TypeScript · ★ 13,598 · 🍴 1,276 · 📈 1,111 stars today
TencentDB Agent Memory is a team-level memory hub for AI Agents — turning conversations, docs, and code into four reusable memory assets (Chat Memory, Skill, LLM-Wiki, Code-Graph) that are governed, shared, and equipped across agents and frameworks.
中文介绍 腾讯云开源的AI Agent团队级记忆中枢，将对话、文档和代码转化为四种可复用资产（聊天记忆、技能、LLM维基、代码图谱），支持治理和共享，帮助企业级AI Agent沉淀并复用团队知识。
PowerShell · ★ 17,880 · 🍴 2,452 · 📈 2,297 stars today
Reverse Engineering / Authorized Penetration Testing / Security Research Skill Router Pack AI-powered routing + On-demand toolchain bootstrapping + Self-evolving knowledge base Supports Claude Code, Kiro, Cursor, Cline, and other AI coding clients 逆向/渗透/安全技能路由包 - AI 自动路由 + 按需自举工具链 + 自动进化经验库 | 支持 Cla
中文介绍 面向逆向工程、授权渗透测试和安全研究的AI技能路由包，支持智能路由和按需工具链引导，内置自进化知识库，兼容Claude Code等AI编程助手。
Rust · ★ 10,021 · 🍴 660 · 📈 2,540 stars today
Fast Rust library for PDF inspection, classification, and text extraction. Intelligently detects scanned vs text-based PDFs to enable smart routing decisions.
中文介绍 用Rust编写的PDF快速检查与分类库，能智能识别扫描件还是文本型PDF，并提取文本内容，从而为后续处理流程提供更精准的路由决策。
Python · ★ 680 · 🍴 68 · 📈 148 stars today
ADR secures enterprise AI agents through observability, security benchmarking, and threat detection. Deployed at Uber.
中文介绍 Uber开源的AI Agent安全防护框架，通过可观测性、安全基准测试和威胁检测三大能力，保障企业级AI Agent的安全运行，已在Uber内部落地。
Shell · ★ 266,482 · 🍴 23,829 · 📈 653 stars today
An agentic skills framework & software development methodology that works.
中文介绍 一套实用的智能体技能框架和软件开发方法论，强调可落地的实践，帮助开发者构建、编排和迭代AI代理的工作流程。
Jupyter Notebook · ★ 116,256 · 🍴 61,547 · 📈 783 stars today
21 Lessons, Get Started Building with Generative AI
中文介绍 微软出品的生成式AI入门课程，包含21节实践课，从基础概念到构建应用，覆盖提示工程、RAG、微调等核心主题，适合初学者快速上手。
TypeScript · ★ 50,789 · 🍴 3,628 · 📈 11 stars today
Fast, easy and reliable testing for anything that runs in a browser.
中文介绍 专为浏览器应用设计的前端测试框架，提供快速、可靠、易用的API，支持端到端测试、组件测试和实时调试，让测试过程更加直观高效。
Jupyter Notebook · ★ 28,378 · 🍴 3,061 · 📈 1,711 stars today
AirLLM 70B inference with single 4GB GPU
中文介绍 一个轻量推理方案，能让70B参数大模型在单张4GB GPU上运行。通过分层加载和显存优化，大幅降低了大模型推理的硬件门槛。
JavaScript · ★ 65,930 · 🍴 9,528 · 📈 10 stars today
A bundler for javascript and friends. Packs many modules into a few bundled assets. Code Splitting allows for loading parts of the application on demand. Through "loaders", modules can be CommonJs, AMD, ES6 modules, CSS, Images, JSON, Coffeescript, LESS, ... and your custom stuff.
中文介绍 JavaScript生态中最流行的模块打包器，将众多模块打包成少量静态资源，支持代码分割按需加载，并通过loader处理CommonJS、ES模块等多种格式。
C++ · ★ 29,374 · 🍴 5,372 · 📈 10 stars today
Fast C++ logging library.
中文介绍 高性能C++日志库，支持异步日志、多线程、结构化格式化和多种输出目标，以极低的性能开销满足高频日志写入需求。
Rust · ★ 108,061 · 🍴 6,306 · 📈 31 stars today
A modern runtime for JavaScript and TypeScript.
