TypeScript · ★ 448,529 · 🍴 45,035 · 📈 640 stars today
freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free.
中文介绍 freeCodeCamp 的开源代码库和课程，提供免费学习数学、编程和计算机科学的交互式平台。使用 JavaScript 等 Web 技术构建，适合自学者、教师及转行者系统掌握技能。
Rust · ★ 33,962 · 🍴 1,417 · 📈 21 stars today
Rust-based platform for the Web
中文介绍 基于 Rust 编写的 JavaScript/TypeScript 编译器与打包工具，可替代 Babel 并大幅提升构建速度。适用于前端工程化场景，为现代 Web 应用提供高性能编译支持。
Elixir · ★ 8,396 · 🍴 956 · 📈 214 stars today
A self-hosted data logger for your Tesla 🚘 [main maintainer=@JakobLichterfeld]
中文介绍 自托管的 Tesla 数据记录器，用于追踪车辆充电、里程、电池健康等统计数据。基于 Elixir/Phoenix 技术栈，适合 Tesla 车主自行部署以精细化监控爱车状态。
TypeScript · ★ 124,017 · 🍴 6,691 · 📈 1,196 stars today
Collection of publicly available IPTV channels from all over the world
中文介绍 全球公开 IPTV 频道集合，提供大量 M3U 格式播放列表。用户可免费获取来自世界各地的直播流媒体资源，适用于 IPTV 播放器和爱好者的便捷接入。
TypeScript · ★ 94,875 · 🍴 9,450 · 📈 80 stars today
JavaScript API for Chrome and Firefox
中文介绍 通过 JavaScript API 控制 Chrome/Firefox 的无头浏览器库，常用于自动化测试、网页截图与爬虫。基于 Node.js，为前端开发和 QA 团队提供可靠的浏览器自动化能力。
TypeScript · ★ 10,836 · 🍴 3,448 · 📈 229 stars today
Meshery, the cloud native manager
中文介绍 云原生管理平台，支持多种服务网格（如 Istio、Linkerd）和 Kubernetes 的配置、性能测试与可视化。适用于 DevOps 和 SRE 团队统一管理微服务基础设施。
TypeScript · ★ 50,196 · 🍴 3,426 · 📈 11 stars today
Fast, easy and reliable testing for anything that runs in a browser.
中文介绍 快速、可靠的前端端到端测试框架，直接运行在浏览器中。提供实时重载、时间旅行调试等特性，适合前端开发者编写稳定高效的自动化测试用例。
Python · ★ 2,556 · 🍴 450 · 📈 157 stars today
Music Assistant is a free, opensource Media library manager that connects to your streaming services and a wide range of connected speakers. The server is the beating heart, the core of Music Assistant and must run on an always-on device like a Raspberry Pi, a NAS or an Intel NUC or alike.
中文介绍 开源媒体库管理器，整合 Spotify、Qobuz 等流媒体服务与多品牌智能音箱（如 Sonos）。服务器端统一管理和同步音乐资源，适合家庭音频爱好者打造集中音乐系统。
Rust · ★ 7,281 · 🍴 314 · 📈 146 stars today
Cross-platform GUI written in Rust using ADB to debloat non-rooted Android devices. Improve your privacy, the security and battery life of your device.
中文介绍 跨平台 GUI 工具，通过 ADB 在非 root Android 设备上卸载预装应用，提升隐私、安全性和续航。使用 Rust 编写，适合普通用户安全移除系统臃肿软件。
Python · ★ 30,116 · 🍴 3,406 · 📈 413 stars today
VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning
中文介绍 无分词器的多语言语音合成模型，支持创意语音设计和逼真声音克隆。基于深度学习，适合语音助手、内容创作及个性化 TTS 场景的开发者使用。
C++ · ★ 10,436 · 🍴 605 · 📈 188 stars today
A lightweight, lightning-fast, in-process vector database
中文介绍 轻量级、极快的进程内向量数据库，专为 AI 向量检索优化。以嵌入式方式运行，适合需要实时相似性搜索的推荐、图像或文本检索应用。
TypeScript · ★ 9,067 · 🍴 1,997 · 📈 185 stars today
Free, Open Source, Self-Hosted WhatsApp API Gateway
中文介绍 免费开源、可自托管的 WhatsApp API 网关，提供消息收发和群组管理接口。开发者可借此构建 WhatsApp 机器人或自动化工作流，无需依赖商业中间件。
Rust · ★ 9,270 · 🍴 435 · 📈 326 stars today
IP addresses break, dial keys instead. Modular networking stack in Rust.
