Python · ★ 6,336 · 🍴 447 · 📈 1,265 stars today
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 60-95% fewer tokens, same answers. Library, proxy, MCP server.
中文介绍 通过压缩工具输出、日志、文件和 RAG 片段，在内容送入 LLM 之前减少 60-95% 的 token 消耗，同时保持回答质量不变。支持作为 Python 库、代理服务器或 MCP 服务器三种方式集成，适合需要控制 LLM API 成本的开发者和企业在生产环境中使用。
Python · ★ 141,080 · 🍴 9,610 · 📈 3,618 stars today
Python tool for converting files and office documents to Markdown.
中文介绍 微软开源的 Python 工具，可将多种文件格式（包括 Word、Excel、PowerPoint 等 Office 文档、PDF 及其他常见文件）批量转换为 Markdown 格式。适合需要将文档内容标准化后喂给 LLM 或构建知识库的开发者，简化文档预处理流程。
JavaScript · ★ 203,902 · 🍴 31,280 · 📈 1,533 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 面向 AI 编程 Agent 的性能优化框架，专为 Claude Code、Codex、Cursor 等工具设计。提供技能管理、本能行为、记忆系统和安全机制，以研究优先的理念提升 Agent 在复杂编程任务中的执行效率，适合深度使用 AI 编程助手的开发者和团队。
Python · ★ 59,136 · 🍴 5,713 · 📈 1,182 stars today
🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl!
中文介绍 自适应 Web 爬虫框架，能从单次请求扩展到大规模全站爬取，自动处理页面结构变化、反爬检测等常见问题。框架设计灵活，覆盖数据采集的完整链路，适合需要稳定抓取动态或频繁变更网站的数据工程师和爬虫开发者。
Python · ★ 12,523 · 🍴 1,536 · 📈 1,722 stars today
Hermes WebUI: The best way to use Hermes Agent from the web or from your phone!
中文介绍 Hermes Agent 的 Web 前端界面，支持在浏览器和手机上便捷使用 Hermes 智能体。提供友好的 Web UI，使用户无需命令行即可与 Hermes Agent 交互，适合希望通过图形界面或移动端访问本地部署 AI Agent 的普通用户。
TypeScript · ★ 4,491 · 🍴 581 · 📈 124 stars today
A modern platform for visual, flexible, and extensible graph-based investigations. For cybersecurity analysts and investigators.
中文介绍 面向网络安全分析师和调查人员的现代化可视化调查平台，基于图谱（Graph）方式呈现关联关系，支持灵活扩展。可用于 OSINT 情报收集、威胁溯源、攻击链分析等场景，帮助安全团队直观梳理复杂实体关系，提升调查效率。
Python · ★ 25,099 · 🍴 2,880 · 📈 783 stars today
VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning
中文介绍 OpenBMB 推出的免 Tokenizer 多语言文字转语音（TTS）系统 VoxCPM2，支持多语言语音生成、创意音色设计和高保真声音克隆。无需传统 Tokenizer 即可处理多语种文本，适合需要高质量语音合成、个性化音色定制的开发者和语音 AI 应用场景。
Jupyter Notebook · ★ 18,446 · 🍴 5,223 · 📈 574 stars today
Code for Machine Learning for Algorithmic Trading, 2nd edition.
中文介绍 《Machine Learning for Algorithmic Trading》第二版配套代码仓库，涵盖量化交易中机器学习模型的完整实现，包括特征工程、时序预测、强化学习等技术在金融市场中的应用。适合量化研究员、算法交易开发者及机器学习从业者学习和参考。
Python · ★ 6,368 · 🍴 1,488 · 📈 30 stars today
中文介绍 面向生产环境的 Agentic RAG（检索增强生成）课程代码仓库，聚焦将 RAG 技术与 AI Agent 结合部署至实际生产系统的实践内容。适合希望将 RAG 从原型推向生产落地的 AI 工程师和后端开发者学习参考。
TypeScript · ★ 24,612 · 🍴 2,177 · 📈 680 stars today
Memory engine and app that is extremely fast, scalable. The Memory API for the AI era.
