供比传统 photogrammetry 更灵活的轻量渲染管线。

browser-use/video-use

Python · ★ 26,728 · 🍴 3,179 · 📈 746 stars today

Edit videos with coding agents

中文介绍 Video-Use 让编码 Agent 通过编程方式编辑视频,把剪切、转场、字幕、配音等操作抽象为 Agent 可调用的动作。适合内容创作者与自动化团队,用自然语言描述即可完成原本需要专业剪辑软件的视频生成与修改。

TNT-Likely/PanWatch

Python · ★ 1,745 · 🍴 321 · 📈 95 stars today

盯盘侠 PanWatch · 自托管 AI 盯盘助手,集成 TradingAgents 多 Agent 投资决策 | A股/港股/美股实时监控、持仓管理、智能分析、全渠道推送

中文介绍 盯盘侠 PanWatch 是自托管的 AI 盯盘助手,集成 TradingAgents 多 Agent 决策框架,覆盖 A 股、港股、美股的实时监控、持仓管理、智能分析与全渠道推送。面向个人投资者与小团队,用开源方案搭建个性化的 AI 投研工作台。

The Tasteful Agent: Measuring and Improving Taste in Long-Horizon Tasks

👍 116

LLM agents increasingly work on long-horizon tasks, and the decisions they make along the way, such as which hypothesis to test or which implementation to build on, determine the outcome of the whole run. Making these decisions well is becoming a key capability for both engineering and research agen

中文介绍 针对长程 LLM Agent 的决策质量研究「品味」(taste)。提出方法用于衡量并改善 Agent 在长时程任务中的判断与选择,提升其推理与实现能力的上限。

HappyWorld-Bench

👍 36

Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, interaction, and modification. We introduce HappyWorld-Bench, a comprehensive benchmark that evaluates whether generated worlds remain reliable as agent

中文介绍 提出 HappyWorld-Bench,用于评测世界模型在生成、探索、交互与修改下的质量、一致性与响应性,考察生成世界在智能体操作下是否仍然可靠。

Ovis-Embedding: Pushing the Frontiers of Universal Omni-Modal Embeddings

👍 34

In this report, we introduce Ovis-Embedding, a state-of-the-art omni-modal embedding family built on native integration of text, image, video, and audio. Instead of assembling separate modality towers, Ovis-Embedding uses a shared multimodal backbone to encode different modalities in a common repres

中文介绍 介绍 Ovis-Embedding,一个统一文本、图像、视频与音频的全模态 embedding 家族,采用共享多模态 backbone 将各模态编码到统一表示空间。

The Past Frames the Future: Memory for Autoregressive Video Generation

👍 33

Advances in generative models have improved video fidelity, enabling long-horizon generation, interactive world modeling, and evolving visual environments. Autoregressive (AR) video generation extends visual sequences through causal rollouts. However, a fundamental bottleneck emerges: as the generat

中文介绍 研究自回归视频生成中的长期一致性瓶颈。提出显式记忆机制(The Past Frames the Future),在因果展开过程中保留历史帧信息,以提升长程视频生成的质量与一致性。

Bellman Policy Optimization

👍 27

Reinforcement learning with verifiable rewards (RLVR) improves the reasoning capabilities of large language models (LLMs). We introduce Bellman Policy Optimization (BPO), a critic-free method derived from Policy Mirror Descent (PMD). For autoregressive generation with terminal rewards, BPO uses the

中文介绍 提出 Bellman Policy Optimization(BPO),一种源自 Policy Mirror Descent 的无 critic 方法,用于在可验证奖励下提升 LLM 的推理能力,支持终端奖励的自回归生成。

Circuit Hypernetworks for Quantum-Augmented Diffusion Language Models

👍 26

Language models can be adapted by changing the computations applied to individual tokens. Quantum circuits offer one such approach, but evaluating wider circuits inside a large model can be computationally demanding. Here we introduce HyperQ, which adds token-conditioned quantum residual branches to

中文介绍 介绍 HyperQ,将 token 条件的量子残差分支嵌入扩散语言模型中,通过 circuit hypernetwork 提升模型可定制性,同时降低在大规模模型内部评估较宽量子电路的计算开销。

Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World

👍 23

Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning in dynamic environments requires VLMs to perceive local state transitions caused by object motion and viewpoint changes and integrate them over long trajectories to maintain an upd

中文介绍 提出 Spatial-Interactor,让 VLM 通过与可观察物理世界的交互学习空间推理,感知物体运动与视角变化引起的局部状态转换,并在长轨迹中整合信息以维持更新。

SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue

👍 23

Long-term conversational memory in multi-party settings requires more than retrieving relevant content from long-term conversations: it must distinguish who said what, whom each statement concerns, how individuals perceive one another, what information is shared by the group, and how states change o

中文介绍 提出 SpeakerMem-R1,一种以发言人为中心的双轨记忆机制,用于多方对话中的长期记忆,区分「谁说了什么」「针对谁」「群体共享信息」等维度,提升对话理解。

JEV-as-a-Judge: Accept When Confident, Escalate When Unsure

👍 23

LLM-as-a-judge enables evaluation across diverse tasks, but inference cost and confidence reliability become critical at scale. We study whether a decision-only judge can provide an economical first pass and identify when stronger evaluation is needed. Comparing jev-as-a-judge with sixteen generativ

中文介绍 研究 JEV-as-a-Judge 策略,使 LLM 判官在置信时接受、不确定时升级到更强模型,以降低大规模评估的推理成本,并与多种判官范式比较。

StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training

👍 23

Vector Quantization (VQ) is fundamental to discrete visual tokenizers that power modern autoregressive and masked image generation models. While recent shared-projection codebook methods have substantially advanced codebook utilization, training stability remains a critical and underexplored challen

中文介绍 介绍 StableVQ,为向量量化 tokenizer 的稳定训练提供实用指导,重点解决共享投影 codebook 方法下常见但少被研究的不稳定性问题。

From Pattern Recognizers to Personalized Companions: A Survey of Large Language Models in Mental Health

👍 16

The rising global prevalence of mental health conditions, together with longstanding barriers in traditional healthcare, such as limited resources, high cost, stigma, and privacy concerns, has created an urgent need for accessible and scalable support. Large Language Models (LLMs) have emerged as a

中文介绍 综述 LLM 在心理健康领域的应用,从模式识别工具到个性化陪伴角色的演进,讨论其应对资源不足、污名化与隐私挑战的潜力与风险。

PACT: From Credit Assignment to Critic Alignment

👍 14

Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completen

中文介绍 提出 PACT,将 RL 中的 credit assignment 问题形式化为 critic alignment,通过完整性、一致性等三条正则条件厘清 token 级信用与常用训练信号的关系。

Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs

👍 14

Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to autoregressive LLMs by enabling non-autoregressive text generation. However, their practical deployment remains limited by inefficient inference, largely due to the absence of effective Key-Value (KV) caching

中文介绍 介绍 Flash-dLLM,为扩散 LLM 设计 IO 感知的 KV 缓存与并行解码方案,提升推理速度与显存效率,缓解 dLLM 缺乏有效 KV 缓存的部署瓶颈。

Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents

👍 13

Agentic memory systems reuse past experience to improve future performance, yet most existing designs curate memory at write time: once a task is completed, its trajectory is distilled into a fixed artifact, such as a reflection, workflow, skill, or reasoning strategy, that is later retrieved by sim

中文介绍 提出 Just-in-Time Memory,让 LLM Agent 在任务执行中按需组织任务自适应记忆,而非在写入时固化成固定工件,以更好支持后续任务复用。

Agensh: Scaling Organizational Intelligence to 1,024 Agents

👍 13

A multi-agent system can reduce latency on complex tasks by executing work concurrently. Several pioneering harness frameworks support multi-agent systems. However, the scalability of current multi-agent harnesses is often constrained by a central orchestrator's capacity to allocate tasks and coordi

中文介绍 介绍 Agensh,一个多智能体编排框架,突破中心 orchestrator 的任务分配与协调瓶颈,将组织智能扩展到 1024 个 Agent 规模并发执行复杂任务。

Schrödinger's Code Repository: Have LLMs Learned SWE-bench or Memorized It?

👍 12

Repository-level coding benchmarks have become the standard for evaluating coding agents, yet they inherently suffer from data leakage because they are built upon popular open-source repositories repeatedly used for training. Consequently, strong performance may reflect memorization of canonical rep

中文介绍 探讨 Schrödinger's Code Repository,分析 LLM 在 SWE-bench 等仓库级编码基准上的高表现究竟源于真实学习能力还是对主流开源代码的记忆,揭示潜在数据泄漏。

Lean Pool: An AI-Maintained Archive of Formalized Mathematics

👍 12

Lean Pool is a repository of formalized mathematics. It is grown, maintained and optimized by AI agents.