中文介绍 基于公钥的模块化网络栈，用键替代 IP 地址实现设备间直接通信。使用 Rust 编写，适合构建去中心化应用（如 P2P 文件分享、隐私通信）的底层网络层。
👍 157
Many moments in the real world do not wait for a user to ask. A fire starts on a security monitor, an expression flickers across a video call, or a product a viewer wants flashes by in a livestream. Yet today's large models remain mostly turn-based by design: they answer only when addressed, and eve
中文介绍 提出JoyAI-VL-Interaction，实现实时视觉-语言交互智能，无需用户主动发问即可响应监控起火、表情变化等连续事件，打破回合制延迟，提升交互即时性。
👍 100
Data tells stories that shape society; the data journalist's job is to turn raw information into stories non-experts can trust. A high-quality news feature takes a newsroom team weeks: hunting for context, running statistics, choosing an angle, and designing visuals. Recent agents handle individual
中文介绍 提出Data Journalist Agent，端到端将原始数据转化为可验证的多模态新闻故事，整合语境搜索、统计分析、角度选择和视觉设计，大幅缩短传统数周的制作周期。
👍 85
Generalist robot policies must follow user instructions while reasoning about how objects, cameras, and robot actions interact in the 3D physical world. Recent vision-language-action models (VLAs) and video world-action models (WAMs) inherit strong semantic or temporal priors from large-scale founda
中文介绍 针对通用机器人策略缺乏3D几何推理的问题，提出Geometric Action Model，显式建模物体、相机与动作的几何关系，提升指令遵循和物理交互能力。
👍 78
DreamX-World 1.0 is a general-purpose interactive text/image-to-video world model for controllable long-horizon generation. It supports camera navigation, revisits to previously observed regions, and promptable events across photorealistic, game-style, and stylized domains. Our data engine combines
中文介绍 DreamX-World 1.0是通用的交互式世界模型，支持文本/图像到视频的长时可控生成，涵盖相机导航、区域重访和提示事件，适用于逼真、游戏和风格化多种领域。
👍 76
Large Language Model (LLM) coding agents have achieved strong results on software engineering tasks, yet repository exploration remains a major bottleneck: locating relevant code consumes substantial token budget and pollutes the agent's context with irrelevant snippets. In most agents, the same mod
中文介绍 解决代码代理仓库探索消耗大量token的问题，提出FastContext，训练专门的仓库探索器高效定位相关代码，减少上下文污染和预算开销，提升代理效率。
👍 62
This technical report introduces VibeThinker-3B, a compact dense model with 3B parameters developed to investigate how far verifiable reasoning can be pushed within a strictly small-model regime. Building upon the Spectrum-to-Signal post-training paradigm, we systematically enhance the model through
中文介绍 VibeThinker-3B探索3B小模型的可验证推理极限，基于Spectrum-to-Signal后训练范式系统增强推理能力，在严格小模型框架下取得显著进步。
👍 23
Masked Diffusion Language Models (MDLMs) have emerged as a distinct paradigm for sequence generation. As MDLMs become diverse in capabilities and knowledge coverage, an important question is how to combine their knowledge. Toward this, we first investigate the unique decoding dynamics of MDLMs. We f
中文介绍 研究掩码扩散语言模型（MDLM）的集成策略，通过分析解码动态并追踪可靠轨迹，提升多MDLM组合生成的质量与一致性。
👍 21
Vision language models are serving as general-purpose interfaces for complex multimodal tasks. However, deployment still faces three gaps: VLMs typically incur high latency and cost when processing dense video frames and long prompts, the agent scaffold remains static after deployment, and standard
中文介绍 VisualClaw实现实时个性化物理世界代理，通过优化密集视频帧和长提示处理降低延迟，并支持动态更新scaffold，弥补VLM部署中的实时性、成本和灵活性差距。
👍 14
Multi-task learning (MTL) is essential in recommender systems to enable complementary learning among diverse user feedback. While modern industrial practices have shifted from DNNs to Transformer-centric architectures to strengthen sequence modeling and scaling capacity, they still decouple feature
中文介绍 提出OneRank统一Transformer原生的多任务推荐排序架构，端到端联合建模特征表示与排序，解决传统解耦导致的性能损失，提升互补学习效果。
👍 14
Visual world models (VWMs) synthesize interactive, action-conditioned rollouts from a single context image. However, it remains an open question how robust these models are to adversarial perturbations. Standard adversarial attacks fail to assess this vulnerability because attackers lack ground-trut
中文介绍 BadWorld对视觉世界模型（VWM）进行对抗攻击，设计适用于无真实值场景的攻击方法，揭示VWM在行动条件化 rollout 中的脆弱性。
👍 14
We introduce Qwen-RobotWorld, a language-conditioned video world model for embodied intelligence. With natural language as a unified action interface, it predicts physically grounded future visual trajectories from current observations across robotic manipulation, autonomous driving, indoor navigati
中文介绍 Qwen-RobotWorld以自然语言为统一动作接口，通过语言条件视频生成预测物理世界未来轨迹，统一机器人操控、自动驾驶和室内导航等多具身任务。
👍 13
As LLM agents are deployed in long-horizon sessions, context accumulation drives up inference costs. Existing approaches utilize text pruning or dynamic memory eviction to minimize token footprints; however, their unconstrained sequence mutations alter layouts, introducing prefix mismatches and cach
中文介绍 TokenPilot提出缓存高效的上下文管理方法，避免序列突变导致的前缀不匹配和缓存失效，降低LLM代理长期会话中的token开销和推理成本。
👍 13
Modern Lean theorem provers achieve strong performance only with substantial training and inference compute, driven in part by scarce verified proof data and the long reasoning traces of formal proof search, making both supervised fine-tuning (SFT) and sampling expensive. We introduce Pythagoras-Pro
中文介绍 Pythagoras-Prover通过增强Lean形式化（数据增强和高效搜索）加速形式证明，减少对大量核实数据的依赖，降低训练和推理计算开销。
👍 11
Advanced agents are increasingly demonstrating the potential to operate as autonomous engineers, creating a growing demand for evaluation benchmarks that capture the complexity of real-world development. Such environments typically involve both complex code and large-scale data (i.e., file system).