中文介绍 面向 AI 时代的高性能记忆引擎与应用，提供极速、可扩展的 Memory API，可为 AI 应用持久化存储和检索用户记忆、上下文信息。适合构建个性化 AI 助手、对话系统或需要长期记忆能力的 AI 产品开发者集成使用。
Python · ★ 8,343 · 🍴 1,073 · 📈 66 stars today
Talk to any LLM with hands-free voice interaction, voice interruption, and Live2D taking face running locally across platforms
中文介绍 本地跨平台运行的 LLM VTuber 框架，支持免手动触发的语音交互、语音打断和 Live2D 实时面部动画驱动。可对接任意大语言模型，让用户与带有虚拟形象的 AI 进行自然对话，适合虚拟主播爱好者、AI 陪伴应用开发者及 VTuber 内容创作者。
👍 0
The scaling of Large Language Models (LLMs) has driven significant performance gains but created substantial challenges in inference efficiency. While Mixture of Experts (MoEs) architectures address this by decoupling model size from inference cost, training MoEs from scratch is often unstable and c
👍 2
Deep Research Agents have shown strong capability in multi-step information retrieval, reasoning, and long-form report generation, but existing benchmarks and systems remain predominantly text-centric, with limited evaluation of whether visual elements are factually reliable and well aligned with th
👍 31
Search agents are often trained as policies over growing transcripts: the model must decide how to search while also remembering what it has seen, which evidence is useful, which constraints remain open, and which claims have actually been checked. We argue that this formulation puts too much routin
👍 7
Reinforcement learning (RL) improves large language model (LLM) agents by teaching them which actions lead to high rewards, but provides little supervision on what those actions do to the environment. World modeling (WM) can fill this gap, yet existing approaches often require separate simulators, e
👍 0
Human annotation is the empirical foundation of much NLP research, from dataset construction to model evaluation, but papers often leave unclear who produced the annotations and how the annotation process was controlled. We provide the first large-scale, task-level audit of human annotation reportin
👍 1
Large language models solve complex problems by generating lengthy chains of explicit reasoning tokens. While effective, this makes reasoning expensive, length-sensitive, and constrained to (discrete) natural language. While latent reasoning offers a continuous alternative, determining useful struct
👍 11
Autoregressive (AR) video diffusion enables variable-length synthesis, but long-horizon generation often suffers from accumulated errors and identity drift. For efficiency, existing methods commonly adopt sliding-window attention during generation. This creates an irreversible generation trajectory:
👍 12
The Model Context Protocol (MCP) has emerged as a transformative standard for connecting large language models (LLMs) with external data sources and tools, and has been rapidly adopted across personal applications and development platforms. However, existing benchmarks predominantly focus on generic
👍 9
Building capable visual web agents requires long-horizon reasoning, precise grounding, and robust interaction with dynamic real-world websites. Despite rapid progress, the strongest systems remain largely proprietary, while open agents still depend heavily on supervised post-training over large coll
👍 5
Selecting the best response from multiple small-model samples using a stronger scorer is a simple inference-time strategy, but fails when the small model has already committed to incorrect reasoning paths. PRM guided search avoids this by scoring candidate continuations during generation, but requir
👍 11
In open-ended environments, exploration is fundamental for autonomous agents, yet current language model agents struggle with this. Effective exploration requires memory, but retaining raw interaction histories is computationally expensive over long trajectories. While latent memory offers a solutio
👍 5
Computer use agents (CUAs) today are primarily deployed as single serial agents. This setup is suboptimal for complex long-horizon tasks that benefit from task decomposition, parallel execution, and consistent re-planning based on new information. In this paper, we argue that we should instead move
👍 6
Reusable skills are a key mechanism for extending agent capabilities, allowing agents to accumulate experience and solve increasingly complex tasks. Yet most existing skill-learning methods store reusable experience as text-only assets, such as instructions, reasoning traces, or summarized trajector
👍 1
Charts are a primary medium for conveying quantitative and relational information, yet systematically evaluating chart parsing models remains difficult. Existing benchmarks focus on narrow chart types and leave diagrammatic structures such as flowcharts and mind maps largely unaddressed, while model
👍 4
Agentic search requires language model agents to explore many sources and answer complex information-seeking questions. Scaling test-time compute is a promising way to improve these agents, but current approaches can fail, because correct answers are often sparse and score-based selection depends on
👍 2
Effective real-world assistance requires AI agents with robust Theory of Mind (ToM): inferring human mental states from their behavior. Despite recent advances, several key challenges remain, including (1) online inference with robust uncertainty updates over multiple hypotheses; (2) efficient reaso