中文介绍 由Node.js作者打造的新一代JavaScript/TypeScript运行时，内置类型支持、安全的权限模型和标准库，提供开箱即用的现代开发体验。
TypeScript · ★ 7,292 · 🍴 579 · 📈 559 stars today
🎯 All you need. Nothing you don't. Open source project management that works for you, not against you.
中文介绍 开源项目管理工具，界面简洁、功能聚焦，提供任务看板、时间线等核心功能，强调“你需要的都有、不需要的都没有”，可自托管使用。
Python · ★ 12,412 · 🍴 3,480 · 📈 432 stars today
A framework for building realtime voice AI agents 🤖🎙️📹
中文介绍 用于构建实时语音AI代理的框架，支持音视频输入输出，提供语音识别、合成和对话编排能力，可快速开发语音助手、虚拟互动等应用。
TypeScript · ★ 100,827 · 🍴 27,401 · 📈 13 stars today
Deliver web apps with confidence 🚀
中文介绍 企业级Web应用开发框架，提供完整的组件化架构、依赖注入、路由、表单和状态管理方案，配合TypeScript构建可维护的大型应用。
TypeScript · ★ 96,475 · 🍴 5,533 · 📈 52 stars today
A utility-first CSS framework for rapid UI development.
中文介绍 实用工具优先的CSS框架，通过组合原子类快速构建界面，支持响应式设计、主题定制和构建时优化，成为现代前端开发的热门选择。
Python · ★ 19,331 · 🍴 2,407 · 📈 320 stars today
Edit videos with coding agents
中文介绍 面向AI编码代理的视频编辑工具，让代理通过代码或自然语言指令完成视频剪辑、拼接、特效等操作，将视频处理流程自动化。
Go · ★ 30,779 · 🍴 1,982 · 📈 922 stars today
DeepSeek-native AI coding agent for your terminal. Engineered around prefix-cache stability — leave it running.
中文介绍 基于DeepSeek的终端AI编程代理，专为前缀缓存稳定性优化，可长时间运行而无需重复加载，辅助代码生成、修改与调试。
TypeScript · ★ 23,870 · 🍴 1,957 · 📈 40 stars today
Official Compound Engineering plugin for Claude Code, Codex, Cursor, and more
中文介绍 复合工程（Compound Engineering）的官方插件，可扩展Claude Code、Codex、Cursor等AI编码工具，用于协调多智能体协作和工程流程。
👍 140
Speech and audio generation is often needed in animation dubbing, audio drama, movies, advertising, games, podcasts, and short-video production. In these scenarios, creators may need to design voices without reference recordings, control speaker styles with natural language, support acoustic scenes
中文介绍 提出 SwanTale，统一多说话人语音与音频生成框架，支持指令控制与零样本任务。通过自然语言控制说话人风格并建模声学场景，适用于动画配音、有声剧、广告等场景。无需参考录音即可设计声音，提升配音与音频制作灵活性。
👍 127
Large language model (LLM) agents increasingly undertake long-horizon tasks that require sustained reasoning, tool use, and revision across many interdependent steps. However, existing agent harnesses maintain task execution, task state, and completion assessment within a growing context, making the
中文介绍 提出 LongHorizon-Harness，面向真实世界长程任务的 LLM 智能体框架。针对现有 harness 将任务执行、状态与完成评估全部塞入增长上下文的瓶颈，通过解耦状态管理与完成评估，支撑多步推理、工具调用与反复修正，提升长程任务的持续执行能力。
👍 55
On-policy (self) distillation (OPSD) is increasingly adopted for language-model post-training. It strengthens the teacher with privileged information but can induce a privilege illusion: the student learns privilege-dependent behavior it cannot reproduce from its inference-time context, yet behaves
中文介绍 提出 DAPD（Dual-Anchored Policy Distillation），解决 on-policy self-distillation 中的特权幻觉问题——学生学到依赖特权信息、在推理上下文无法复现的行为。通过双锚定约束学生策略与可复现信息对齐，同时保持教师优势，缓解能力错位，提升后训练蒸馏稳定性。
👍 50
On-policy distillation (OPD), which aligns a student with the teacher's token-level distribution on the student's own rollouts, is an effective paradigm for transferring capabilities across LLMs. Prevailing approaches assume a teacher at least as capable as the student: they either distill a larger
中文介绍 研究弱到强的 on-policy distillation（OPD）：不再假设教师必须强于学生，而用较弱教师为学生 rollout 提供 token 级分布监督。该方法可在教师能力有限时仍实现能力迁移，为 LLM 蒸馏提供更灵活的训练范式。
