mattpocock/skills
Shell · ★ 252,569 · 🍴 21,316 · 📈 2,666 stars today
Skills for Real Engineers. Straight from my .agents directory.
中文介绍 面向真实工程师的技能集，来自作者本人的 .agents 目录。提供可直接用于 AI 编程助手的技能片段，帮助开发者提升编码效率，适合使用 Claude Code 等工具的程序员。
2026-09-06
← 历史归档affaan-m/ECC
JavaScript · ★ 249,860 · 🍴 37,611 · 📈 1,325 stars today
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
中文介绍 代理（agent）性能优化系统，包含技能、直觉、记忆、安全和研究优先的开发支持。适用于 Claude Code、Codex、Opencode 和 Cursor 等 AI 编程工具，旨在提升代理的自主性和准确性。
DietrichGebert/ponytail
JavaScript · ★ 127,899 · 🍴 6,845 · 📈 2,813 stars today
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
中文介绍 让 AI 代理像最懒的资深开发一样思考，强调“最好的代码是没写的代码”。通过减少不必要的代码生成，简化开发流程，适用于希望提高代码质量和效率的开发者。
NousResearch/hermes-agent
Python · ★ 241,985 · 🍴 49,718 · 📈 573 stars today
The agent that grows with you
中文介绍 一款能与你共同成长的 AI 代理，强调持续学习和适应能力。可能采用自适应记忆机制，适用于需要个性化交互和长期陪伴的开发者或普通用户。
fmtlib/fmt
C++ · ★ 25,567 · 🍴 3,043 · 📈 133 stars today
A modern formatting library
中文介绍 现代 C++ 格式化库，提供类型安全、可扩展的字符串格式化功能，替代 printf 和 iostream。性能优异，支持用户自定义类型，广泛应用于 C++ 项目中，是 https://github.com/fmtlib/fmt 的官方仓库。
anthropics/skills
Python · ★ 174,546 · 🍴 20,670 · 📈 472 stars today
Public repository for Agent Skills
中文介绍 Anthropic 官方公开的 Agent Skills 仓库，包含用于增强 AI 代理能力的技能定义和示例。开发者可借鉴或复用这些技能，提升 Claude Code 等代理的定制化和功能性。
cathrynlavery/diagram-design
HTML · ★ 31,670 · 🍴 2,040 · 📈 852 stars today
38 editorial diagram types for Claude Code, Codex, and Pi. Self-contained HTML + SVG. No shadows. No Mermaid slop.
中文介绍 提供 38 种编辑级图表设计模板，采用自包含的 HTML+SVG 格式，无阴影、不依赖 Mermaid。适用于 Claude Code、Codex 和 Pi 等工具，方便快速生成专业美观的图表。
anomalyco/opencode
TypeScript · ★ 204,665 · 🍴 26,701 · 📈 725 stars today
The open source coding agent.
中文介绍 开源编码代理，旨在帮助开发者自动完成编程任务。它可能集成代码生成、调试和重构功能，适用于希望提升开发效率的个人或团队。
ruvnet/ruflo
TypeScript · ★ 70,687 · 🍴 8,414 · 📈 127 stars today
🌊 The original agent meta-harness. Deploy intelligent multi-player swarms, coordinate autonomous workflows, and build conversational AI systems. Features adaptive memory, self-learning intelligence, RAG integration, and native Claude Code / Codex / Hermes and many more Integrated
中文介绍 原创代理元框架，支持部署智能多代理集群、协调自主工作流和构建对话式 AI 系统。具备自适应记忆和自学习能力，适用于复杂 AI 应用开发场景。
humanlayer/skills
TypeScript · ★ 2,681 · 🍴 74 · 📈 408 stars today
中文介绍 HumanLayer 提供的技能集，但无详细描述。可能包含用于增强 AI 代理与人类协作的功能模块，推测用于提升交互效率，具体内容需查看仓库文档。
blader/humanizer
Python · ★ 43,458 · 🍴 3,644 · 📈 988 stars today
Agent skill that removes signs of AI-generated writing from text
中文介绍 智能体技能，用于从文本中去除 AI 生成痕迹，使内容更自然。适合需要伪原创或去 AI 味的写作场景，可用于博客、文案等。
BraveOPotato/FckSignups
TypeScript · ★ 2,876 · 🍴 198 · 📈 50 stars today
A list of tools that are open-source, in-browser, and require no-signups!