👍 4
True video intelligence demands more than recognizing what is visible: it requires reasoning about why events unfold, predicting what would change under different conditions, and deciding what to do next. We refer to this progression, from perception through causal reasoning and simulation to strate
👍 1
Large-scale multilingual bitext often contains two distinct problems: non-parallel sentence pairs and low-quality translations. We decompose model-based assessment for such data into two independent components: parallelism assessment with multilingual embeddings and reference-free quality estimation
👍 6
Multimodal large language models (MLLMs) have shown strong capabilities in perception, reasoning, and action generation. However, their ability to sustain exploration in dynamic open worlds remains unclear. Existing embodied and game-based benchmarks often compress interaction into short-horizon tas
👍 25
Watermarking embeds statistical signatures in AI-generated text for detection and attribution. We reveal a fundamental vulnerability: when users access multiple models (today's reality), watermarks trivially fail. Watermarks perturb output distributions away from the original, and in competitive mar
👍 26
Speculative decoding accelerates LLM inference by drafting multiple tokens and verifying them in parallel with the target model. However, its practical speedup is constrained by the trade-off between draft quality and drafting cost: autoregressive drafters model causal dependencies among draft token
👍 6
Large Vision-Language Models (LVLMs) map visual inputs into dense token sequences, imposing a quadratic computational bottleneck for inference. Elastic visual-token compression addresses this by training a single model that can run at multiple visual-token budgets. However, existing approaches strug
👍 1
A retrieval-augmented generation (RAG) system deployed over a multi-author institutional corpus can give a different answer to the same question depending on which source it retrieves -- a failure mode the dominant single-gold-answer paradigm cannot diagnose. We argue that source-dependence is a mis
👍 1
Reasoning models are evaluated on single-turn benchmarks but deployed in multi-turn dialogue, where users push back on correct answers. Under sustained adversarial pressure we find a previously undocumented failure mode: the chain-of-thought stays factually correct from first turn to last while the
👍 1
LLM-generated reviews for scientific papers are gaining considerable traction and are even being officially piloted by major conferences. We have to assume that not only reviewers are using LLM-assistance, but also that authors use LLMs to revise their papers before submitting. In this work, we perf
👍 1
AI systems are fallible, and humans can make mistakes in deciding whether to trust AI over their own judgment. Thus, improving human-AI collaboration requires understanding when, why, and how humans decide to rely on AI. We study two distinct reliance decisions: the delegation choice -- deciding whe
👍 54
As agent capabilities advance, existing benchmarks, such as τ^2-Bench, are becoming increasingly saturated. Yet constructing new benchmark tasks remains complex, costly, and labor-intensive. Moreover, the standard approach, in which scenarios are first written in natural language and then mapped to
👍 5
Large language model (LLM) unlearning has emerged as a crucial post-hoc mechanism for privacy protection and AI safety, yet auditing whether target knowledge is truly erased remains challenging. Existing output-level metrics fail to detect when this knowledge remains recoverable from internal repres
👍 3
Academic researchers need efficient and reliable methods for collecting high-quality information from trusted sources, but modern tools for AI-assisted research still suffer from the tendency of Large Language Models (LLMs) to produce factually inaccurate or nonsensical output, commonly referred to
👍 3
This paper addresses the challenge of integrating 3D meshes as a native modality within Multimodal Large Language Models (MLLMs). Diffusion-based large reconstruction models decouple semantic understanding from geometric reasoning, operating as stateless reconstructors conditioned on dense 2D pixel
@elpresidank · 116 粉丝 · 2.9M 阅 · 543 赞 · 35 转
Most AI agent memory is built on embeddings. And there's now a proof that this entire class of system is going to forget what you stored in it — and confidently make up things you never stored at all.
中文介绍 从数学层面论证基于向量嵌入的 AI agent 记忆系统存在根本性缺陷：不仅会遗忘已存储的信息，还会自信地「捏造」从未存储过的内容。作者提出以拓扑结构替代平坦嵌入空间来组织上下文，认为结构化记忆才能真正逃脱遗忘陷阱，适合关注 agent 长期记忆可靠性的开发者深思。
@1salman · 363 粉丝 · 2.0M 阅 · 682 赞 · 45 转
Everyone keeps asking whether AI favors specialists or generalists. I think that is the wrong question. AI does not pick a side. It changes the tradeoff. The old world forced a choice. You could go
中文介绍 反驳「AI 时代该做专才还是通才」的二元论，认为 AI 改变的不是选择结果，而是选择本身的代价。过去专精与广博互相排斥，AI 出现后两者可以按需切换——真正的竞争优势在于能快速在深度与广度之间动态切换的人，而非固守某一侧。
@eng_khairallah1 · 61.9K 粉丝 · 693.5K 阅 · 511 赞 · 71 转
Obsidian has 2,700+ community plugins. Over 100 of them are AI-related. Save this :) And the CEO of Obsidian personally published official Claude Skills for the platform - 12,900+ GitHub stars in