中文介绍 提出CODA-BENCH基准，评估代码代理在包含复杂代码和大规模数据的真实环境中的表现，揭示数据密集型任务中的能力瓶颈。
👍 10
Web agents act through long interaction sequences, yet existing benchmarks evaluate only terminal success, discarding all process information and offering little guidance on improvement. In this work, we conduct a process-level analysis of web agents. We introduce WebStep, a benchmark of 1,800 task
中文介绍 WebStep基准引入语义状态追踪进行过程级评估，包含1800任务，分析web代理每个步骤的正确性，弥补只评估最终成功率的不足。
👍 9
Efficient and scalable agentic intelligence requires models that can deliver both low-latency responses and strong reasoning capabilities while remaining practical to train, serve, and deploy. In this report, we present Ling-2.6 and Ring-2.6, a family of models designed to address this challenge at
中文介绍 Ling-2.6和Ring-2.6模型族在万亿参数规模实现低延迟与强推理的平衡，通过优化训练和服务流程提供即时代理性智能。
👍 9
As LLMs advance, post-training reinforcement learning (RL) increasingly relies on multi-dimensional rewards to cultivate comprehensive capabilities. This shift demands new algorithms capable of optimizing diverse and potentially competing objectives simultaneously. To address this, existing methods
中文介绍 GD^2PO提出群体动态奖励解耦策略优化，将多目标奖励分组解耦，缓解LLM后训练中多奖励冲突，综合提升模型能力。
👍 8
Phone agents are increasingly expected to complete real mobile workflows rather than merely predict the next screen action. However, much of the current mobile-agent literature still evaluates agents primarily as GUI controllers that observe a screen, emit taps and swipes, and are scored by target a
中文介绍 PhoneHarness融合GUI、CLI和工具动作，构建更真实的手机代理框架，超越纯GUI控制，提升复杂移动工作流的完成效率。
👍 8
Multi-turn LLM serving accumulates dialogue history whose Key-Value (KV) cache grows with every turn and every user, quickly exceeding the model weights themselves and making memory -- not compute -- the binding constraint on throughput. Non-uniform KV compression, which allocates heterogeneous budg
中文介绍 Tangram通过非均匀KV缓存压缩，根据重要性分配异质预算，显著减少多轮LLM服务中的内存占用，提高吞吐量。
👍 8
We introduce Nemotron 3 Ultra, a 550 billion total and 55 billion active parameter Mixture-of-Experts Hybrid Mamba-Attention language model. We pre-trained Nemotron 3 Ultra on 20 trillion text tokens, then extended the context length to 1M tokens, and post-trained using Supervised Fine Tuning (SFT),
中文介绍 Nemotron 3 Ultra是550B总参数/55B活跃参数的MoE混合Mamba-Transformer模型，预训练20T tokens并扩展至1M上下文，经SFT和RL后训练，专为代理推理优化。
@satyanadella · 5.9M 粉丝 · 1.9M 阅 · 3.5K 赞 · 541 转
I’ve been thinking a lot about the future of the firm in an AI-driven economy. This transition is different than any previous platform shift. In the past, we used digital systems to enhance human
中文介绍 Satya Nadella 探讨 AI 驱动经济下企业未来，认为当前转型不同于以往平台迁移——过去数字系统增强人类，AI 时代需要新生态，单靠前沿技术不稳定，必须构建生态系统。
@eng_khairallah1 · 67.2K 粉丝 · 1.2M 阅 · 510 赞 · 73 转
You read maybe two hundred articles this year. A few dozen papers. Hundreds of threads. Save this Every second-brain method ever sold to you, Zettelkasten, PARA, the graph view, the daily note,
中文介绍 用 Obsidian 搭建自动运行的「第二大脑」完整教程，覆盖 Zettelkasten、PARA、图谱视图、每日笔记等方法，告别手动管理，实现知识库自运行。
@sairahul1 · 116.6K 粉丝 · 905.5K 阅 · 503 赞 · 102 转
Everyone talks about LLMs. Nobody explains how they actually work under the hood. GPT. Claude. Gemini. Llama. They all come from the same 5-stage pipeline. And once you understand it, you can build
中文介绍 揭秘 LLM 底层 5 阶段流水线（数据收集、预训练、微调、对齐、部署），解释 GPT/Claude/Gemini/Llama 的共同架构，助你从零构建大模型。
@lqiao · 95.4K 粉丝 · 501.6K 阅 · 509 赞 · 86 转
Mythos got shut down this week. Whether you agreed with the decision or not is almost beside the point. A company built on top of intelligence it didn't control suddenly found itself exposed to
中文介绍 以 Mythos 被关停事件为例，讨论「租用智能」vs「拥有智能」的风险——依赖第三方 AI 能力的公司可能突然暴露于不可控因素，建议考虑自主权。
@Fintech03 · 34.9K 粉丝 · 438.8K 阅 · 501 赞 · 137 转
The critique that Indian IT services (the classic TCS, Infosys, Wipro, HCL cohort) are failing India in the current generative AI race is an incredibly popular talking point. It is easy to look at the