中文介绍 汇集了无需注册即可使用的开源浏览器工具列表。涵盖多种实用工具，方便用户在线直接使用，省去注册麻烦，适合追求便捷的用户。
WorldFlowAI/everything-claude-code
JavaScript · ★ 2,341 · 🍴 361 · 📈 139 stars today
Claude Code toolkit - agents, commands, skills, rules, and hooks for productive AI-assisted development
中文介绍 Claude Code 工具包，包含代理、命令、技能、规则和钩子，助力高效 AI 辅助开发。提供预配置规则和技能，适合希望快速上手 Claude Code 的开发者。
magnitudedev/magnitude
TypeScript · ★ 3,185 · 🍴 228 · 📈 686 stars today
Open source inference server that runs the best local models for your hardware, plugged into the agent you already use. Works with Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, Oh My Pi, and Cline.
中文介绍 开源推理服务器，根据硬件自动运行最佳本地模型，并集成到现有代理中。支持 Pi、OpenCode、Hermes 等多种代理，适合注重隐私和本地部署的用户。
bikini/exploitarium
Python · ★ 4,699 · 🍴 1,252 · 📈 232 stars today
A single archive of public exploit PoCs and vulnerability research writeups. At the time I post these, none have been reported. Feel free to report them yourself and take credit for the CVE if handed out lulz. Please do not abuse these. I do this so to allure people into the field, and I've always f
中文介绍 公开漏洞利用 PoC 和漏洞研究文档的单一归档。作者发布时称这些未报告，鼓励研究者自行报告并获取 CVE 编号，适合安全研究人员参考和使用。
nvm-sh/nvm
Shell · ★ 94,910 · 🍴 10,424 · 📈 57 stars today
Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions. $nvm: 3ArcxqLtXMmBnWbbtfwQgVL3MNnDsggzgGDtXMnjpump
中文介绍 Node 版本管理器，使用 POSIX 兼容的 bash 脚本管理多个活动的 Node.js 版本。支持版本切换、安装等，是 Node.js 开发者的常用工具，适用于 macOS/Linux 系统。
Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments
👍 265
As terminal-based code agents become prevalent, agent trajectories have accumulated at scale, while realistic, executable environments remain scarce. However, environments are what agent post-training actually requires: each can be re-queried into many verifiable tasks and provides execution feedbac
中文介绍 为解决终端代理训练环境稀缺的问题，提出Terminal-Universe，将大量代理轨迹转化为可执行、可扩展的终端环境。每个环境可重复产生可验证任务并提供执行反馈，从而为代理后训练提供充足监督信号。
Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning
👍 159
Large language models achieve superior performance on tasks that require extended reasoning, but long chains of thought make the KV cache a severe memory bottleneck. Existing KV cache compression methods share one paradigm: score each cached token by some estimate of how much it will matter later, a
中文介绍 针对大模型长推理中KV缓存的内存瓶颈，提出随机注意力机制，重新思考缓存驱逐策略。不同于传统基于重要性打分的方法，该方法从随机性出发，高效选择保留的键值对，在保持推理性能的同时显著降低内存占用，实现高效长链推理。
Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training
👍 149
Large language models offer broad capabilities, but adapting them to evolving domains, tools, and requirements often entails repeated post-training. Autonomous systems automate parts of this process by proposing updates, training candidates, and using evaluation feedback to select subsequent proposa
中文介绍 关注自主LLM后训练中的经验迁移问题，提出条件性迁移方法，智能判断何时复用先前经验。通过评估反馈动态调整更新策略，避免低效迁移，提升多轮后训练的适应性和效率。
Rethinking On-Policy Distillation of Large Language Models II: One Training Example
👍 74
On-policy distillation (OPD) combines student-generated rollouts with dense token-level supervision from a teacher. Existing work has mainly studied its algorithmic behavior, leaving the role of training data unclear. We examine this role at the data-minimal limit by training on a single query. One-
中文介绍 在数据极简极限下（仅用单个查询）研究on-policy distillation的训练数据作用。揭示单条数据对蒸馏效果的关键影响，为理解OPD的算法行为与数据依赖提供新视角。
Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM
👍 73
Hybrid LLMs pair softmax attention with linear-attention layers such as Gated DeltaNet (GDN), whose recurrent state summarizes the context in fixed size. Early community 4-bit quantizations of Qwen3.8-27B (48 GDN layers, 16 attention layers) left the GDN block in 8- or 16-bit precision -- especially
中文介绍 研究混合大模型（如Qwen3.8-27B）中Gated DeltaNet层为何能抗4-bit量化。归因于其循环结构特性，并通过NVFP4 W4A4量化在保持性能的同时大幅减少内存占用，实现高效部署。
Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction
👍 45
Online 3D reconstruction models perform poorly on long videos. This happens because regressing poses relative to a fixed first-frame anchor forces extrapolation far beyond the training distribution. Small drifts accumulate and amplify into significant geometric collapse. However, we observe that per
中文介绍 针对在线3D重建在长视频上性能下降的问题，提出Scal3R，采用多相对位姿查询而非固定首帧锚点，避免外推误差累积，实现可扩展的在线重建。
Editable Visual Design
👍 40
While diffusion base models such as GPT-Image-2 and Nano-Banana exhibit remarkable visual expressiveness, their end-to-end generation inherently yields flattened bitmaps with error-prone text, precluding layer-wise post-editing. Conversely, code-based visual generation via Coding Agents provides pre
中文介绍 提出可编辑视觉设计方法，结合扩散模型与编码代理，生成分层可编辑的视觉内容，克服端到端生成难以进行图层级修改的局限。
WHALE: A Simple Recipe for Joint Harness-Weight Optimization
👍 30
Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness u
中文介绍 提出WHALE方法，联合优化代理的模型参数与可执行harness代码。通过交替更新打破冻结瓶颈，提升整体代理性能。
Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding
👍 29
Streaming video understanding requires multimodal large language models (MLLMs) to process continuous visual inputs and respond to user queries under strict causality and bounded memory. Existing approaches typically compress historical observations into an external memory bank and retrieve query-re
中文介绍 为流式视频理解提出渐进式潜在记忆演化方法，无需外部检索，在严格因果和有限内存下动态更新记忆，提升长视频理解能力。
Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering
👍 26
Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employ CLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semanti
中文介绍 针对知识型VQA中检索仅注重视觉相似度的问题，提出实体对齐检索方法，增强实体级语义匹配，提升长尾实体问答的准确率。
ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes
👍 25
Compact token sequences are essential for efficient 3D generation. However, existing 3D tokenizers typically organize latent representations either over spatial regions or as fixed-size sets of global tokens, both suffering sharp reconstruction degradation when compressed to extremely low token budg
中文介绍 提出ZipTok3D，采用紧凑标记前缀实现高保真3D令牌化。在极低令牌预算下，相比现有方法显著改进重建质量，适用于高效3D生成。
RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests
👍 24
Coding agents are now commonly evaluated on the SWE-bench family of benchmarks, whose tasks are built from curated GitHub issues: long, structured, and information-rich. Real user requests, however, are typically far shorter and less structured. To characterize this gap, we define a six-category inf
中文介绍 构建RealSWE基准，评估编码代理在真实用户请求（短小且非结构化）下的表现。定义了六类信息缺口，揭示与SWE-bench任务差异及对代理性能的影响。
CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation
👍 24
MLLM-based embedding models remain limited in compositional retrieval, often failing to distinguish scenes containing the same concepts but different attribute-object bindings. Yet the same backbone can resolve such distinctions when used as a cross-attentive reranker, motivating us to distill its c
中文介绍 提出CORE方法，通过重排序器蒸馏提升多模态大模型嵌入的组合推理能力。使嵌入模型能区分相同概念但不同属性绑定的场景，增强检索精度。
DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training
👍 23
Reinforcement Learning from Verifiable Rewards works well when a task has a programmatic checker, but most long-horizon agent domains have none. We work in the outcome-blind setting, where ground-truth success signals are not available. Multi-criteria rubrics are a popular way to supply such a rewar
中文介绍 针对长期任务缺乏可验证奖励的问题，提出DRACO框架，使用动态评分细则进行细粒度信用分配。在结果盲设下提供有效的训练信号，提升代理学习效果。
WorldReward: Reward Modeling for Camera-Conditioned World Models
👍 23
Camera-conditioned world models generate interactive videos in which commanded actions should induce the expected scene changes while appearance, geometry, and temporal dynamics remain coherent. Existing rewards assess these requirements separately: geometry-based rewards estimate trajectory executi