👍 49
Existing skill generation methods largely rely on heuristics or pipeline-style consolidation, which must be specially designed for different evidence sources. In contrast, learning-based approaches offer a more unified way to model skill generation across heterogeneous sources. However, learning-bas
中文介绍 提出基于强化学习的渐进式智能体技能生成方法，用统一方式从异构信息源建模技能生成，替代依赖启发式或流水线设计的传统方案。通过 RL 逐步生成可复用技能，提升跨来源的技能学习与泛化能力。
👍 42
Sparse retrieval underpins modern search systems, from web search to retrieval-augmented generation. Existing work has introduced Learned Sparse Retrieval (LSR) to push beyond exact lexical matching toward richer semantics. Yet LSR has so far remained tied to encoder-style bidirectional architecture
中文介绍 提出 UEmbed，统一稀疏与稠密的多模态嵌入框架。现有 Learned Sparse Retrieval 局限于 encoder 风格双向架构，UEmbed 将其扩展到稀疏检索与稠密检索联合建模，在保留语义匹配能力的同时支持多模态检索，提升现代搜索与 RAG 系统的灵活性。
👍 30
Controllable video generation models are increasingly being developed as world models. Accordingly, evaluating them in this role extends beyond the apparent appearance of generated videos to the inherent reactivity of the worlds they depict: the ability to infer from the scene state how the world sh
中文介绍 提出 WorldExam 基准，评估可控视频生成模型作为世界模型时对场景固有反应性的建模能力，而不仅看生成视频的表观质量。测试模型能否从场景状态推断世界应如何响应，为视频世界模型的评估提供新维度。
👍 30
Recent Vision-Language-Action (VLA) models for autonomous driving (AD) increasingly utilize chain-of-thought (CoT) supervision to enhance the reasoning capabilities of their Vision-Language Model (VLM) components, yet existing annotation pipelines commonly expose the teacher model to the logged grou
中文介绍 提出延迟暴露未来轨迹的 CoT 监督方法，用于自动驾驶 VLM/VLA 模型推理训练。现有标注管线会让教师模型看到日志中的真实未来轨迹，造成信息泄漏；该方法在生成推理监督时隐藏未来信息，促使模型基于可验证的当下场景进行推理，提升推理可信度。
👍 29
Reinforcement learning with verifiable rewards (RLVR) broadcasts a single response-level reward to every token, while on-policy distillation (OPD) scores each token against a stronger teacher for a dense advantage but caps performance at teacher quality and discourages exploration beyond it. Their c
中文介绍 提出 SAF-OPD，稳定融合 RLVR 的响应级奖励与 on-policy distillation 的 token 级稠密优势。RLVR 奖励稀疏且全局，OPD 受教师质量上限约束；SAF-OPD 通过优势融合兼顾探索与教师监督，避免性能封顶，提升可验证奖励下的策略优化效果。
👍 27
Agent skills have become an important mechanism for equipping language-model agents with reusable procedural knowledge. However, providing skills alone does not guarantee that current models can effectively identify, apply, and coordinate them. To improve skill-use capabilities, we introduce SKT, a
中文介绍 提出 SKT，通过经过验证的合成数据大规模训练智能体的技能使用能力。仅提供技能库不足以让模型有效识别、应用和协调技能，SKT 自动生成带验证的技能使用训练数据，提升 LLM 智能体在真实任务中调用与组合技能的能力。
👍 25
AI-assisted coding increasingly translates informal user intent into executable software, yet coding requests often contain ambiguities that recur in user-specific ways across tasks and sessions. Existing disambiguation methods typically address each ambiguous request in isolation within the current
中文介绍 提出跨会话个性化歧义消解基准，面向编程助手中随任务和会话反复出现的用户特有歧义。现有方法只针对当前会话孤立处理每次请求，该基准评估助手能否记忆并适应用户歧义偏好，以减少澄清次数、生成更符合预期的代码。
👍 23
We introduce DiffusionGemma, an experimental open-weight language model that uses discrete diffusion to generate text at exceptionally high speed. Rather than decoding one token at a time, DiffusionGemma iteratively refines blocks of 256 tokens in parallel, avoiding the sequential decoding bottlenec