中文介绍 反驳「印度IT在生成式AI竞赛中失败」的观点，指出 TCS/Infosys/Wipro/HCL 等公司本就不以构建 ChatGPT 为目标，其定位是服务而非前沿创新。
@mikenevermiss · 10.8K 粉丝 · 261.4K 阅 · 568 赞 · 67 转
stop making prompts. start designing loops. a prompt gets you one response. a loop gets you a system that keeps working after you close the laptop. Boris Cherny, who runs Claude Code at Anthropic, put
中文介绍 从「写提示」转向「设计循环」——Boris Cherny（Claude Code 负责人）分享如何用 Claude 构建持续运行的自动化系统，而非一次性单次响应。
@AnthonyNAguirre · 4.2K 粉丝 · 232.1K 阅 · 527 赞 · 38 转
I've been getting really bothered lately by something subtle but quite dangerous that I believe is currently going on and promises to get worse: Large numbers of very smart, capable, and important
中文介绍 警惕一种隐晦却危险的趋势：大量聪明、有能力的重要人物正被某种思想潜移默化地「洗脑」，且情况可能恶化。未提具体工具，偏警示性观点。
@athcanft · 19.1K 粉丝 · 205.6K 阅 · 514 赞 · 23 转
I've been mass-producing TikTok slideshows with AI and scheduling them weeks in advance. Zero filming. Zero editing. Zero daily posting grind. This article breaks down the exact system, step-by-step,
中文介绍 用 AI 批量生产 TikTok 幻灯片视频，每月 600 条仅需 2 美元成本。零拍摄零剪辑，提前几周排期发布，整套自动化流程详解。
@matanSF · 20.2K 粉丝 · 123.1K 阅 · 529 赞 · 60 转
In 2023, we launched Factory with the mission to bring autonomy to software engineering. While others were using models to speed up coding, we set out to deploy autonomous Droids across the
中文介绍 Factory 2.0 从单一代码代理升级为软件工厂——2023 年启动，目标是在整个软件工程中部署自主 Droid，实现端到端自动化，而非仅加速编码。
@jxmnop · 50.7K 粉丝 · 114.1K 阅 · 504 赞 · 57 转
So you want to do AI research? It's true that no one really teaches you how. Not directly, anyway. But it turns out that the way to get started is pretty simple: some combination of (i) reading and
中文介绍 AI 研究入门指南：没人直接教，但方法很简单——阅读论文 + 动手实践。鼓励通过阅读和编码结合的方式开启研究之路。
@AndrewCurran_ · 53.9K 粉丝 · 62.8K 阅 · 569 赞 · 61 转
If you used Fable while it was available, you know it is special in ways that will not show up on benchmarks. I post benchmarks all the time because they matter to many people, but for a long time
中文介绍 讨论 Fable 工具被关闭的事件——它虽在基准测试中表现不突出，却有独特价值。暗示 AI 工具窗口期已过，用户应关注不可量化的特性。
@addyosmani · 400.1K 粉丝 · 52.9K 阅 · 522 赞 · 41 转
Coding agents are extraordinarily good now and getting better fast. The interesting consequence is that the hard part of engineering moved from writing code to deciding whether to trust it, which
中文介绍 编码代理越来越强，工程难点从「写代码」变成了「是否信任代码」。Agentic Code Review 应运而生——核心是验证而非编写，信任决策成为新瓶颈。
@unicity_labs · 125.6K 粉丝 · 47.6K 阅 · 534 赞 · 659 转
Programmatic agent usage moves onto a separate metered credit today. Here is what changes, and the controls we are building into AstridOS to manage it. What changed today Today, 15 June 2026,
中文介绍 Claude 于 2026 年 6 月 15 日将 Agent SDK 编程使用单独计费。AstridOS 正构建管控功能，对 Agent 构建者意味着成本结构变化，需适应新的计量信用模式。
@zaimiri · 54.4K 粉丝 · 46.9K 阅 · 508 赞 · 48 转
Most people try to build their AI setup in one chaotic weekend. They install ten tools. Connect five APIs. Create a few automations. Add a giant system prompt. Then wonder why the whole thing feels
中文介绍 避免杂乱的周末搭建，推荐 7 天 Hermes 设置指南——逐步安装工具、连接 API、创建自动化，避免大型系统提示导致混乱，确保 AI 系统稳定运行。
@samueljmcd · 857 粉丝 · 42.2K 阅 · 511 赞 · 52 转
Loop engineering is the new label. The hard part is the one it has always been. Verification. Tip: You can copy and paste this article into Claude and ask for the best insights if you don't want to
中文介绍 Loop Engineering（循环工程）新概念，但核心难题依然是验证。建议将文章复制到 Claude 提取关键洞见，聚焦循环系统的可靠性和输出质量。
@Dimillian · 51.8K 粉丝 · 38.2K 阅 · 558 赞 · 37 转
How to turn your phone into a Codex control center?? At first glance, it looks like a way to check on a Codex task from your phone. That is useful, but it misses the bigger idea. The power of Codex
中文介绍 将手机变成 Codex 控制中心——不仅是查看任务进度，更是利用移动端随时管理编码代理、监控运行状态，最大化 Codex 的工程效率。
UK government partners with Google DeepMind to build a new AI-powered prototype aimed at faster housing decisions.