中文介绍 提出WorldReward，为相机条件世界模型设计统一奖励建模。整合几何与外观一致性要求，评估交互视频中动作引发的场景变化是否协调。
PACE: Towards Surfacing Hidden Conflicts in User Requests
👍 22
Personalized assistants should not only comply with user requests but also assess whether those requests are appropriate given the user's current circumstances. However, prior work has primarily focused on accurately executing requests, overlooking the need for assistants to account for context and
中文介绍 提出PACE方法，使个性化助手不仅执行用户请求，还能评估请求是否适应当前情境，发现其中隐含冲突，提高互动安全性。
Using Grounded Theory for Agent Behavior Analysis at Scale
👍 18
Understanding agent behavior requires methods that scale to thousands of trajectories and surface new patterns in long, often unfamiliar tasks where pre-built classifiers fall short. We propose to bring grounded theory into agent trajectory analysis: a six-decade-old qualitative method from the soci
中文介绍 将社会学扎根理论引入代理轨迹分析，用于大规模行为模式发现。该方法无需预构建分类器，能自动浮现新模式，适合长任务分析。
Environment Evolution for Terminal Agents
👍 17
Scaling interactive and verifiable environments is critical for training terminal agents. As frontier models become more capable, environments synthesized from scratch become less challenging and thus provide limited learning signals. Recent co-evolution methods iteratively synthesize environments n
中文介绍 针对终端代理训练环境难度不足的问题，提出环境演化方法，迭代合成更具挑战性的环境，为前沿模型提供持续学习信号。
Principia: Relational Physics Tests for Video Models
👍 16
Evaluating physical reasoning in video models is difficult because absolute motion measurements depend on frame rate, object scale, and camera calibration, all of which are often ambiguous or unavailable in generated video. We propose a different approach. When two objects in the same scene obey the
中文介绍 提出Principia基准，通过测试视频中物体间的相对运动是否符合物理关系（如碰撞与引力），避免依赖绝对运动测量，从而评估视频模型的物理推理能力。
Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs
👍 15
Long-video language models cannot look at every frame: an hour sampled once per second is 3,600 images, and a system keeps only a small fixed slice of that pool. Which frames survive that slice is usually treated as a preprocessing detail; we test whether it should be. Published selectors make the c
中文介绍 针对长视频MLLM的视觉令牌分配问题，进行受控研究。比较不同帧选择器，测试其对模型性能的影响，发现选择策略应被重视为重要环节。
Grok Imagine Odyssey Contest Winners
@imagine · 79.2K 粉丝 · 714.5K 阅 · 532 赞 · 86 转
Homer had a lyre. You had Grok Imagine. Thousands of you quoted our launch post and built original Odyssey scenes in Grok Imagine. Image, video, and the voices. Three to five minutes, with at least
中文介绍 Grok Imagine 举办《奥德赛》主题创作竞赛，数千人用图像、视频和声音功能构建原创场景，时长要求 3-5 分钟以上，展示多模态生成能力。
Five things I learned from using Astra
@dkundel · 23.9K 粉丝 · 225.2K 阅 · 520 赞 · 44 转
Back in February, getting Codex to build LEGO models took a custom skill, complicated instructions, and hours of iteration. And it was still not great. I came away thinking it wasn’t something models
中文介绍 博主总结使用 Astra 的五点心得：对比二月时让 Codex 搭乐高模型需自定义技能和复杂指令且效果不佳，现在工具进步明显，工作流大幅简化。
CarryGo, here we go!
@CarryGo_AI · 9 粉丝 · 192.8K 阅 · 1.0K 赞 · 2 转
Inside CarryGo lies a world model—helping organizations see, simulate, act and learn as a system. At the product level, CarryGo is a world model, built first for business. At the system level, it is
中文介绍 CarryGo 发布：底层为世界模型，帮企业实现「看见、模拟、行动、学习」的系统级能力，产品层先面向商业场景，系统层则更宏观。
Applied AI Doesn't Work
@vasuman · 96.1K 粉丝 · 109.9K 阅 · 517 赞 · 32 转
The world has spent a fortune on AI. For most enterprises, nothing changed. I've spoken to over 300 CEOs, CIOs and CFOs at the largest companies on Earth. I learned most companies apply AI onto their
中文介绍 作者与超 300 位 CEO/CIO/CFO 交流后指出：大多数企业只是把 AI 硬套进现有流程，导致巨额投入却几乎没带来改变，应用 AI 的方式需重构。
How I make content with Astra
@nickbaumann_ · 20.0K 粉丝 · 74.8K 阅 · 511 赞 · 38 转