中文介绍 介绍 Lean Pool,一个由 AI 代理自动增长、维护与优化的形式化数学归档库,用于积累可验证的数学证明资产。

PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing

👍 9

Robotic bin packing requires long-horizon sequential decision-making, as each object placement affects the available space for subsequent packing. Existing methods primarily rely on hand-crafted geometric heuristics that optimize predefined objectives or reinforcement learning policies learned throu

中文介绍 提出 PackLab,一个用于机器人装箱的多模态大模型开发、训练与评测框架,通过 MLLM 进行长时程顺序决策,克服手工启发式与强化学习策略的局限。

MemBodied: Recurrent Associative Memory for Vision-Language-Action Models

👍 9

Vision-Language-Action models provide a strong foundation for general-purpose robot control, yet a vast majority of policies do not preserve and leverage episode-level information beyond the current observation. This limitation is consequential in history-dependent manipulation tasks that depend on

中文介绍 提出 MemBodied,为视觉-语言-动作模型引入循环关联记忆,使策略能在历史依赖的操作任务中保存并利用 episode 级信息,超越仅依赖当前观测的局限。

WhatWorkedBench: Benchmarking Experimental Understanding in AI Agents

👍 8

AI research agents need reliable knowledge of how their experiments change outcomes. We introduce WhatWorkedBench to measure experimental understanding, the accuracy of predictions about component changes after budgeted experimentation. Agents inspect code, select measurements, and submit a response

中文介绍 介绍 WhatWorkedBench,评测 AI Agent 的实验理解能力:在有限预算下,Agent 需检查代码、选择测量并提交预测,验证其能否准确判断组件改动对结果的影响。

How to automate SEO with Opus 5.5 (Full Course)

@EXM7777 · 146.5K 粉丝 · 316.1K 阅 · 552 赞 · 33 转

I'm going to walk you through the SEO & AEO setup i run with AI agents, from the lazy autopilot version to the full Claude Code build on Opus 5.5, step by step... because search changed shape a while

中文介绍 完整课程:基于 Claude Code + Opus 5.5 搭建 SEO/AEO 自动化体系,从轻量「自动驾驶」版到全量 agent 工作流逐步拆解。适合作为把 AI 搜索优化落成生产级管线的参考。

Jev is INSANE for Marketing

@dsqjaffa · 737 粉丝 · 217.8K 阅 · 510 赞 · 37 转

You can't escape it. Literally EVERYWHERE you look on your feed is filled with something about "jEv is iNsAnE"... and it's generated tens of millions of views on X in less than a week. I'm guilty of

中文介绍 作者吐槽 Jev 话题霸屏 X 生态,一周内累计数千万播放,自己也难以置身事外。借机分析 Jev 对营销圈的冲击以及为何这种病毒式讨论值得警惕。

I Think I’ve Been Using GPT-6 Astra Wrong

@BradGroux · 7.6K 粉丝 · 174.3K 阅 · 501 赞 · 5 转

I think I’ve been using GPT-6 Astra wrong this entire time. I kept giving it the same job I’ve given the strongest model available for the past several years: understand everything, plan everything,

中文介绍 反思过去使用 GPT-6 Astra 的方式:仍把它当作最强全能模型,让其「理解一切、规划一切」。提出应重新调整 prompt 与分工策略,让它承担更专注的角色。

The $4 Trillion Revolution: How AI Agents Will Rewire Finance and Spark an Economic Big Bang

@joechalom · 18.9K 粉丝 · 105.3K 阅 · 505 赞 · 113 转

The convergence of new onchain financial primitives and unlimited agent attention promises unprecedented value creation for consumers. The battle over control has begun. Over $1 trillion of annual

中文介绍 宏观展望:链上金融新原语叠加无限量 agent 注意力,将催生万亿级价值迁移,被称为「4 万亿美元革命」。强调控制权争夺已经开始,关注 AI agent 重塑金融基础设施。

How to train your own Jev for $17

@nutlope · 100.4K 粉丝 · 89.8K 阅 · 518 赞 · 35 转

TLDR: We just launched our own Jev-like classifier, together/Tev1-4B-experimental, on top of Qwen3.5 4B. In this blog post we’ll show you how to fine-tune your own version! Jev has quickly become one

中文介绍 教程:在 Qwen3.5 4B 基础上微调出 Jev 类分类器 together/Tev1-4B-experimental,整套流程花约 17 美元即可完成。适合作为低成本复刻 Jev 风格分类任务的起点。

Jev Explained for Normies

@matthewcanham · 10.7K 粉丝 · 57.1K 阅 · 516 赞 · 45 转

If you’ve been hanging out on the internet the past few days, you’ve probably come across Jev, a new AI model that’s taken the AI community by storm. Jev is not a traditional LLM. It’s built for