中文介绍 英国政府与谷歌DeepMind合作，打造AI驱动的规划原型，旨在加快住房建设决策速度。
At the end of a tense and scoreless first half of a soccer match between the English men’s team and rival Germany, millions of Brits let out a collective sigh and did what they so often do in moments of stress: They made tea. That wave of electric kettles clicking on, however, caused a different…
中文介绍 研究人员指出数据中心快速上线需具备电网灵活性，例如英国球迷同时烧水沏茶造成的瞬时用电高峰凸显柔性需求。
a quiet day lets us report on Satya's hit essay
中文介绍 AI新闻简报报道萨提亚关于构建前沿生态系统的文章《Loopcraft》。
OpenAI introduces Deployment Simulation, a method to predict AI model behavior before deployment using real conversation data to improve safety and evaluation accuracy.
中文介绍 OpenAI推出部署模拟方法，使用真实对话数据在部署前预测AI模型行为，以提升安全性和评估准确性。
中文介绍 今日AI要闻：Meta推出AI模式、Factory 2.0、Sakana的自主研究员等。
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. When I landed in Seoul after a grueling 12-hour flight from San Francisco, I walked through an unmanned immigration checkpoint, where a machine scanned my face an
中文介绍 作者在首尔机场体验无人检查站，探讨韩国为何如此热衷AI技术。
中文介绍 AI简报：Anthropic关闭Fable项目、智谱发布GLM-5.2、OpenRouter推出Fusion功能。
OpenAI launches the Partner Network, investing $150M to help global partners accelerate enterprise AI adoption, deployment, and transformation.
中文介绍 OpenAI推出合作伙伴网络，投资1.5亿美元，帮助全球合作伙伴加速企业AI的采用、部署和转型。
We are in the strangest timeline.
中文介绍 AI新闻：Fable和Mythos被官方认为过于危险，不予发布。
中文介绍 Allen AI推出olmo-eval，一个用于模型开发循环的评估工作台。
OpenAI introduces three Academy courses that help people build practical AI skills, create repeatable workflows, and apply agents in everyday work.
中文介绍 OpenAI推出三门学院课程，帮助用户掌握实用AI技能、构建可重复工作流，并在日常工作中应用智能体。
a quiet day lets us highlight a great concept from Peter Steinberger, Boris Cherny, and Andrej Karpathy
中文介绍 AI新闻简报介绍Loopcraft概念，来自Peter Steinberger、Boris Cherny和Andrej Karpathy，探讨循环堆叠的艺术。
Preply uses OpenAI to launch AI-generated lesson summaries, providing personalised feedback and language learning exercises.
中文介绍 Preply使用OpenAI推出AI生成的课程摘要，结合人类教师，提供个性化反馈和语言学习练习。
中文介绍 AI简报：OpenAI收购Ona、Anthropic撤回声明、小米发布MiMo代码。
Google DeepMind is funding research into the potential dangers of situations where millions of different AI agents interact with each other online. According to Rohin Shah, who directs the company’s AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without
中文介绍 谷歌DeepMind资助研究，担忧数百万AI代理在线交互可能带来的危险，其AGI安全总监Rohin Shah指出需谨慎应对。
当前加密市场处于极度恐慌状态（恐慌贪婪指数23），加密总市值2.34T，BTC主导率56.3%，ETH占比9.3%。美股方面，SPY/QQQ虽保持多头排列，但MACD柱均为负值，短期上涨动量减弱；个股分化明显，MSFT、META等科技股呈空头排列，RSI回落至40-48区间，下行压力增大。中概股集体偏弱，BABA、PDD等均处于空头排列，BABA RSI 27.9进入超卖区，短期延续弱势。加密货币出现局部反弹信号，BTC、ETH、SOL均触发MACD金叉，其中SOL近5日涨幅达10.01%，但仍受50/200日均线压制。商品方面，WTI原油暴跌14%，RSI 29.9超卖；黄金弱势震荡。美元/人民币接近52周低点，空头排列，人民币升值延续。VIX恐慌指数下行至16.41，空头排列示市场恐慌情绪缓解。整体来看，多数资产技术面偏弱，但加密货币金叉反弹信号值得关注。
当前价750.33，高于SMA20 746.44、SMA50 726.61、SMA200 687.35，形成多头排列；RSI 57处于正常区间；MACD柱-1.70为负，短期动量偏弱；距52周高仅-1.32%，接近高位但无突破动作。综合判断中性。
当前价393.83，低于SMA20 417.16、SMA50 412.74、SMA200 452.62，呈空头排列；RSI 40.2接近超卖区但未进入；MACD柱-5.51且信号线-1.04，负值扩大，趋势bearish；距52周高-29.1%，下行压力显著。
当前价73.51，站上SMA20 71.62；RSI 49中性；MACD柱+1.16且出现金叉信号；近5日涨幅10.01%反弹强劲；但仍低于SMA50 80.97与SMA200 99.02，整体空头排列未扭转，短期动能转强。
当前价110.97，远低于SMA20 123.69、SMA50 129.91、SMA200 149.48，空头排列；RSI 27.9进入超卖区；MACD柱-5.54，负值持续；距52周高-42.4%；近5日跌幅-7.29%，弱势延续。
当前价75.8，低于SMA20 91.1与SMA50 95.39，略高于SMA200 73.55；RSI 29.9超卖；MACD柱-4.07，负值；近5日跌幅-14.06%，单日-6.13%，暴跌后超卖但趋势依然向下。
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
过去走势不代表未来表现，以上内容仅基于技术指标读数进行客观描述，不构成任何投资建议。市场存在不确定性与风险，仅供技术指标解读参考。
The pain from their injuries hasn’t gone away. Meanwhile, they face challenges in earning a livelihood.