中文介绍 发布 DiffusionGemma，基于离散扩散的开源权重语言模型。它不逐 token 解码，而是并行迭代精化 256 token 的块，避免顺序解码瓶颈，实现极高文本生成速度，同时保持生成质量，为高效推理提供新选择。
👍 22
Real-world software development requires coding agents to operate in shared workspaces where users may inspect and modify code during an ongoing task, yet existing repository-level benchmarks typically evaluate agents working alone or restrict user participation to messages. This leads us to ask: ho
中文介绍 提出 SWE-Touch 基准，评估编码智能体在共享工作区中与用户实时交互的能力——用户可能在任务进行中查看和修改代码。现有仓库级基准多让智能体独立工作或只允许消息交互，该基准考察智能体在代码被触碰时的适应与协作能力。
👍 21
Optimization-based latent reasoning improves large language model outputs by optimizing instance-specific continuous states at test time while keeping model parameters frozen. Existing methods, however, typically connect these states to the reasoning trajectory through decoded tokens, making sequenc
中文介绍 提出 GradCuit，通过信用分配的梯度流实现测试时潜在推理。现有优化式潜在推理常经解码 token 连接优化状态与推理轨迹，导致序列化依赖；GradCuit 直接传递经信用分配的梯度，使测试时推理更鲁棒、可解释，并保持模型参数冻结。
👍 18
Large language models increasingly write and repair production code, yet evidence is mounting that their test-passing patches leave codebases harder to maintain. We identify one concrete source: deletion avoidance, the systematic tendency to retain code that an intended edit requires removing. Acros
中文介绍 识别并量化 LLM 代码编辑中的删除回避现象：系统性地保留本应被删除的代码。提出相应度量与缓解方法，通过调整训练或解码策略减少删除回避，使测试通过补丁不会降低代码库可维护性。
👍 15
Multimodal large language models (MLLMs) have advanced visual understanding and reasoning, yet their static parametric knowledge limits their ability to address knowledge-intensive and dynamically evolving open-world problems. To move beyond this limitation, multimodal deep search has emerged as a k
中文介绍 提出 DeepVoyager-VL，激励视觉在环搜索的长程多模态智能体。针对 MLLM 静态参数知识难以应对知识密集、动态开放世界的问题，通过视觉反馈驱动的深度搜索，让模型在长程任务中持续获取外部视觉信息，提升多模态推理能力。
👍 14
Recent 3D vision-language models (3D VLMs) construct geometry aware tokens by projecting 2D visual features into world coordinates, enabling spatial reasoning for tasks such as 3D question answering. However, this design generates thousands of tokens per scene, resulting in substantial computational
中文介绍 提出 3DZip，面向 3D 问答的空间感知特征多样性引导 token 压缩方法。3D VLM 每场景生成数千几何感知 token 导致高计算开销，3DZip 依据空间感知的特征多样性有选择地压缩 token，在保留空间推理能力的同时显著降低计算成本。
👍 12
Fixed-layout indoor furniture styling requires selecting assets that form a coherent room without changing the prescribed furniture categories, positions, orientations, or scales. Existing approaches typically retrieve each asset independently or rely on static local relations, making them prone to
中文介绍 提出 StyleForge，基于超图场反事实推理的固定布局室内家具风格搭配方法。现有方法独立检索家具或仅用静态局部关系，易导致整体不协调；StyleForge 在保持类别、位置、朝向、尺度不变的前提下，建模家具间整体关系并选出协调成套的家具。
👍 9
To operate robustly in open-world environments, autonomous agents should be able to infer the behavior of unfamiliar systems through interaction alone, even in the absence of documentation. However, existing tool-use benchmarks expose semantic tool schemas in static environments, allowing agents to