I've been using Astra in Codex to edit videos we've shipped on the official ChatGPT Instagram. I've also been using it to pull clips from longer videos and create product animations. And what's really
中文介绍 博主分享用 Codex 中的 Astra 制作内容：剪辑官方 ChatGPT Instagram 视频、提取长视频片段、制作产品动画，效率显著提升。
AI Engineering Skills Map: Using coding agents
@AndrewYNg · 1.9M 粉丝 · 48.9K 阅 · 918 赞 · 136 转
A key AI engineering skill is using coding agents. Your skill at steering them both to write code and to carry out non-code tasks, such as analyzing data or managing system operations, allows you to
中文介绍 吴恩达提出关键 AI 工程技能是使用编码智能体：不仅写代码，还能用于数据分析、系统运维等非编码任务，驾驭能力决定产出上限。
Rethinking skills and prompts for GPT-6 Astra
@pvncher · 35.4K 粉丝 · 44.6K 阅 · 508 赞 · 37 转
Coding agents have come a long way, and best practices are changing fast. What used to require a lot of handholding and scaffolding no longer does. If you’ve been using agents in your projects over
中文介绍 针对 GPT-6 与 Astra 时代，作者认为编码智能体的最佳实践快速演变：过去需要大量人工引导和脚手架的工作，现在已不再需要，技能与提示词策略应更新。
Grok Bot for Work
@joshkim · 8.3K 粉丝 · 42.4K 阅 · 525 赞 · 41 转
I can't do my job without Grok Bot. It happened really quickly. A few weeks ago, I wasn’t leveraging agents like the AI-pilled experts you see on X and LinkedIn. I was still a human in the loop:
中文介绍 作者表示工作离不开 Grok Bot：几周前还不像专家那样用智能体，仍「人在回路」，如今已重度依赖，转变发生得很快。
Codex has changed my life as a solopreneur
@jonnym1ller · 20.7K 粉丝 · 36.0K 阅 · 510 赞 · 23 转
It's difficult to exaggerate how absurdly life-changing Codex has been for me as a solopreneur in recent months. And yet people have been saying 'but what do you actually use it for' (so I decided to
中文介绍 独立开发者分享 Codex 带来的巨大改变：近期作为一人公司获益匪浅，并回应「你到底拿它做什么」的疑问，详细列举实际用途。
OpenClaw Power, MacBook Simplicity: Five Days With Grok Bot
SpaceXAI’s Grok Bot has the same level of programming power as OpenClaw, but it’s programmable at a different level of abstraction.
中文介绍 Grok Bot 具有与 OpenClaw 同等的编程能力，但在不同的抽象层级上可编程，兼顾了类似 MacBook 的简易性。
Architecting memory and storage in the AI era
The era of AI inference has arrived. Imagine a healthcare system analyzing millions of data points in real time to accelerate life-saving medical research, or an intelligent assistant instantly resolving thousands of complex customer needs at once. These real-world breakthroughs rely on advanced inf
中文介绍 AI 推理时代已至，医疗系统可实时分析数百万数据点加速研究，智能助手可同时处理复杂客户需求，内存和存储架构正面临新挑战。
Data from drones in Ukraine is fueling a new Wild West marketplace
Battlefields in Ukraine are littered with the remnants of drones, which are now firmly established as a critical weapon of modern warfare. But behind all that wreckage, there’s a new gold mine for the defense sector. The data drones generate will far outlast the wars in which they are used to fight,
中文介绍 乌克兰战场无人机残骸遍布，其生成的数据远超战争本身寿命，为国防领域带来新金矿，催生野性西部式的数据市场。
[AINews] GPT-6 Astra: OpenAI’s biggest LLM launch of all time
new SOTA computer use and coding, 2.5x pricier per token, but WAY cheaper per task, less monitorable. overall, a very successful launch of OpenAI’s new frontier model class.
中文介绍 OpenAI 发布迄今最大规模模型 GPT-6 Astra，在计算机使用和编程上达到新 SOTA，每 token 价格贵 2.5 倍，但每任务成本更低、监控难度更高。
GPT-6 Astra 🚀, Grok Bot Enterprise 🤖, Runway world model 🌎
中文介绍 今日 AI 要闻包括 GPT-6 Astra 发布、Grok Bot 企业版推出以及 Runway 世界模型亮相。
GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour
We spent 20B+ tokens of GPT-6 Astra to explore everything. Here’s our learnings.
中文介绍 GPT-6 Astra 可作为自动化 AI 工程师，每小时成本低于 6 美元，作者花费超 200 亿 token 探索其能力并分享经验。
Introducing WeatherNext 3, our most advanced and accurate global weather AI model
中文介绍 DeepMind 推出 WeatherNext 3，号称其最先进、最准确的全球天气 AI 模型。
Daybreak for Frontline Defenders: $1B to protect essential services
OpenAI introduces Daybreak for Frontline Defenders. A $1 billion commitment expands access to frontier cyber AI, training, and support for essential services.