中文介绍 整理了 30 个 Obsidian 进阶工作流、插件与配置方案，重点覆盖普通用户不熟悉的玩法。Obsidian 现有 2700+ 社区插件，其中 100+ 与 AI 相关；Obsidian CEO 亲自发布了官方 Claude Skills 集成，GitHub 已获 12900+ Star，是目前 AI 笔记工具结合最值得关注的动向之一。
@prukalpa · 23.1K 粉丝 · 583.2K 阅 · 506 赞 · 80 转
A field guide to what it is, what it is not, and where it fits in your AI architecture. I have had some version of the same conversation with a CIO almost every day this year. Their team has read
中文介绍 为企业 AI 架构师提供「企业上下文层」的实战指南，厘清它是什么、不是什么以及在整体 AI 架构中的定位。作者几乎每天都与 CIO 进行同类对话，发现大量团队对这一概念存在误解，文章试图建立清晰的概念边界，适合正在规划企业级 AI 基础设施的技术决策者参考。
@theonejvo · 22.1K 粉丝 · 504.3K 阅 · 861 赞 · 1 转
Over the past year, @pewdiepie, has been turning into one of the most visible champions of private, self-hosted computing, and it has been a genuine pleasure to watch. What began in late 2025 as an
中文介绍 记录了对 PewDiePie 自托管 AI agent 系统的提示注入攻击实验：通过构造恶意 Cocomelon 网站触发 agent 漏洞，随后协助其修复防护。PewDiePie 近年来成为私有化、自托管计算的知名倡导者，本文展示了真实 agent harness 在间接提示注入场景下的脆弱性及防御思路。
@monokern · 1.2K 粉丝 · 263.1K 阅 · 505 赞 · 72 转
Most people treat research as a manual task. You open 10 tabs. You watch videos. You read articles. You take notes somewhere. An hour later you have a pile of information you're not sure what to do
中文介绍 分享将 Claude Code、NotebookLM 与 Obsidian 串联成「越用越聪明」研究系统的工作流：用 Claude Code 自动抓取与整理资料，NotebookLM 做多源摘要与问答，Obsidian 沉淀结构化笔记并形成知识图谱。目标是让每次研究的产出自动喂养下一次，彻底摆脱「打开10个标签页却不知道怎么用」的困境。
@garrytan · 853.3K 粉丝 · 180.6K 阅 · 503 赞 · 43 转
In January I got back into coding and I built Garry's List. Over five hundred thousand lines of Rails and the tests to police it. I was proud of it. I shouldn't have been. The thing worth being proud
中文介绍 Garry Tan 以亲身经历为例反思 AI 编程的误区：他用 AI 写了 50 万行 Rails 代码，但真正值得骄傲的不是代码量，而是产品决策本身。他警告开发者不要把 agent 当成流水线工厂（Foxconn 模式），堆砌代码只是在积累技术债，核心竞争力应回归到判断力与架构思维。
@dair_ai · 124.6K 粉丝 · 84.0K 阅 · 504 赞 · 83 转
1. SkillOpt Microsoft Research treats a compact natural-language skill document as the trainable state of a frozen agent, then learns that document through rollouts, reflection, and bounded edits
中文介绍 本周 AI 顶会论文速览，重点推荐：①微软研究院 SkillOpt——将自然语言技能文档作为冻结 agent 的可训练状态，通过 rollout 与反思迭代优化；另含多篇涵盖推理、多模态、agent 协作等方向的新作，适合跟踪前沿研究进展的研究者与工程师收藏。
@trq212 · 263.1K 粉丝 · 75.7K 阅 · 542 赞 · 36 转
Last week, we released dynamic workflows in Claude Code. Claude can now write its own harness on the fly, custom-built for the task at hand. While the default Claude Code harness is built for coding,
中文介绍 介绍 Claude Code 新发布的「动态工作流」功能：Claude 可以在执行任务时即时编写专属 harness，而非依赖固定的编程默认框架。这意味着 Claude Code 不再局限于代码任务，可根据任意目标自动生成最适合的执行结构，显著扩展了其作为通用 agent 的适用边界。
@mvanhorn · 27.6K 粉丝 · 54.5K 阅 · 545 赞 · 44 转
Three months ago I posted "Every Claude Code Hack I Know." It hit 913K views. @kevinrose had asked what IDE to use, and my answer was: "No IDE. Just plan.md files and voice." This used to be called
中文介绍 作者三个月前发布的「Claude Code 全部技巧」获 91.3 万浏览后，再次更新 2026 年 6 月版「Agentic 工程全技巧」。核心理念延续：不用 IDE，只用 plan.md 文件加语音控制；内容升级为更广泛的 agentic 工程实践，覆盖多 agent 协作、任务分解与状态管理等新场景。
@0xJeff · 80.5K 粉丝 · 47.0K 阅 · 512 赞 · 45 转
It’s been a little more than 2 months now and Hermes has taught me so many different things. One of the key things is that Agents like Hermes or OpenClaw don’t fail on intelligence, they tend to fail
中文介绍 作者使用 AI 分析 agent「Hermes Analyst」60 天后总结 6 个工作流与 6 条核心教训。最关键发现：Hermes、OpenClaw 等 agent 的失败根源不是智能不足，而是上下文管理问题——信息组织方式、任务边界定义与工具调用时机才是决定 agent 可靠性的真正瓶颈。
GitHub pioneered the modern AI coding era with Copilot, and the resulting explosion in agentic coding has led to notable strains on the most popular developer platform in the world. Here's the plan.
中文介绍 GitHub首席产品官Kyle Daigle阐述了GitHub针对AI代理编程时代的战略规划。Copilot引领了现代AI编程浪潮，随之而来的代理式编程爆发式增长对这一全球最大开发者平台造成了显著压力，GitHub正就此制定应对方案以维持平台稳定性与可扩展性。
中文介绍 Hcompany在Hugging Face发布Holo3.1模型，定位为快速、本地化运行的计算机使用代理（Computer Use Agent），支持在本地设备上执行自动化操作任务。
Travelers built an AI-powered Claim Assistant with OpenAI to guide customers through filing claims, provide 24/7 support, and scale operations during peak demand.
中文介绍 美国保险公司Travelers与OpenAI合作，基于OpenAI技术构建了AI驱动的「理赔助手」，已在全国范围内部署。该系统可引导客户完成理赔申请流程、提供全天候24小时支持，并在理赔高峰期实现运营规模弹性扩展。
The global health care sector is under increasing strain. Decades of chronic underinvestment and constraints in recruitment have coincided with a surge in demand for services for aging populations. Gaps in provision are already taking a toll, with fragmented access to care and high rates of stress a
中文介绍 MIT《技术评论》分析了代理式AI在全球医疗领域的应用前景。全球医疗体系面临长期投资不足、人员招募受限与老龄化人口服务需求激增的三重压力，碎片化的医疗服务可及性问题突出，代理式AI被视为缓解上述矛盾、重塑人性化医疗的潜在路径。
Discover new Codex plugins, sites, and annotations that help analysts, marketers, designers, investors, and other teams get more done with AI.
中文介绍 OpenAI发布Codex新插件、站点集成与代码注释功能，面向分析师、市场营销人员、设计师、投资者等多类职能角色及多种工具和工作流场景，旨在帮助不同团队借助AI提升工作效率。
This article is from Making AI Work, MIT Technology Review’s limited-run newsletter examining how to apply LLMs across industries. To receive it in your inbox,sign up here. From accounting to design to market research and product development, there’s a staggering breadth of skills needed to run a bu
中文介绍 MIT《技术评论》「让AI发挥作用」系列通讯探讨中小企业如何利用大语言模型提升竞争力，涵盖会计、设计、市场调研及产品开发等广泛应用场景，为中小企业主提供跨行业的AI落地实践建议。
OpenAI calls for global action on youth AI safety, proposing an international institute to strengthen safeguards, standards, and opportunities for young people.