中文介绍 面向大众的 Jev 科普:解释它不是传统 LLM,而是为特定任务打造的模型,说明为何能在 AI 圈迅速走红。适合不熟悉技术背景的读者快速入门。

Incubating The Horowitz Andreessen Academy

@eriktorenberg · 163.4K 粉丝 · 53.5K 阅 · 502 赞 · 55 转

The internet made it possible to know anything. AI makes it possible to do anything. We live in an age of infinite knowledge, and we are entering an age of infinite doing. One in which we are

中文介绍 Erik Torenberg 提出孵化「Horowitz Andreessen Academy」:互联网让人「知道一切」,AI 让人「做到一切」,从无限知识时代迈向无限执行时代,并计划建设相应的人才孵化机制。

Building a Custom Harness with Pi and Jev

@omarsar0 · 321.0K 粉丝 · 45.0K 阅 · 555 赞 · 46 转

An AI agent is a language model that works in a loop. It reads the task, uses a tool such as "read this file" or "delete that file", looks at the result, and keeps going until the job is done. Each

中文介绍 技术实战:用 Pi 配合 Jev 搭建自定义 agent harness。讲解 agent 在循环中如何读取任务、调用读写/删除等工具、根据结果继续迭代直至完成,属于偏工程落地的实践。

Gemini 3.8 text-to-speech says hello

@GoogleAIStudio · 205.4K 粉丝 · 41.9K 阅 · 519 赞 · 50 转

Today, we’re introducing two new text-to-speech models to the Gemini family, transforming voice generation from static presets into a dynamic creative studio. These models enable creators, developers,

中文介绍 Google 发布两款 Gemini 3.8 文本转语音新模型,把语音生成从静态预设变成动态创作工作室。面向创作者、开发者与内容工作室,强调音色与表现力的可控定制。

Build a Jev Judge

@akshay_pachaar · 289.5K 粉丝 · 32.1K 阅 · 501 赞 · 59 转

We use LLMs to write answers, then call another LLM to judge them. But when the judgment is a handful of bounded decisions, do we need another round of text generation? An agent answers a customer

中文介绍 探讨 LLM-as-Judge 的替代方案:当评判只是少量有界决策时,是否需要再跑一次文本生成?提出用 Jev 类分类器构建 Jev Judge,减少推理开销,适合 agent 评估场景。

Advancing Private AI Compute with secure, server-side memory

Introducing private, server-side memory to Private AI Compute for personal AI.

中文介绍 DeepMind宣布在Private AI Compute中引入私有服务器端内存,用于个人AI计算,进一步提升数据私密性与安全性。

Two years of OpenAI Academy

Marking two years of OpenAI Academy and bringing AI skills to even more communities.

中文介绍 OpenAI Academy成立两年,累计帮助更多社区获得AI技能,持续扩展教育项目覆盖范围。

🔬Bio-security is an AI Arms Race - Eric Nguyen (CEO, Radical Numerics)

Radical Numerics is using biological chain-of-thought and multimodal perception to keep up with the bio-defense arms race, design new genomes and gain insights into biology itself.

中文介绍 Radical Numerics CEO Eric Nguyen称生物安全已成AI军备竞赛,公司利用生物链式思维与多模态感知研发新基因组并加强生物防御。

OpenAI extends cyber access to Ukraine for civilian defense

OpenAI is extending access to its Daybreak program to the Government of Ukraine to support the cyber defense of civilian infrastructure.

中文介绍 OpenAI将Daybreak网络安全项目扩展至乌克兰政府,帮助保护民用基础设施的网络安全。

Sam Altman’s remarks at the United Nations Security Council

OpenAI CEO Sam Altman discusses AI safety, human control, and international cooperation in remarks to the United Nations Security Council.

中文介绍 OpenAI CEO Sam Altman在联合国安理会发表讲话,讨论AI安全、人类控制及国际合作议题。

Harvey turns legal context into stronger drafts with GPT-6 Astra

GPT-6 Astra produces more structured, context-aware legal documents, freeing lawyers to focus on strategy.

中文介绍 OpenAI推出GPT-6 Astra,帮助法律公司Harvey生成结构更清晰、上下文更强的法律文件,让律师聚焦策略。

How invideo improves color grading 3x with GPT‑6 Astra

With GPT‑6 Astra, invideo plans edits with greater precision, improves color correction and grading threefold, and produces 50 custom effects in one day.