中文介绍 OpenAI 启动 Daybreak for Frontline Defenders 计划，承诺投入 10 亿美元，扩大前沿网络 AI 获取、培训和关键服务支持。
NeoMME: an efficient Multimodal-native and Multilingual Encoder
中文介绍 NeoMME 是一个高效的多模态原生多语言编码器，旨在提升多模态与多语言处理效率。
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.
中文介绍 Playco 使用 GPT-6 Astra 从灰色盒原型构建三款主题游戏原型，手动修复减少 50%。
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.
中文介绍 Legora 用 GPT-6 Astra 在几分钟内审查了 41 份文件，找出全部四个植入错误，财务审查工作流性能提升近 40%。
GPT-6 Astra: A new generation of intelligence
Introducing GPT-6 Astra, our most intelligent and aligned model yet, with state-of-the-art capabilities across computer use, coding, cybersecurity, and science.
中文介绍 OpenAI 发布 GPT-6 Astra，号称迄今最智能、最对齐的模型，在计算机使用、编程、网络安全和科学领域能力达 SOTA。
[AINews] Muse Spark 1.3 matches GPT-5.6-Sol, confirming Meta Superintelligence as the newest Frontier Lab, >90% discount for training
an epic comeback story for Meta
中文介绍 Meta 的 Muse Spark 1.3 匹配 GPT-5.6-Sol，确认 Meta Superintelligence 为新前沿实验室，训练成本折扣超 90%，成回归传奇。
Safety overview: GPT-6 Astra
GPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.
中文介绍 GPT-6 Astra 是最强大的广泛部署模型，据 Preparedness Framework，首次达到网络安全能力的「关键」级别。
Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps
中文介绍 通过 100 步 GRPO，微调 350M 模型以改善结构化输出性能。
当前市场技术面分化明显。美股方面，标普 500（SPY）和纳斯达克（QQQ）呈现多头排列，SPY 距 52 周高点仅 -1.18%，但苹果（AAPL）当日大跌 2.51%，特斯拉（TSLA）下挫 5.92%，科技股内部分化。加密市场在恐慌贪婪指数 73（贪婪）背景下，比特币（BTC）和以太坊（ETH）均出现 MACD 死叉，尽管价格仍高于中长期均线，短期动量减弱。中概股普遍偏弱，阿里巴巴（BABA）、拼多多（PDD）均呈空头排列，RSI 低于 45，技术面承压。商品外汇中，原油（CL=F）受 5 日 9.69% 涨幅推动强势上涨，而美元兑人民币（USDCNY=X）超卖（RSI 17.2）。宏观方面，10 年美债收益率（^TNX）向上突破，VIX 处于低位，市场风险偏好尚可。整体看，资产间相关性降低，应关注个股独立性。
今日关注
NVDA
英伟达 (NVDA)
偏上行
当前价 230.36，高于 SMA20（220.08）、SMA50（210.57）和 SMA200（196.68），形成多头排列。RSI 为 60.4，处于强势区间但未超买。MACD 柱为 3.83，且 MACD 线（3.83）高于信号线（2.99），呈金叉状态。近 5 日涨幅 5.89%，接近 52 周高点（-2.61%），表明短中期动量向上。
USDCNY=X
美元 / 人民币
偏下行
当前价 6.70，低于 SMA20（6.73）、SMA50（6.76）和 SMA200（6.86），呈空头排列。RSI 极低为 17.2，处于超卖区域，但 MACD 已死叉（MACD=-0.0117，信号=-0.0111），显示下行趋势延续。近 1 日跌 0.26%，近 5 日跌 0.37%，且接近 52 周低点，技术面整体偏空。
PDD
拼多多 (PDD)
偏下行
当前价 82.21，低于 SMA20（86.45）和 SMA50（85.61），且低于 SMA200（98.05），呈空头排列。RSI 为 39.0，低于 50，显示弱势。MACD 柱为负（-1.21），MACD 线低于信号线，死叉状态。近 5 日跌 4.06%，距 52 周高点 -41.03%，动量向下，整体技术面偏空。
CL=F
WTI 原油期货
中性偏多