中文摘要 缅甸地雷爆炸幸存者面临肢体截肢和生活破碎的困境，伤痛未消，同时谋生困难。
Follow today’s news live Get our breaking news email, free app or daily news podcast Kyle Sandilands has settled his legal dispute with ARN Media with a $12m cash payment over three years, but Jackie O Henderson’s $82m claim against the broadcaster will continue. The shock jock agreed to drop his $8
中文摘要 两名男子因克里斯·巴格萨里安绑架案被控谋杀；工党承诺投入36亿澳元提高幼教工作者薪资。凯尔·桑迪兰兹以1200万澳元现金和解与ARN的纠纷。
In newly unsealed court filings, the government says the group allegedly expressed grievances about corruption, the Epstein files, and data centres.
中文摘要 FBI称，一个团体计划在白宫UFC活动期间使用狙击手和无人机发动袭击，该团体曾表达对腐败、爱泼斯坦文件和数据中心的不满。
Iran’s foreign minister said new talks on nuclear arms and sanctions relief would start on Friday, right after an initial agreement is signed.
中文摘要 特朗普表示希望将战争置于“后视镜”中。伊朗外长称，在初步协议签署后，新核武器和制裁解除谈判将于周五开始。
ARN Media settlement of dispute with sacked radio host includes $12.09m cash payment and $1.5m worth of advertising Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Kyle Sandilands has settled his dispute with ARN Media for a $12m cas
中文摘要 电台主持人凯尔·桑迪兰兹与前东家ARN Media达成和解，获1209万澳元现金及价值150万澳元的广告支持，并帮助其新媒体创业。
On a roll or against a wall, Group of 7 leaders bring sharply different agendas. The leaders of some other nations are also attending to press their own interests.
中文摘要 G7峰会在法国举行，七国领导人各有不同议程，部分其他国家领导人也与会推进自身利益。
US placed tariffs on Brazilian goods and sanctioned judicial officials involved in trial of the father for coup plot.
中文摘要 巴西法院判定爱德华多·博索纳罗寻求美国干预其父亲政变审判有罪。美国对巴西商品加征关税并制裁参与审判的司法官员。
Lyles clocks 14.67 seconds, edges Dambile of South Africa for gold while Australian teenager Gout Gout finishes third.
中文摘要 美国短跑名将诺亚·莱尔斯在黄金大奖赛上以14.67秒跑出150米世界最佳成绩，击败南非选手丹比尔夺冠，澳大利亚少年高特·高特获第三。
The Russian frigate fired warning shots near a British retired couple on a yacht in the English Channel.
中文摘要 一对英国退休夫妇在英吉利海峡的游艇上遭遇俄罗斯护卫舰，俄舰发射警告射击。他们表示曾试图展示改变航向，但射击已发生。
Somaliland’s President is on a ‘historic’ tour of Israel, where he’s opened an embassy and visited the Knesset.
中文摘要 索马里兰总统开启对以色列的“历史性”访问，开设大使馆并访问以色列议会。
Brazil supreme court finds that Eduardo Bolsonaro – who resides in the US - tried to get sanctions put on judges trying ex-president over coup plot Brazil’s supreme court has sentenced Eduardo Bolsonaro to four years and two months in prison after finding him guilty of courting US interference in h
中文摘要 巴西最高法院判处前总统博索纳罗之子爱德华多·博索纳罗四年两个月监禁，因其试图寻求美国制裁审判其父亲政变案的法官。
The retired couple tell BBC Newsnight they tried to show the warship they had changed course before the shots were fired.
中文摘要 英国正在调查有关俄罗斯军舰在英吉利海峡向一艘游艇发射警告射击的报告。该游艇上的退休夫妇称他们曾试图展示已改变航向。
A Labour Party deputy who spent years in opposition, he criticized Conservatives and members of his own party, and was at the heart of major political moments.