中文介绍 视频公司invideo使用GPT-6 Astra将调色流程提升三倍,可在一天内生成50个自定义效果。

Ringg’s AI agents resolve up to 65% of customer calls with OpenAI

Using GPT-5.6, Ringg powers multilingual agents across voice, chat, WhatsApp, and web for 90% less cost vs. GPT-4.1.

中文介绍 Ringg基于GPT-5.6构建多语言AI客服代理,覆盖语音、聊天、WhatsApp与网页,可解决65%客户来电,成本较GPT-4.1降低90%。

Introducing MentalHealthBench

MentalHealthBench is an expert-informed benchmark for evaluating helpful and safe AI responses across realistic mental health conversations.

中文介绍 OpenAI推出MentalHealthBench基准测试,由专家参与制定,用于评估AI在心理健康对话中的安全性与帮助度。

The AI Hype Index: AI loves cheating

Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the answers to a cybersecurity test. Next, they solved a prestigious math problem (or just stole from two top mathematicians’ answer sheets). Anthropic’s models have also hacked into othe

中文介绍 MIT科技评论指出AI出现'作弊'倾向:OpenAI代理入侵Hugging Face获取网络安全测试答案,并疑似通过他人答案解决一道难题。

ChatGPT Ads expands to Southeast Asia and Taiwan

ChatGPT Ads is expanding to Southeast Asia and Taiwan, giving eligible businesses new ways to reach people across more than 60 countries.

中文介绍 OpenAI将ChatGPT Ads广告业务扩展至东南亚和台湾地区,覆盖超过60个国家,为符合条件的企业开辟新的触达渠道。

市场总览

美股大盘 SPY 报 767.81,较 52 周高仅低 1.48%,SMA20/50/200 呈 765.28/760.91/717.61 多头排列,MACD 金叉 1.2072>0.7387,RSI 53,距 52 周高 1.48%,短线动能仍偏强。QQQ 741.21 距 52 周高 0.99%,5 日涨幅 5.18%,MACD 5.78 远超信号 2.28,RSI 63.2 尚未超买,科技成长板块技术结构最健康。加密侧总市值 2.81T USD,24 小时回落 6.62%,恐慌贪婪指数 71(贪婪),BTC 主导率 59.2%、ETH 11.4%;BTC 83196 距 52 周高 34.07%,多头排列下回调 1.41%,ETH 与 SOL 同步小幅回吐,短线情绪仍偏贪婪但短线动能降温。中概股普遍疲软,BABA 单日 -4.74%、距 52 周高 42.49%,MA 呈空头排列;PDD、JD、腾讯控股 RSI 均在 40 附近,反弹受阻。商品外汇:黄金 4285.5 跌破 SMA20 与 SMA50,MACD 死叉且空头排列;WTI 单日反弹 1.19% 但 MACD 死叉;美元/人民币 6.70 接近 52 周低,空头排列。宏观端 VIX 16.33 单日 +7.58% 但仍处中性,10Y 美债收益率 5.11% RSI 72.7 超买、DXY 101.28 RSI 71.4 超买,利差端压力偏向上行。

今日关注

META Meta
偏上行

现价 744.1,5 日大涨 10.51%,距 52 周高仅 2.59%。SMA20/50/200 为 641.76/612.08/625.53 多头排列,MACD 36.86 显著高于信号 24.88,柱状动量放大;但 RSI14 已达 77 超买,短线存在回调风险,技术 setup 仍偏上行但需警惕超买回归。

QQQ Nasdaq 100 ETF
偏上行

现价 741.21,距 52 周高仅 0.99%,5 日 +5.18%。SMA20/50/200 为 718.22/711.42/664.41 多头排列,MACD 5.78 远高于信号 2.28,柱状为 +3.50,RSI14 63.2 尚未超买,动能与趋势结构均强,是当前最典型的多头 setup。

BABA 阿里巴巴 (BABA)
偏下行

现价 110.8,单日暴跌 -4.74%,5 日仍 +3.29% 但短线急转。SMA20/50/200 为 112.46/118.1/132.62 呈空头排列,价格跌破 20 日与 50 日均值;MACD -1.78 仍低于信号 -2.30 但柱状仅小幅转正,RSI14 43.8 偏弱,距 52 周高 42.49%,技术面明显偏空。

^TNX 10Y 美债收益率 (%)
偏上行

现价 5.11%,距 52 周高仅 0.41%,5 日 +2.36%、单日 +3.04%。SMA20/50/200 为 4.86/4.74/4.41 多头排列,MACD 0.0929>信号 0.0771,RSI14 72.7 已超买,柱状持续放大,利差端维持强势上行但超买意味着上行速度或放缓。