当前价 91.48，远高于 SMA20（85.44）、SMA50（80.96）和 SMA200（78.81），多头排列。近 5 日涨幅显著达 9.69%，RSI 为 65.3 接近超买但未进入。MACD 金叉（MACD=2.26，信号=1.52），柱线为正。但价格距 52 周高点尚远（-23.43%），且 5 日涨幅过大，短期或有回调风险，故综合为中性偏多。
TSLA
特斯拉 (TSLA)
中性
当前价 354.08，略高于 SMA20（348.44）但低于 SMA50（357.95）和 SMA200（399.56），趋势为 bearish 空头排列。RSI 为 50.9 处于中性，MACD 柱为正（3.66）且 MACD 线高于信号线，略显金叉。但近 1 日大跌 5.92%，距 52 周高点 -29.02%，多空信号混杂，整体呈中性偏弱。
全部资产
^VIX
VIX 恐慌指数
$14.53
+1.47%
- 5 日
- +0.69%
- 距 52w 高
- -58.8%
- RSI(14)
- 44.9
- 趋势
- 空头
- SMA 20 / 50 / 200
- 15.10 / 16.20 / 18.25
- MACD / 信号
- -0.413 / -0.428
MACD 金叉 (4 天前)空头排列
^TNX
10Y 美债收益率 (%)
$4.78
+0.46%
- 5 日
- +1.36%
- 距 52w 高
- -0.6%
- RSI(14)
- 60.1
- 趋势
- 多头
- SMA 20 / 50 / 200
- 4.71 / 4.63 / 4.36
- MACD / 信号
- 0.038 / 0.032
MACD 金叉 (3 天前)接近 52 周高多头排列
DX-Y.NYB
美元指数 DXY
$99.16
+0.16%
- 5 日
- -0.54%
- 距 52w 高
- -2.6%
- RSI(14)
- 42.9
- 趋势
- 中性
- SMA 20 / 50 / 200
- 99.39 / 100.23 / 99.17
- MACD / 信号
- -0.242 / -0.286
接近 52 周高
SPY
S&P 500 ETF
$770.19
-0.39%
- 5 日
- +0.11%
- 距 52w 高
- -1.2%
- RSI(14)
- 55.6
- 趋势
- 多头
- SMA 20 / 50 / 200
- 769.05 / 756.86 / 712.15
- MACD / 信号
- 3.070 / 3.734
接近 52 周高多头排列
QQQ
Nasdaq 100 ETF
$718.96
+0.18%
- 5 日
- +0.35%
- 距 52w 高
- -4.0%
- RSI(14)
- 53.6
- 趋势
- 多头
- SMA 20 / 50 / 200
- 717.51 / 711.09 / 657.71
- MACD / 信号
- 1.183 / 1.464
多头排列
AAPL
Apple
$319.97
-2.51%
- 5 日
- +0.08%
- 距 52w 高
- -7.1%
- RSI(14)
- 53.9
- 趋势
- 多头
- SMA 20 / 50 / 200
- 313.14 / 315.15 / 283.87
- MACD / 信号
- 2.592 / 1.034
多头排列
MSFT
Microsoft
$499.70
-2.04%
- 5 日
- -2.69%
- 距 52w 高
- -9.8%
- RSI(14)
- 59.4
- 趋势
- 多头
- SMA 20 / 50 / 200
- 495.71 / 444.17 / 431.05
- MACD / 信号
- 15.260 / 17.641
金叉(SMA50↑SMA200) (4 天前)多头排列
NVDA
Nvidia
$230.36
+0.84%
- 5 日
- +5.89%
- 距 52w 高
- -2.6%
- RSI(14)
- 60.4
- 趋势
- 多头
- SMA 20 / 50 / 200
- 220.08 / 210.57 / 196.68
- MACD / 信号
- 3.831 / 2.994
MACD 金叉 (1 天前)接近 52 周高多头排列
GOOGL
Alphabet
$338.46
-1.17%
- 5 日
- -2.35%
- 距 52w 高
- -17.2%
- RSI(14)
- 44.8
- 趋势
- 中性
- SMA 20 / 50 / 200
- 343.61 / 348.57 / 335.94
- MACD / 信号
- -2.983 / -2.595
TSLA
Tesla
$354.08
-5.92%
- 5 日
- +1.53%
- 距 52w 高
- -29.0%
- RSI(14)
- 50.9
- 趋势
- 空头
- SMA 20 / 50 / 200
- 348.44 / 357.95 / 399.56
- MACD / 信号
- 3.659 / 0.733
空头排列
META
Meta
$616.77
+1.00%
- 5 日
- +6.70%
- 距 52w 高
- -22.0%
- RSI(14)
- 62.4
- 趋势
- 中性
- SMA 20 / 50 / 200
- 576.88 / 595.49 / 622.29
- MACD / 信号
- 1.900 / -4.970
BTC-USD
Bitcoin
$79,858.53