中文介绍 OpenAI呼吁全球各方联合行动，推进青少年AI安全议题，并提议成立一个国际性机构，专门负责强化针对未成年人的AI安全保障、制定相关标准，同时为年轻群体创造更多AI发展机遇。
Jensen scores a huge win.
中文介绍 英伟达发布Cosmos 3世界模型、Nemotron 3 Ultra大语言模型及RTX Spark新品，CEO黄仁勋（Jensen Huang）此次发布被评价为重大胜利。文章对上述新产品进行了综合报道与解读。
The Next Era of Knowledge Work report explores how Codex is transforming productivity through AI-powered research, data analysis, workflow automation, and content creation.
中文介绍 OpenAI发布《下一个知识工作时代》报告，阐述Codex如何通过AI驱动的研究、数据分析、工作流自动化及内容创作功能，成为面向所有知识工作者的通用生产力工具，标志其定位从编程助手向全场景效率平台延伸。
中文介绍 TLDR AI每日简报（2026年6月2日）聚焦三大要点：Anthropic提交IPO申请文件、OpenAI与AWS达成云服务合作、Perplexity开放搜索相关代码，涵盖当日AI行业重要动态。
Our approach to AI policy and political advocacy, transparency, support for thoughtful regulation and AI safety, and that no outside political group speaks on the company’s behalf.
中文介绍 OpenAI公开阐述其AI政策立场与政治倡导原则，内容涵盖透明度承诺、对审慎监管与AI安全的支持态度，并明确声明不存在任何代表公司发声的外部政治团体，强调公司政策独立性。
中文介绍 JetBrains在Hugging Face发布Mellum2模型，这是一款拥有120亿参数的混合专家（Mixture-of-Experts）架构语言模型，专为开发者工具场景设计，是JetBrains在AI编程辅助领域的最新成果。
Inside xAI: Building Grok Imagine in 3 Months, Videogen vs World Models, and why Grok Imagine is so underrated. For the first time, we do a deep dive with the guy who led it!
中文介绍 xAI Grok Imagine负责人Ethan He首次深度分享了在3个月内构建Grok Imagine视频生成模型的全过程，探讨视频生成模型与世界模型的差异，并阐述为何视频代理模型将成为AI下一个重要发展方向。
中文介绍 IBM研究院发文指出，企业级AI规模化落地不能仅依赖大语言模型本身，代理逻辑（Agent Logic）是实现可扩展企业AI应用的关键所在，文章探讨了超越单纯LLM能力、构建可靠企业AI系统的架构思路。
OpenAI breaks ground on a 1GW data center project in Michigan as part of Stargate, building AI infrastructure to expand access, create jobs, and support communities.
中文介绍 OpenAI在密歇根州正式启动Stargate项目数据中心建设，该项目规模达1吉瓦（1GW），旨在为「智能时代」构建AI基础设施，预计将扩大AI服务可及性、拉动当地就业并支持社区发展。
从当前技术指标整体来看，各大资产类别分化明显。美股方面，SPY（RSI14 75.6）与 QQQ（RSI14 79.4）均处于超买区间，价格贴近 52 周高位，多头排列完整，短期动量强劲；但 GOOGL 单日跌幅 -3.86%、5日跌幅 -6.95%，MACD 大幅背离，META 呈空头排列且 MACD 死叉，个股层面出现明显分化，美股整体呈强势但内部结构并不均匀。加密货币市场承压最为显著：加密恐慌贪婪指数降至 11（极度恐慌），加密总市值 24h 下跌 5.33%，BTC 主导率 56%。BTC（RSI14 22.1）、ETH（RSI14 22.4）、SOL（RSI14 26.4）三者均跌入超卖区间，空头排列完整，MACD 负值扩张，技术面呈系统性下行压力。中概股方面，腾讯单日异动 +10.46%，MACD 金叉初现，但阿里巴巴、拼多多仍维持空头排列；整体中概板块技术面偏弱，局部出现反弹迹象但趋势尚未逆转。商品与外汇方面，黄金期货（4507）低于 SMA20 和 SMA50，RSI14 43.4 偏弱；原油 RSI14 47.8 中性；USDCNY RSI14 29.3 触及超卖，美元指数 DXY 则维持 SMA 多头排列，RSI14 56.9，相对稳健。宏观指标上，VIX 现报 15.77 且 5日下行 -7.29%，美债 10Y 收益率 4.46% 处于 MACD 死叉，整体市场情绪呈现股强、币弱、中概分化的格局。
BTC 当日跌幅 -6.13%，5日跌幅 -8.96%，现价 66,946 已跌破 SMA20（75,674）、SMA50（77,092）、SMA200（79,243）三条均线，呈空头排列。RSI14 读数仅 22.1，深陷超卖区间；MACD 柱值 -1856，远低于信号线 -962，动量持续向下。配合加密恐慌贪婪指数降至 11（极度恐慌），加密总市值 24h 下跌 5.33%，技术面整体承压显著。
QQQ 当前报价 746.16，距 52 周高仅 -0.04%，5日涨幅 +2.17%。价格站于 SMA20（716.13）、SMA50（659.21）、SMA200（619.49）之上，呈明显多头排列。RSI14 读数 79.4，处于超买区间；MACD（21.98）已高于信号线（21.60），形成金叉。短期动量强劲，但 RSI 超买提示上行空间可能有限。
META 现价 597.63，距 52 周高回落 -24.94%。价格低于 SMA20（612.27）、SMA50（618.49）、SMA200（664.75），三线呈空头排列。MACD 为 -4.05，信号线 -3.98，柱状图负值且发生死叉，动量向下。RSI14 为 40.7，接近偏弱区域。近 5 日跌幅 -2.4%，技术结构整体偏空。