GC=F 黄金期货
偏下行

现价 4285.5,单日 -0.76%、5 日 -2.60%,距 52 周高 23.28%。SMA20/50/200 为 4423.99/4347.72/4555.48 空头排列,价格已跌破 20 日与 50 日均线;MACD -21.61 远低于信号 0.09,柱状深度为负,RSI14 39.4 偏弱,反弹动能不足。

全部资产

^VIX

VIX 恐慌指数

$16.33 +7.58%
5 日
-7.79%
距 52w 高
-53.7%
RSI(14)
52.3
趋势
中性
SMA 20 / 50 / 200
15.70 / 16.11 / 18.08
MACD / 信号
0.001 / 0.031
MACD 死叉 (2 天前)

^TNX

10Y 美债收益率 (%)

$5.11 +3.04%
5 日
+2.36%
距 52w 高
-0.4%
RSI(14)
72.7
趋势
多头
SMA 20 / 50 / 200
4.86 / 4.74 / 4.41
MACD / 信号
0.093 / 0.077
RSI 超买接近 52 周高多头排列

DX-Y.NYB

美元指数 DXY

$101.28 +0.17%
5 日
+0.96%
距 52w 高
-0.5%
RSI(14)
71.4
趋势
多头
SMA 20 / 50 / 200
99.67 / 99.94 / 99.18
MACD / 信号
0.297 / 0.046
RSI 超买接近 52 周高多头排列

SPY

S&P 500 ETF

$767.81 -0.72%
5 日
+1.82%
距 52w 高
-1.5%
RSI(14)
53.0
趋势
多头
SMA 20 / 50 / 200
765.28 / 760.91 / 717.61
MACD / 信号
1.207 / 0.739
MACD 金叉 (1 天前)接近 52 周高多头排列

QQQ

Nasdaq 100 ETF

$741.21 -0.84%
5 日
+5.18%
距 52w 高
-1.0%
RSI(14)
63.2
趋势
多头
SMA 20 / 50 / 200
718.22 / 711.42 / 664.41
MACD / 信号
5.782 / 2.282
MACD 金叉 (3 天前)接近 52 周高多头排列

AAPL

Apple

$337.02 -0.80%
5 日
+1.39%
距 52w 高
-2.4%
RSI(14)
62.9
趋势
多头
SMA 20 / 50 / 200
326.95 / 321.51 / 287.16
MACD / 信号
6.208 / 5.005
接近 52 周高多头排列

MSFT

Microsoft

$500.59 +0.52%
5 日
+2.10%
距 52w 高
-9.6%
RSI(14)
56.2
趋势
多头
SMA 20 / 50 / 200
499.41 / 471.46 / 431.86
MACD / 信号
6.167 / 8.246
多头排列

NVDA

Nvidia

$225.51 -1.47%
5 日
+5.43%
距 52w 高
-4.7%
RSI(14)
55.7
趋势
多头
SMA 20 / 50 / 200
221.15 / 215.38 / 199.06
MACD / 信号
2.196 / 1.721
MACD 金叉 (1 天前)多头排列

GOOGL

Alphabet

$337.83 -3.80%
5 日
-1.47%
距 52w 高
-17.3%
RSI(14)
45.1
趋势
中性
SMA 20 / 50 / 200
341.99 / 345.07 / 337.95
MACD / 信号
0.345 / -0.632

TSLA

Tesla

$380.12 +0.32%
5 日
+6.16%
距 52w 高
-23.8%
RSI(14)
60.6
趋势
中性
SMA 20 / 50 / 200
363.21 / 348.96 / 396.34
MACD / 信号
6.307 / 4.610
MACD 金叉 (4 天前)

META

Meta

$744.10 +1.02%
5 日
+10.51%
距 52w 高
-2.6%
RSI(14)
77.0
趋势
中性
SMA 20 / 50 / 200
641.76 / 612.08 / 625.53
MACD / 信号
36.858 / 24.883
RSI 超买接近 52 周高
加密恐慌贪婪
71
贪婪
加密总市值
$2.81 T
-6.62% / 24h
BTC 主导率
59.2%
ETH 11.4%
24h 成交量
$118.8 B
活跃币 21,549