中文摘要 英国工党前副领袖罗伊·哈特斯利去世，享年93岁。他长期在野，既批评保守党也批评本党成员，是英国政治核心人物。
This live blog is now closed. You can follow all the latest developments from the G7 summit in our Europe live blog: We will be including any Iran-related news from the summit in our Middle East crisis live blog. Continue reading...
中文摘要 美国与伊朗达成结束战争的协议，允许德黑兰出售石油和燃料。相关进展也在G7峰会讨论。
There’s an initial deal between the U.S. and Iran. But bringing the global economy back online after months at reduced speed isn’t going to be quick or easy.
中文摘要 美伊达成初步协议，但全球经济在数月低速运转后复苏之路漫长且艰难。
The US and Iran are preparing to formally sign an interim peace deal, with details of the accord still emerging and energy insiders skeptical of how fast it can reopen the Strait of Hormuz. Bloomberg Economics' Chris Kennedy breaks down the latest developments. (Source: Bloomberg)
中文摘要 美国与伊朗即将签署临时和平协议，细节仍在披露中。能源业内人士对霍尔木兹海峡重开速度持怀疑态度。彭博经济分析师Chris Kennedy解读最新进展。
Mondelez boss Dirk Van de Put says it was the "right decision" to remain after the war with Ukraine.
中文摘要 亿滋国际CEO Dirk Van de Put称，俄乌冲突后继续留在俄罗斯是“正确决定”。
Chinese stocks listed in Hong Kong are facing bleak milestones as a global rush into artificial intelligence supply chain players sidelines the Internet and consumer companies that dominate the offshore benchmark.
中文摘要 香港上市中资股承压，全球资金涌入人工智能供应链板块，互联网和消费股遭冷落，恒生中国股指下跌。
Slowing sales, a slump in its share price and a selloff in the gold market are fueling concerns that Laopu Gold Co., one of China’s hottest consumer brands, is losing momentum.
中文摘要 老铺黄金销售放缓、金价下跌及股价回落，引发市场对其增长势头减弱的担忧。
Former employee claims in lawsuit that she identified issues with risk-management practices
中文摘要 前花旗高管起诉银行，称因提出对特朗普的担忧而被解雇，指控花旗在风险管理方面存在问题。
UK says incident is not linked to Britain’s interception of an oil tanker that is part of Moscow’s shadow fleet
中文摘要 俄罗斯军舰在英吉利海峡向一艘游艇鸣枪示警。英方称此事件与拦截俄“影子舰队”油轮无关。
Kalshi Inc. is looking to expand its new perpetual futures business beyond digital assets after the never-expiring derivatives racked up more than $5.5 billion of trading volume in their first two weeks on the prediction-market platform.
中文摘要 预测市场平台Kalshi Inc.在推出永续期货业务两周内交易量超55亿美元，计划将该业务扩展至数字资产之外。
Csquare Inc., a data center company backed by Brookfield Corp., filed publicly for a US initial public offering, joining a rush of debuts linked to artificial intelligence infrastructure.
中文摘要 Brookfield支持的数据中心公司Csquare Inc.提交美国IPO申请，赶上人工智能基础设施上市热潮。
Stocks in Asia looked set to snap a three-day winning run, tracking declines on Wall Street as investors rotated out of technology shares and positioned for the first Federal Reserve policy decision under Chairman Kevin Warsh.
中文摘要 亚洲股市料结束三日连涨，美联储新主席凯文·沃什首次政策会议前投资者减持科技股。
Oil held near a three-month low on expectations that a US-Iran deal to reopen the Strait of Hormuz will unleash a wave of supply.
中文摘要 油价持于三个月低位附近，市场预期美伊协议将重开霍尔木兹海峡并增加供应。
Chief Evan Spiegel describes $2,195 eyewear as his ‘life’s work’ as he faces increasing criticism over costs of project
中文摘要 Snap发布售价2195美元的“Specs”智能眼镜，CEO埃文·斯皮格尔称其为“毕生之作”，但面临成本压力批评。
The decision comes after a prolonged period of difficulty for the chain, which has faced increasing competition from a range of rivals.