腾讯单日大涨 +10.46%，5日涨幅 +9.70%，现价 481.6 接近 SMA50（480.8），但仍低于 SMA200（573.63），距 52 周高回落 -29.49%，呈中性偏整理格局。MACD（-11.40）已上穿信号线（-14.03）形成金叉，RSI14 58.2 处于正常区间。短线动量有所改善，但均线结构尚未完全转强，方向待进一步确认。
USDCNY 现价 6.76，低于 SMA20（6.79）、SMA50（6.83）、SMA200（6.98），呈空头排列，接近 52 周低位。RSI14 读数 29.3，进入超卖区间；MACD（-0.016）形成死叉并低于信号线（-0.014）。5日跌幅 -0.36%，美元相对人民币持续走软，技术面信号偏空。
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数据计算所得技术指标进行客观描述，仅供技术指标解读参考，不构成任何投资建议或买卖依据。过去走势不代表未来表现，市场存在不确定性，请读者自行评估风险。
Follow today’s news live Get our breaking news email, free app or daily news podcast Chaney said: AI companies should be able to unlock the majority of global content with a handful of individual deals – in the same way that every other industry licenses copyrighted content. Government could further
中文摘要 澳大利亚政治动态：所罗门群岛总理抵达澳大利亚，与总理阿尔巴尼斯展开条约谈判。与此同时，工党议员康罗伊指控胡锡在AUKUS核潜艇计划党团投票问题上「重写历史」。此外，人工智能公司版权授权问题引发讨论，有观点认为AI企业应能通过少数协议解锁全球大多数内容。
Over the past two decades, Iran repeatedly threatened to close down the waterway. President Trump underestimated Iran’s ability to do so.
中文摘要 过去二十年间，伊朗曾多次威胁封锁霍尔木兹海峡。分析人士指出，特朗普政府低估了伊朗实施封锁的实际能力，未能充分重视相关兵棋推演和警告信号，最终导致局势演变至当前紧张态势。
The US military said aircraft disabled the engine room of the unladen Botswana-flagged M/T Lexie as it passed through international waters US forces fired a Hellfire missile to disable a tanker attempting to break through the American blockade and reach Iran, amid an enduring standoff in the strait
中文摘要 美军在霍尔木兹海峡对峙期间，向一艘试图突破封锁驶向伊朗的博茨瓦纳籍油轮「M/T Lexie」发射地狱火导弹，击中该空载油轮机舱使其失去动力。油轮当时正在国际水域航行。此举是美方维持对伊封锁行动的一部分。
Syrian commission confirms the deaths of Rania al-Abbasi’s six children, missing under Bashar al-Assad’s rule.
中文摘要 叙利亚相关委员会证实，在阿萨德政权统治期间失踪的拉尼娅·阿尔-阿巴西的六名子女已确认死亡。此前这些孩子的下落长期不明，此次调查结果揭开了这一家庭悲剧的真相。
A colossal 400-kilogram flag was draped over Rome’s Colosseum as Italy marked the 80th anniversary of its republic.
中文摘要 意大利庆祝共和国成立80周年，罗马斗兽场悬挂了一面重达400公斤的巨型国旗以示纪念。意大利于1946年6月2日通过公投废除君主制，正式建立共和国，此次活动是该国重要的历史纪念节点。
British actor Idris Elba was knighted by King Charles III at Windsor Castle on Tuesday.
中文摘要 英国演员伊德里斯·艾尔巴于周二在温莎城堡接受英王查尔斯三世授勋，正式获封爵士。艾尔巴以出演《路德》《火线》等影视作品闻名，此次受封是对其艺术成就的官方认可。
Demonstrators in Nairobi have carried an ‘Ebola’ coffin to protest a proposed US quarantine facility.
中文摘要 肯尼亚首都内罗毕爆发抗议活动，示威者抬着标有「埃博拉」字样的棺材游行，反对美国在当地设立隔离检疫设施的计划。抗议者对该设施的安全性及其潜在影响表达强烈不满。
The electoral mission described the first round of voting in Colombia's presidential race as 'a lesson in democracy'.
中文摘要 欧洲选举观察团驳斥了有关哥伦比亚总统选举存在舞弊的传言，将首轮投票描述为「民主的典范」。观察团对整体选举程序表示认可，认为相关舞弊指控缺乏依据。
After a musician fell ill during a live performance of the score from “La La Land,” the composer Justin Hurwitz asked for a sight reader. A 21-year-old student stepped up.
中文摘要 悉尼一场《爱乐之城》管弦乐现场演出中，一名演奏员突然病倒，作曲家贾斯汀·赫尔维茨当场向观众席寻求能即兴视奏的替补。一名21岁的音乐学生挺身而出，成功填补空缺，使演出得以顺利完成。
Prime Minister Benjamin Netanyahu of Israel vowed to continue Israel’s campaign against Hezbollah, even as Lebanese and Israeli officials met in Washington.