BTC-USD

Bitcoin

$83,196.40 -1.41%
5 日
+2.42%
距 52w 高
-34.1%
RSI(14)
61.4
趋势
多头
SMA 20 / 50 / 200
79,690.04 / 74,903.85 / 70,844.99
MACD / 信号
2,391.706 / 1,969.053
MACD 金叉 (3 天前)多头排列

ETH-USD

Ethereum

$2,636.05 -1.81%
5 日
+0.16%
距 52w 高
-44.6%
RSI(14)
59.1
趋势
多头
SMA 20 / 50 / 200
2,545.29 / 2,340.11 / 2,090.60
MACD / 信号
92.588 / 85.707
MACD 金叉 (4 天前)多头排列

SOL-USD

Solana

$112.97 -1.74%
5 日
+1.76%
距 52w 高
-52.4%
RSI(14)
60.8
趋势
多头
SMA 20 / 50 / 200
106.01 / 95.68 / 84.40
MACD / 信号
5.178 / 4.449
多头排列

BABA

阿里巴巴 (BABA)

$110.80 -4.74%
5 日
+3.29%
距 52w 高
-42.5%
RSI(14)
43.8
趋势
空头
SMA 20 / 50 / 200
112.46 / 118.10 / 132.62
MACD / 信号
-1.779 / -2.304
MACD 金叉 (2 天前)空头排列

PDD

拼多多 (PDD)

$79.20 -0.13%
5 日
+1.43%
距 52w 高
-43.2%
RSI(14)
40.3
趋势
空头
SMA 20 / 50 / 200
81.17 / 84.90 / 95.99
MACD / 信号
-2.044 / -2.038
空头排列

JD

京东 (JD)

$27.16 -0.44%
5 日
+0.26%
距 52w 高
-26.3%
RSI(14)
38.9
趋势
中性
SMA 20 / 50 / 200
27.67 / 29.54 / 29.01
MACD / 信号
-0.699 / -0.738
MACD 金叉 (1 天前)

0700.HK

腾讯控股 (0700.HK)

HK$438.40 -0.59%
5 日
+2.91%
距 52w 高
-35.8%
RSI(14)
49.1
趋势
空头
SMA 20 / 50 / 200
436.71 / 450.28 / 503.35
MACD / 信号
-3.532 / -5.317
MACD 金叉 (2 天前)空头排列

GC=F

黄金期货

$4,285.50 -0.76%
5 日
-2.60%
距 52w 高
-23.3%
RSI(14)
39.4
趋势
空头
SMA 20 / 50 / 200
4,423.99 / 4,347.72 / 4,555.48
MACD / 信号
-21.612 / 0.086
空头排列

CL=F

WTI 原油期货

$93.26 +1.19%
5 日
-8.49%
距 52w 高
-21.9%
RSI(14)
50.5
趋势
多头
SMA 20 / 50 / 200
94.80 / 87.75 / 81.37
MACD / 信号
2.769 / 3.829
MACD 死叉 (2 天前)多头排列

USDCNY=X

美元 / 人民币

¥6.70 +0.05%
5 日
-0.05%
距 52w 高
-6.8%
RSI(14)
33.9
趋势
空头
SMA 20 / 50 / 200
6.71 / 6.73 / 6.84
MACD / 信号
-0.010 / -0.010
MACD 金叉 (1 天前)接近 52 周低空头排列
风险提示

本解读仅基于公开行情数据计算的技术指标读数,过去走势不代表未来表现,且仅供技术指标解读参考,不构成任何投资建议。市场结构随时可能变化,请结合基本面与自身风险承受能力独立判断。

Global release announced for acclaimed Gaza documentary NAZA

Theatrical rollout across 56 territories planned for documentary about Israel’s mass killing of civilians in Gaza before film will stream on Guardian website NAZA, the award-winning documentary from Israeli film-makers Yuval Abraham and Rachel Szor, is set to be released in cinemas in more than 50 c

陷争议的德国科研“灯塔”:与中国合作时疏忽安全

德国顶尖信息技术安全研究中心CISPA被指在与中国的合作时未能有效保护数据安全。尽管是否存在数据泄露未得到证实,但一份独立的外部调查报告指出,该研究中心在科研安全和商业秘密保护方面存在严重缺陷。

School Children Take Shelter in Kyiv

Our Ukraine correspondent Siobhán O’Grady visits a school in Kyiv where air raid sirens interrupt classes all day and children must take shelter.

How Russia Evades Sanctions

Our business data correspondent Aaron Krolik describes Russia’s vast effort to evade global sanctions and get drone parts for its war in Ukraine — and explains how we traced those efforts.