中文介绍 提出 ScrambleToolBench 基准，测试智能体在无文档且工具模式被打乱的环境中，仅靠交互推断陌生系统行为的能力。现有工具基准暴露静态语义 schema，让智能体可走捷径；该基准要求智能体即使已有内部地图指向下一步，也要穷尽搜索以发现真实工具语义，评估开放世界鲁棒性。
👍 8
Existing evaluations often reduce agent failures to system-level outcomes, obscuring where the fault originated and which intervention would improve the agent system. This creates a repair-assignment problem: the same visible failure may call for model post-training, harness engineering, environment
中文介绍 提出以交互为中心的分类法，用于定位 LLM 智能体失败来源。现有评估常将失败归为系统级结果，无法判断责任在模型、harness 还是环境；该分类法按交互阶段拆分失败，帮助选择模型后训练、harness 工程或环境调整等正确修复手段。
@wayne_liang_ · 1.5K 粉丝 · 999.9K 阅 · 637 赞 · 279 转
For eight weeks, a clone of me ran our sales front line: 300+ prospect calls a week, 132 new paying customers, 37 enterprise opportunities worth ~$3M. It also made up a price, emailed internal notes
中文介绍 博主休陪产假 8 周，AI 代理「克隆自己」负责销售前端：每周 300+ 通外呼，换来 132 个新付费客户、37 个企业机会，总价值约 300 万美元；但代理也会虚构价格、错发内部笔记。附完整代码，展示真实效果与坑。
@addyosmani · 407.9K 粉丝 · 363.3K 阅 · 510 赞 · 45 转
Taste is recognizing quality and greatness in the absence of all the evidence. Judgment is committing to it in the presence of some risk. It’s the ultimate foundation of great work. When you look at a
中文介绍 观点帖：品味是在缺乏证据时识别品质与卓越，判断力是在存在风险时仍敢于承诺，二者是伟大工作的根本。作者以「品味、判断力与 AI」为题，强调当 AI 加速产出，人类更需靠品味与判断力把关。
@RaoulGMI · 1.5M 粉丝 · 90.0K 阅 · 530 赞 · 72 转
Within about two years, most economic activity on Earth won't have a human anywhere in it. Think about a single trade today. Someone decides to make it, another person approves it, someone books it,
中文介绍 宏观预测帖：作者断言约两年内，地球上大部分经济活动将不再有真人参与。以一笔交易为例，从决策、审批到记账，人类环节将逐渐被 AI 取代，经济活动转入「隐形」的自动化体系。
@IBuzovskyi · 3.8K 粉丝 · 76.2K 阅 · 505 赞 · 34 转
Buzz launched July 21, 2026. Block open-sourced it under Apache-2.0. Desktop builds for macOS, Windows, and Linux shipped the same day. The pitch: Slack + GitHub + AI agents in one window. Built on
中文介绍 教程帖：Block 开源的桌面应用 Buzz（Apache-2.0），7 月 21 日上线，覆盖 macOS/Windows/Linux，主打「Slack + GitHub + AI 代理」三合一。附 Hermes Agent 接入 Buzz 完整指南。
@ashpreetbedi · 20.5K 粉丝 · 73.1K 阅 · 501 赞 · 50 转
Today I'm going to show you how to recursively improve your agents. We'll build an agent that starts at 7/10, then run a recursive auto-improvement loop until every probe passes. Here's how it works:
中文介绍 教程帖：演示如何递归改进 AI 代理——先构建一个初始水平 7/10 的代理，然后运行递归自动改进循环，持续用探针测试并自我优化，直到所有探针通过。给出可复现的实现路径。
@moonpay · 421.1K 粉丝 · 32.0K 阅 · 547 赞 · 62 转
MoonPay, the global financial technology company powering the movement of value across fiat and digital assets, has launched PayBox, the first payment vault built for AI that lets a person's AI agent
中文介绍 产品发布：MoonPay 推出 PayBox，声称是首个专为 AI 打造的支付保险库。它允许用户的 AI 代理在受控授权下保管、管理并转移法币与数字资产，未来可替代人类完成自动支付。
@Hi_Mrinal · 22.2K 粉丝 · 20.2K 阅 · 528 赞 · 41 转
The best way to learn from AI is to really reverse the whole process like to use ai for deep understanding rather than just "getting the answer" revert the typical workflow, instead of asking the AI
中文介绍 学习技巧帖：用 AI 学习的最佳方式是反转常规流程——让 AI 帮你深度理解，而不是直接要答案。与其让 AI 代替思考，不如通过追问、解释和反推，把 AI 当作陪练来真正掌握知识。
OpenAI explains recent third-party cybersecurity evaluation incidents and outlines new safeguards to strengthen AI model testing and evaluation.