中文摘要 以色列与黎巴嫩和平谈判在华盛顿重启，以色列和黎巴嫩官员出席会谈。与此同时，以色列总理内塔尼亚胡宣示将继续对真主党的军事行动，以色列对黎巴嫩的打击仍在持续，局势尚未出现实质性缓和。
The action against Nobitex and several of its executives accused it of helping the government evade sanctions and finance militants.
中文摘要 美国财政部宣布对伊朗最大加密货币交易所Nobitex及其多名高管实施制裁，指控该平台协助伊朗政府规避国际制裁、为武装组织提供资金支持。此举是美国对伊朗经济施压行动的最新举措。
中文摘要 该条目为伊朗战争及以黎局势滚动直播的最新动态入口，暂无具体摘要内容。相关详情请参阅同系列直播报道。
The Canadian trade minister met with the U.S. Trade Representative on Tuesday.
中文摘要 加拿大贸易部长与美国贸易代表于周二举行会谈，就美加贸易关系进行磋商。就在会谈开始前数小时，特朗普再度将加拿大称为美国「第51个州」，此番言论再次引发两国外交紧张。
Plans for luxury resorts on the Adriatic Sea have long worried environmentalists. The developments are now a flashpoint for broader frustrations with the government.
中文摘要 阿尔巴尼亚爆发抗议活动，反对与特朗普女婿库什纳相关的亚得里亚海豪华度假村开发项目。该计划长期引发环保人士担忧，当前抗议规模持续扩大，已演变为民众对政府施政广泛不满情绪的集中爆发点。
Analysts say 'strategic leaks' aim to influence public perceptions amid diplomatic impasse in US-Israel war on Iran.
中文摘要 坊间流传特朗普曾严厉斥责以色列总理内塔尼亚胡，但多位分析人士对美以关系出现裂痕的传言持怀疑态度。分析人士认为，这些「战略性泄露」意在借美伊战争外交僵局之机影响外界对美以关系的观感。
Gold edged lower, after US jobs data reinforced bets that the Federal Reserve will keep interest rates higher for longer.
中文摘要 受美国强劲就业数据影响，市场押注美联储将长期维持高利率，黄金价格小幅下跌。就业数据表现稳健，进一步削减了美联储近期降息的预期，令贵金属承压。
A newly-published report suggests a moderate lifestyle in retirement costs £32,700 for one person and £45,400 for two.
中文摘要 一份新发布的报告显示，英国四分之三的在职人员养老金储蓄不足以支撑退休后的"适度"生活水平。报告测算，单人适度退休生活每年需花费3.27万英镑，双人则需4.54万英镑。
The Bank of England is asking the public which animals should appear on future banknotes.
中文摘要 英格兰银行正向公众征询意见，从入选候选名单的18种动植物中选取形象印制于未来版本纸币上，其中包括海雀（Puffin）和大黄蜂（Bumblebee）等物种。
Technology committee calls on government to trigger break clause in flashpoint NHS contract
中文摘要 英国议会技术委员会表示，数据分析公司Palantir不应在英国公共数据系统中扮演重要角色，并呼吁政府触发其与NHS（国家卫生服务）合同中的解约条款，以终止这一争议合同。
UK regulator focuses on teams with partnerships with digital assets companies that are not authorised to operate in Britain
中文摘要 英国金融监管机构正审查英超球队与未获英国授权经营的加密货币公司之间的赞助合作，相关俱乐部或面临法律追究。监管重点在于这些数字资产企业是否合规。
Taiwan’s central bank is extending its reach ever deeper into foreign-exchange markets, seeking to ensure currency stability as the artificial intelligence boom intensifies a divide between the island’s tech sector and the rest of the economy.
中文摘要 台湾央行正持续深化对外汇市场的干预力度，以维护新台币汇率稳定。背景是人工智能热潮加剧了台湾科技产业与其他经济部门之间的分化，导致汇率波动压力上升。
Fund with net assets of $31bn aimed at retail investors limits withdrawals in latest sign of exodus from sector
中文摘要 私募信贷资管机构Cliffwater旗舰基金遭遇大规模赎回申请，赎回比例达净资产的17%。该基金净资产规模约310亿美元，主要面向零售投资者，为此已限制提款，成为私募信贷行业资金外流趋势的最新例证。
Asian stocks tracked Wall Street higher as renewed enthusiasm for the artificial-intelligence trade drove the S&P 500 to a ninth day of gains. The yen traded close to the 160-per-dollar level.
中文摘要 受人工智能交易热情重燃推动，标普500指数录得连续第9个交易日上涨，亚洲股市随之跟涨。与此同时，日元汇率逼近160日元兑1美元关口。
Oil rose a third day on pessimism over the prospects of the US and Iran reaching a peace deal and as fighting flared up in the Persian Gulf.
中文摘要 受美伊和平谈判进展停滞及波斯湾地区冲突再度升级影响，国际油价连续第三个交易日上涨。市场对美伊达成协议的悲观情绪加剧了供应担忧，推动油价走高。
The company said its own workers are testing a "wearable access badge" and a desktop device.