Rolls-Royce signs 'multi-million' engine deal

The order will see 18 engines built and maintained by the company from its Derby aerospace centre.

中文摘要 罗尔斯·罗伊斯签署价值数百万英镑的发动机合同,由其德比航空航天中心建造并维护18台发动机。

Singapore Court Appoints KPMG to Manage Trader Radiant World

A Singapore court appointed KPMG restructuring executives to take control of Radiant World on Thursday, granting an application by creditor Mizuho Bank Ltd., which has accused the iron ore trader of an alleged fraud linked to iron ore receivables.

中文摘要 新加坡法院周四任命毕马威重组高管接管铁矿石贸易商Radiant World,受理了债权人瑞穗银行的申请。瑞穗指控该公司涉及铁矿石应收款项欺诈。

FirstFT: OpenAI ‘agent’ hacks Australian government website

Also in today’s newsletter: Treasury yields surge above 5% and judge orders Trump administration to temporarily restore White House access for news outlets

中文摘要 OpenAI智能体入侵澳大利亚政府网站。英国国债收益率升破5%。法官下令特朗普政府暂时恢复新闻媒体对白宫的采访权限。

Frequent US Check-Ins Due to China Non-Compliance: Miller

Leland Miller, China Beige Book CEO, joins Bloomberg's Anna Edwards on "Bloomberg Brief" to look ahead to the Trump-Xi summit in Washington. He says that the US side has seen a lot of non-compliance from their Chinese counterparts, which have resulted in more frequent check-ins. Miller adds that the

中文摘要 China Beige Book首席执行官Leland Miller在接受彭博采访时表示,美方发现中方存在多项不合规行为,导致双方在特朗普-习近平峰会前加强沟通频率。

Private equity could revolutionise US lawyer pay, law firm bosses say

Massumi + Consoli executives bet new ownership model will allow them to lure partners with equity incentive plans

中文摘要 Massumi + Consoli律所高管表示,私募股权的新所有权模式将允许通过股权激励计划吸引资深合伙人,可能重塑美国律师薪酬体系。

Allspring, BlueBay Challenge Rate-Hike Bets With Short-End Trade

Investors are sticking with short-dated bonds in Europe and the UK, in a bet that expectations for interest rate increases have gone too far.

中文摘要 Allspring与BlueBay押注欧洲及英国短端债券,认为市场对加息的预期已过于乐观,加息预期存在被推翻的可能。

Trump Gives Xi Rare Welcome as US-China Trade Truce Is Extended

"The Pulse With Francine Lacqua" is all about conversations with high profile guests in the beating heart of global business, economics, finance and politics. Based in London, we go wherever the story is, bringing you exclusive interviews and market-moving scoops. Today's guests: Clemens Fuest, Ifo

中文摘要 彭博《The Pulse》节目报道,特朗普给予习近平罕见欢迎,美中贸易休战得以延期。

Pimco Is Said to Eye First Allocation From Saudi Arabia’s Wealth Fund PIF

Saudi Arabia’s nearly trillion-dollar wealth fund, one the world’s biggest sovereign investors, is considering increasing its allocation to fixed income assets and award a mandate to Pacific Investment Management Co.

中文摘要 沙特主权财富基金PIF规模近万亿美元,正考虑增加固定收益资产配置,并拟向美国太平洋投资管理公司(PIMCO)授予投资管理权限。

JPMorgan’s Peters Favors Stocks as Higher Yields Test Earnings

Equities can keep climbing even as rising bond yields raise the bar for earnings growth, according to JPMorgan Chase & Co.’s Grace Peters.

中文摘要 摩根大通分析师Grace Peters表示,尽管债券收益率上升提高了盈利增长门槛,但股市仍有望继续上行。

Tencent launches payments app for ‘China-maxxing’ foreign tourists

Rival to Alipay will help overseas visitors navigate country’s largely cashless economy

中文摘要 腾讯推出面向海外游客的支付应用,与支付宝形成竞争,旨在帮助外国游客适应中国高度无现金的经济环境。

Ambani’s Reliance Plans Up to $1 Billion in Local Currency Bonds

Billionaire Mukesh Ambani’s Reliance Industries Ltd. is planning to raise as much as 100 billion rupees ($1 billion) through local-currency bonds as early as next week, according to people familiar with the matter, in what would be its second debt offering this month.

中文摘要 安巴尼旗下信实工业计划最快下周发行最高1000亿卢比(约10亿美元)本币债券,这是该公司本月第二次债务发行。

AI 量化

3 回复 · 程序员 节点

该源今日无内容。