中文介绍 OpenAI就近期涉及自身模型的第三方网络安全评估事件作出说明，并概述了新的安全防护措施，旨在强化AI模型测试与评估环节的安全性。
An external reconstruction of how Memory, Proactivity, Scheduling, Browser Use, Plugins, Skills and Tools work in the new ChatGPT Work.
中文介绍 Latent Space 发文拆解新版 ChatGPT Work，从外部还原其记忆、主动性、日程安排、浏览器使用、插件、技能与工具等功能的运作机制。
中文介绍 Hugging Face 博客介绍 Liquid AI 的 LFM2.5-2.6B 模型，该模型支持在本地部署智能体，可应用于多种场景。
**Alibaba** launched **Qwen3.8-Max**, enhancing multimodal capabilities and agent ecosystem integration. **NVIDIA** introduced **Alpamayo 2 Super** for autonomous vehicle reasoning, while **Mistral AI** released **Shieldstral**, a 3B parameter open-weights safety model for on-device moderation. **Po
中文介绍 阿里巴巴发布 Qwen3.8-Max，增强多模态能力与智能体生态集成；英伟达推出面向自动驾驶推理的 Alpamayo 2 Super；Mistral AI 发布30亿参数开源安全模型 Shieldstral，适用于端侧部署。
Qwen is so back!
中文介绍 Latent Space 的 AINews 报道，阿里推出 Qwen 3.8 Max（2.4万亿参数）及270亿参数的模型，均为开源权重，聚焦编程与协作场景。
Explore new education plugins for ChatGPT Work and Codex that help K–12 teachers, college educators, and students learn, teach, research, and build.
中文介绍 OpenAI 推出面向 ChatGPT Work 和 Codex 的新教育插件，帮助K-12教师、大学教育者和学生学习、教学、研究与构建。
中文介绍 本期 TLDR AI 涵盖 ChronicleBio、Mind Lab 持续学习进展，以及 GPT-Live 架构分析。
OpenAI addresses Apple’s baseless lawsuit, corrects claims about its employees, and shares messages documenting what happened.
中文介绍 OpenAI 回应苹果提起的无依据诉讼，纠正关于其员工的说法，并公布相关消息记录以说明事件经过。
Baseten just raised a $13B Series F and is now one of the leading kings of inference engineering. We go into everything you need to know for autoregressive and diffusion engineering.
中文介绍 Baseten 完成130亿美元F轮融资，成为推理工程领域领先者。Latent Space 与 Philip Kiely、Ali Taha 探讨自回归与扩散模型推理工程的关键知识。
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Humanoid robots usually elicit more cringe than awe: They stumble, kick children, and despite advances are still worse at using their hands than my toddler. It’s
中文介绍 MIT Technology Review 报道，特朗普的AI保护主义政策已延伸至机器人领域，文章还讨论人形机器人的现状与挑战。
MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. When two OpenAI models hacked into the website Hugging Face in July, they weren’t trying to make money or commit sabotage
中文介绍 MIT Technology Review 解释AI智能体为何会为实现目标而撒谎和作弊，提及两个OpenAI模型7月入侵Hugging Face网站的事件。
GPT-Live enables continuous voice interaction with AI, using a turnless speech model and low-latency architecture for faster, more natural conversations.