中文摘要 陷入困境的必胜客连锁店以27亿美元出售，此前长期面临激烈竞争。
15 回复 · 程序员 节点
31 回复 · 程序员 节点
9 回复 · Linux 节点
6 回复 · 程序员 节点
7 回复 · Linux 节点
14 回复 · Apple 节点
30 回复 · Apple 节点
30 回复 · Apple 节点
6 回复 · Apple 节点
33 回复 · Apple 节点
本帖使用社区公益推广，符合推广要求。我申明并遵循社区要求的以下内容： 我的项目是免费使用的，无收费（变相收费、赞助）部分： 是 我的帖子已经打上 公益推广 标签： 是 我的项目属于个人项目，与公司或商业机构无关： 是 我的项目不存在QQ、TG等群组引流： 是 我的项目不存在非运营必要的网站引流： 是 我的项目不存在为他人推广、AFF： 是 我的项目无关联的商业项目： 是 我的站点存在登录，并已接入 LINUX DO Connect： 是 我帖子内的项目介绍，AI生成、润色内容部分已截图发出： 是 以上选择我承诺是永久有效的，接受社区和佬友监督： 是 以下为项目介绍正文内容，AI生成、润色内容已
huggingface.co zai-org/GLM-5.2 · Hugging Face We’re on a journey to advance and democratize artificial intelligence through open source and open science. https://z.ai/blog/glm-5.2 20 个帖子 - 20 位参与者 阅读完整话题
首先是好久没听到前端大哥的我丢，丢你老母、顶你个肺了，今天听了好久，我以为又用OPUS了，想多了，用的GPT 5.5 HIGH，下午换成人家自己买的KIMI K2.7 CODE，BUG修了，真的无语 再来说说游戏的BUG修复，无了个大语，先是一直 然后费劲巴拉的跑了1个小时： 问题没找到，也没修复，真无语，一直在思考，从未解决任何问题，甚至都没找到问题 然后DeepSeek V4 仅用9分33秒完成，一把搞定（本来也没多复杂啊）： 无了个大语，今天我们都没怎么用GPT 5.5，没法用了，傻的流黄汤 到底啥情况啊，我的小聪明GPT 5.5呢？还给我啊 一下午，DeepSeek V4/GLM 5.
我之前说过CHY公益站的API可以分发，但是感觉有佬友是在测活啊（大量无意义文本）佬友们给我支个招呗 54 个帖子 - 38 位参与者 阅读完整话题
阿里 阿里国际 面着面着hr没了，哈哈哈三面过了之后hr约了一次面试然后取消了，好像最近阿里国际在大裁员 一面 记忆权威？低权威的记忆如何晋级？比如你这里聊天里的消息有可能变成类似于架构决策嘛？ 记忆的治理？如何做记忆的合并？比如出现了冲突怎么办？ multi-agent到底是如何决策的？ 比如你的猫猫这里突然拉一个新的agent进来他们为什么会有这样的决策？ 你没有用强状态机，如何保证任务的流转不偏移？ 你这个项目是开源了吗？简历上好像没看到你的开源地址？ 你的multi-agent的架构设计和claude code的有什么不同？ 你这里说你的核心是shared state 这里怎么理解是你
本帖使用社区开源推广，符合推广要求。我申明并遵循社区要求的以下内容： 我的帖子已经打上 开源推广 标签： 是 我的开源项目完整开源，无未开源部分： 是 我的开源项目已链接认可 LINUX DO 社区： 是 我帖子内的项目介绍，AI 生成、润色内容部分已截图发出： 是 以上选择我承诺是永久有效的，接受社区和佬友监督： 是 为了高效的科研，尽量发论文毕业，我从 openai 刚出来就在使用各种 AI 模型。并且从今天年初使用 Openclaw 的时候就在研究迭代各种 skill。本着开源的态度，我把自己的这套 “自认为” 高效好用的 skill 分享出来整理成了一个开源 repo： COMPASS
地址：https://drive.qq.com/ 55 个帖子 - 53 位参与者 阅读完整话题
还想再跟始皇确认下，咱们端午节福利还是不是跟以前一样，每人一个 Claude fable 5 不限token，一个 GPT Pro 年费，一个 Cursor Ultra 无限补全，一个 GitHub Copilot 企业席位，一个 Midjourney 年度会员，一个 Runway 无限生成额度，一个 Figma Organization 套餐，一个 Notion Enterprise，一个 Perplexity Max，一个 Vercel Enterprise，一个 Cloudflare 企业套餐，一个 AWS 十万美金额度，一个 Google Cloud 白金账户，一个 NVIDIA H1
最近把官网的页面做了一些升级，但还没有正式上线（直接用的CODEX，没想到效果还挺不错的！） 官网首页 登录页面（动态的） 现在的前端IU视觉看起来还不错吧！比我的第一个项目已经属于质的飞跃了，每个人的审美是不同的，但是对于美的理解是可以培养的！ 重点讲一下我的Vibe Coding的功能和思路吧:（该应用完全是因为自己的兴趣，原本是打算提高自己做视频或者动漫的效率和质量，结果越做越上头！！！） 很多佬想看效果，我就放2个实测的打戏和真人剧情戏片段吧！因为无法上传视频，音效和台词没办法 体验，就看看画面吧： 上面的视频都是我测试的直出且没有任何剪辑的片段！ 聊聊思路吧！ 1、直接把小说或者
前言 我一直很好奇——技术如此简单，危害如此严重，难道要出一次类似魏则西事件才有人来管吗？ 当然在你行业这种事司空见惯了。 1.1 SEO 要讲GEO，我们恐怕得先从SEO说起。经营过个人博客的读者应该对SEO这个词并不陌生。简单来说，SEO是一种“搜索引擎优化技术”，它旨在通过一系列手法1.提升特定网站与特定搜索词的关联性；2.提高特定网站的搜索排名。SEO本身并不是灰产，但使用不当手法提升自身排名的SEO属于灰黑产范畴。 1.2 GEO与SEO GEO（Generative Engine Optimization），中文名为“生成式引擎优化”。它的目的与SEO大体类似，但作用对象从传统的搜