中文摘要 微软正测试一款面向办公室员工的可穿戴人工智能设备，包括"可穿戴门禁徽章"和桌面装置两种形态，目前正由公司内部员工试用。
SpaceX is planning to set the terms of its initial public offering as early as Wednesday afternoon, according to people familiar with the matter, ahead of what’s expected to be the biggest ever listing.
中文摘要 据知情人士透露，SpaceX计划最早于周三下午确定IPO发行条款，预计此次上市将成为有史以来规模最大的首次公开募股之一。
GameStop Corp. shares climbed as much as 13% in after-hours trading after the video-game retailer reported a record quarterly profit driven by its collectibles business.
中文摘要 GameStop公布史上最高单季利润，增长主要由其收藏品业务（包括手办和卡牌）驱动。财报发布后，该公司股价盘后一度上涨逾13%。
16 回复 · 程序员 节点
11 回复 · 程序员 节点
6 回复 · 程序员 节点
21 回复 · 程序员 节点
14 回复 · Apple 节点
6 回复 · 程序员 节点
8 回复 · Apple 节点
4 回复 · Apple 节点
7 回复 · Linux 节点
13 回复 · Apple 节点
还得练啊，都过去一个月了，还没看到无限bug team 随便用 54 个帖子 - 51 位参与者 阅读完整话题
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 项目地址： GitHub： github.com GitHub - physics-dimension/PriceAI: AI 订阅渠道比价工具：聚合 ChatGPT、Claude、Gemini、Grok... AI 订阅渠道比价工具：聚合 ChatGPT、Claude、Gemini、Gr
挺多号又刷新了额度。让我遇上Bug了？ 准备明天重开公益站 还有人需要吗？ 130 个帖子 - 121 位参与者 阅读完整话题
前段时间已经关门了，现在我又找到了方法，我觉得德国沃达丰这张手机卡是那种你不申请，绝对血亏，会后悔到拍大腿的卡，它的成本已经低到没人性了，低到每年3毛钱就行，你不申请，也许过几天这个羊毛就彻底没了。 最近火热的giffgaff卡，保号都没这么的便宜 全流程时间预估：从申请到激活使用，整个过程顺利的话大约需要 10分钟作用， 遇到问题也不超过24 小时。其中填表和 KYC 认证只需 10-20 分钟，主要等待时间在于官方审核（通常 10 分钟到 24 小时不等，大部分人 1小时内能收到。建议尽早申请，趁政策还没收紧！ 教程如下，从如何注册申请——KYC详细步骤——如何注册绑定账号——如何充值保号
我在很早之前就说过，后面我们会有很多小单子放出来，给大家接来练手，这玩意儿不可能赚不到钱，只是没想到市场变化的这么快，B级的单子做不了了（没市场），起步都是A+，但这也不是问题，全流程我们已经跑通，所以要扩招人员，下半年猛猛冲击高质量的短剧和网络大电影。 1.这次扩招我们需要6名左右的视频制作人员，我们扩招的人一定要有责任心，人品好，技术都可以慢慢打磨，基本跟着流程走一部就很快能学会了，怎么理解“责任心？” 就比如一部剧只有10天的制作周期，每人分配到10-15集，如果交片的质量不行，和做不完15集，我们是需要担责的，违约会扣我们很多钱。【一定不能在你这出了问题，比如大家都在加班熬夜赶片，你说
看了很多强调上下文重要性的文章，用cc接国模愈发不顺手，总感觉cc的上下文屎山是不是越来越大了，于是捡起之前折腾了一半的pi，我的想法和作者部分一致，上下文工程是第一要务，准确的控制注入模型的每一个字肯定能获得比黑盒更舒服更有包裹感(误 。 点进这个帖子的应该或多或少都是听说过pi的，如果不了解请看作者的文章 agent，无 MCP、无后台 bash、无内置 to-do、无 plan mode、无子 agent，本体的内置提示词不超过1000token，仅内置4 个工具：read、write、edit、bash 关于package是什么 在pi中，Package 可以是： Extension：
前女友好像经常半夜跑电报，我想知道她来电报做啥？有没有知道的？二次元？看片？还是啥。 107 个帖子 - 81 位参与者 阅读完整话题
目前手中有GG,德国沃达丰，O2,美国手机卡，如果有护照的，强烈建议德国沃达丰和O2,买个小白卡成本20多元，但是可以写入6个手机号，目前我申请了3个沃达丰和3个02，使用wise定时保号，沃达丰一个月充值0.01元，o2目前充值0.01，官网变成一年保号一次就可以，所以6个号一年成本就2块人民币左右，最不推荐的就是GG 85 个帖子 - 40 位参与者 阅读完整话题
人家的订阅都是用量用完了限制使用，他MiniMax浓眉大眼的是接着奏乐接着舞，给你直接扣余额。 扣余额吧就扣余额，你余额是0的他还能给你扣成负的。 放眼世界应该没第二家了吧？ 问客服，有没有开关可以控制，切换到正常的模式，问就是没有。 欠费是没多少，就一块多，但是无语是挺无语的。 46 个帖子 - 27 位参与者 阅读完整话题
辞职在家 目前接单软件开发 一个月也赚几千块 咋办现在 家里一来人就问上班没 干啥工作 家里人也是 现在压力很大很焦虑 一说我就上火 已经不愿意跟任何人说话了 天天给自己关在自己小屋子里面 咋办 100 个帖子 - 75 位参与者 阅读完整话题