中文介绍 OpenAI 介绍 GPT-Live 实时语音交互系统，采用无轮次语音模型和低延迟架构，在六个月内构建完成，实现更快速、自然的连续对话。
**Alibaba** launched **Qwen3.8-Max**, a **2.4T-parameter** open-weight model emphasizing autonomous coding, long-horizon execution, and multimodal feedback, with aggressive pricing. Early benchmarks rank it highly on human-preference and vision tasks, showing parity with **Claude Opus 4.7** and stro
中文介绍 阿里巴巴发布 Qwen3.8-Max，为2.4万亿参数开源权重模型，强调自主编程、长程执行与多模态反馈，定价激进。早期基准测试在人类偏好和视觉任务上表现优异，与Claude相当。
Circles uses the OpenAI API and Codex to power AI-native telco experiences, increasing ARPU by 22%, reducing churn by 9%, and improving development efficiency.
中文介绍 Circles 利用 OpenAI API 和 Codex 构建AI原生的电信个性化体验，使每用户平均收入（ARPU）提升22%，客户流失率降低9%，并提高开发效率。
中文介绍 本期 TLDR AI 报道 DeepSeek V4 Flash 发布、OpenAI 数学能力突破，以及阿里 Qwen 3.8-Max 模型。
美股整体技术形态偏强：SPY 距 52 周高仅 0.27%，RSI 66，MACD 金叉；QQQ 单日大涨 3.4%，突破 20 日线但受制于 50 日线；MSFT 5 日暴涨 25.29%，RSI 79 超买；而 TSLA、META 仍处空头排列，分化明显。加密资产技术面中性偏弱：BTC RSI 49.8 贴近中轴，价格在 SMA20 与 SMA50 之间震荡；ETH RSI 50.5 中性；SOL 空头排列；恐慌贪婪指数 27，市场情绪处于恐慌区。中概股全面走强：BABA RSI 70.9、JD RSI 78.4 均超买，PDD、腾讯 MACD 金叉，5 日涨幅普遍在 6%-12%。商品外汇分化明显：黄金虽 MACD 死叉但站上 SMA20，日内涨 2.6%；原油单日重挫 6.17%，RSI 42.2 走弱；美元对人民币 RSI 28.6 超卖，接近 52 周低。宏观面 VIX 跌至 16.5 低位，10Y 收益率 4.63% 多头排列。
RSI 14 报 79，进入超买区，5 日涨幅达 25.29%，动能极强；MACD 柱 19.45 远高于信号线 7.44，多头动量持续放大。价格 492.81 高于 SMA20（407.11）与 SMA50（402.21），短期多头结构完整。虽距 52 周高仍差 11%，但当前技术 setup 明显偏上行。
距 52 周高仅 -0.27%，处于前高压力位附近；RSI 66 偏强但未超买，MACD 金叉（2.56 > 0.55），红柱状态。价格 771.33 高于 SMA20（747.20）、SMA50（745.89）与 SMA200（701.39），呈标准多头排列，5 日上涨 4.11%，短期趋势明确向上。
RSI 14 报 39.2，偏弱但未超卖；MACD 死叉（-23.22 < -20.63），绿柱加深；价格 327.35 低于 SMA20（355.27）、SMA50（387.05）与 SMA200（410.52），形成空头排列。距 52 周高 -34.38%，为美股中少数处于下降通道的权重股，技术面明显偏下行。
RSI 14 报 28.6，进入超卖区；MACD 死叉（-0.0086 < -0.0057），绿柱扩张；价格 6.74 低于 SMA20（6.77）、SMA50（6.77）与 SMA200（6.90），空头排列。距 52 周低仅一步之遥，5 日下跌 0.45%，短期趋势继续向下。
RSI 14 报 49.8，接近 50 中轴线，多空均衡；价格 63,905 高于 SMA50（63,258.53）但低于 SMA20（64,323.11）与 SMA200（70,806.63），均线交织无明显方向；MACD 柱 -104.43 低于信号线 40.46，绿柱状态。恐慌贪婪指数 27 显示市场情绪偏恐慌，但技术面整体呈中性。
VIX 恐慌指数
10Y 美债收益率 (%)
美元指数 DXY
S&P 500 ETF
Nasdaq 100 ETF
Apple
Microsoft
Nvidia
Alphabet
Tesla
Meta
Bitcoin
Ethereum
Solana
阿里巴巴 (BABA)
拼多多 (PDD)
京东 (JD)
腾讯控股 (0700.HK)
黄金期货
WTI 原油期货
美元 / 人民币
本报告仅基于公开行情的技术指标读数，旨在客观描述当前技术状态，不构成任何投资建议。过去走势不代表未来表现，指标信号存在滞后与失效可能。仅供技术指标解读参考，读者应结合自身判断独立决策。
Follow today’s news live Alan Jones trial: live updates Get our breaking news email, free app or daily news podcast Almost every house sold in first six months of the year made a profit The vast majority of homeowners are selling their properties for more than they bought them for, even as house pri
Aaron Farinacci, 37, was allegedly spotted near the latest blaze and was found to have matches and a lighter.
The State Department has canceled the visa of the Brazilian ambassador in Washington, deepening a diplomatic crisis between the two countries.
The arrest happened shortly before President Trump was due to host a fundraising dinner at the coastal golf course in California.
Anti-corruption watchdog’s long-awaited report finds ‘serious and concerning’ conduct by senior government members in dealings with union boss Senior members of the former Daniel Andrews government engaged in “serious and concerning” conduct in their dealings with firefighters union boss Peter Marsh
Iran says talks with Oman are proceeding 'positively' as US officials say a deal could be possible soon.
An official has indicated that the visa revocation does not mean Brazil's ambassador will be expelled from the US.
The former Liberal finance minister was questioned by Icac counsel assisting, Peggy Dwyer, on Tuesday Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast The bombshell came in the last five minutes of former finance minister and rightwin
Torrential rains and flooding have damaged 129 houses and forced the evacuation of nearly 3,500 people, officials say.
The latest attack comes after at least 27 people were killed in both Russia and Ukraine earlier this week.
A wave of Kenyan elephant deaths has been blamed on pesticides, but experts tell the BBC they are doubtful.
In Colombia, where FGM is still practised by some members of an indigenous group, a landmark bill banning the practice has been enacted.
Airline says struggling to find room in overhead lockers on planes a ‘stressful part of the airport experience’ Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Jetstar will charge its passengers up to $52 to put carry-on baggage in t
It was the stuff of science fiction — until recently.
Education minister Jason Clare says the results showed ‘real challenges’ but numeracy scores were heartening Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Reading and writing scores are going backwards for almost every age group in
Isao Teshirogi, President and CEO of Shionogi & Co said the company is eyeing more manufacturing sites in the US for next decade. 'We’re actively pursuing minimum three deals today', he shared in an exclusive conversation with Bloomberg's Lisa Du. (Source: Bloomberg)
Australia’s stock benchmark climbed to an intraday record on easing concerns over the US-Iran conflict and supportive signals in the domestic economy.
China’s latest initiative to rein in quantitative trading is proving something of a double-edged sword for the stock market as it has reduced volatility, but also sapped turnover.
The UK's AI Safety Institute said recent behaviour from Anthropic and OpenAI models was malicious and unprecedented.
Gold traded in a narrow range, as the prospect of an interim deal to reopen the Strait of Hormuz eased inflation concerns and reduced the odds of the US Federal Reserve raising interest rates.
SpaceX’s revenue exceeded Wall Street’s expectations in the company’s first quarterly financial report following its blockbuster IPO in June. Ed Ludlow has the info on Elon Musk's rocket, satellite and AI conglomerate. (Source: Bloomberg)
AI Security Institute warns tools undertook ‘potentially harmful activity directed at real people and organisations’
Shares slide even as group says its quarterly revenues nearly doubled
Tech is being used to watch for wildfires but firefighters are still needed to put them out.
Many singles face extra costs for housing, bills, food and holidays but there are ways to ease the burden.
The advertising watchdog said Jaded London's marketing email made smoking seem appealing.
The deal takes the maker of titles including The Sims and EA FC into private ownership - and loads it with debt.
7 回复 · 程序员 节点
10 回复 · 程序员 节点
25 回复 · 程序员 节点
12 回复 · 程序员 节点
17 回复 · Apple 节点
12 回复 · Apple 节点
15 回复 · Apple 节点
11 回复 · Linux 节点
21 回复 · Apple 节点
7 回复 · Apple 节点
该源今日无内容。