+0.23%
- 5 日
- +1.67%
- 距 52w 高
- -36.7%
- RSI(14)
- 66.9
- 趋势
- 中性
- SMA 20 / 50 / 200
- 76,463.64 / 69,095.45 / 69,684.17
- MACD / 信号
- 3,332.168 / 3,439.861
MACD 死叉 (1 天前)
ETH-USD
Ethereum
$2,482.60
+1.08%
- 5 日
- +0.64%
- 距 52w 高
- -47.9%
- RSI(14)
- 64.2
- 趋势
- 多头
- SMA 20 / 50 / 200
- 2,389.63 / 2,090.01 / 2,036.08
- MACD / 信号
- 119.766 / 132.374
金叉(SMA50↑SMA200) (5 天前)MACD 死叉 (4 天前)多头排列
SOL-USD
Solana
$103.30
+1.32%
- 5 日
- +0.29%
- 距 52w 高
- -59.2%
- RSI(14)
- 65.0
- 趋势
- 多头
- SMA 20 / 50 / 200
- 96.99 / 83.77 / 82.31
- MACD / 信号
- 6.128 / 6.454
金叉(SMA50↑SMA200) (3 天前)MACD 死叉 (1 天前)多头排列
BABA
阿里巴巴 (BABA)
$113.24
+1.28%
- 5 日
- -4.76%
- 距 52w 高
- -41.2%
- RSI(14)
- 40.9
- 趋势
- 空头
- SMA 20 / 50 / 200
- 120.98 / 116.17 / 135.46
- MACD / 信号
- -2.039 / -0.483
空头排列
PDD
拼多多 (PDD)
$82.21
+0.71%
- 5 日
- -4.06%
- 距 52w 高
- -41.0%
- RSI(14)
- 39.0
- 趋势
- 空头
- SMA 20 / 50 / 200
- 86.45 / 85.61 / 98.05
- MACD / 信号
- -1.208 / -0.441
空头排列
JD
京东 (JD)
$28.26
+1.87%
- 5 日
- -1.67%
- 距 52w 高
- -23.3%
- RSI(14)
- 39.7
- 趋势
- 中性
- SMA 20 / 50 / 200
- 29.27 / 29.47 / 29.13
- MACD / 信号
- -0.652 / -0.477
金叉(SMA50↑SMA200) (9 天前)
0700.HK
腾讯控股 (0700.HK)
HK$442.80
+2.26%
- 5 日
- -2.72%
- 距 52w 高
- -35.2%
- RSI(14)
- 45.7
- 趋势
- 空头
- SMA 20 / 50 / 200
- 448.90 / 455.21 / 516.89
- MACD / 信号
- -5.230 / -4.142
空头排列
GC=F
黄金期货
$4,476.60
-0.34%
- 5 日
- -0.03%
- 距 52w 高
- -19.9%
- RSI(14)
- 52.3
- 趋势
- 中性
- SMA 20 / 50 / 200
- 4,467.15 / 4,238.98 / 4,520.46
- MACD / 信号
- 56.031 / 79.642
MACD 死叉 (4 天前)
CL=F
WTI 原油期货
$91.48
+0.20%
- 5 日
- +9.69%
- 距 52w 高
- -23.4%
- RSI(14)
- 65.3
- 趋势
- 多头
- SMA 20 / 50 / 200
- 85.44 / 80.96 / 78.81
- MACD / 信号
- 2.265 / 1.518
MACD 金叉 (3 天前)多头排列
USDCNY=X
美元 / 人民币
¥6.70
-0.26%
- 5 日
- -0.37%
- 距 52w 高
- -6.8%
- RSI(14)
- 17.2
- 趋势
- 空头
- SMA 20 / 50 / 200
- 6.73 / 6.76 / 6.86
- MACD / 信号
- -0.012 / -0.011
MACD 死叉 (今天)RSI 超卖接近 52 周低空头排列
以上内容仅为基于公开数据的技术指标描述与解读，不构成任何投资建议。技术指标反映过去走势，过去走势不代表未来表现，实际行情可能因宏观事件或市场情绪发生反转。请读者结合自身情况独立判断，仅供技术指标解读参考。
Australia news live: Taylor proposes laws against flag burning and to lock in 26 January as Australia Day
Follow today’s news live Get our breaking news email, free app or daily news podcast Duniam targets ‘frivolous’ asylum seeker claims The opposition is “very close” to “having a full migration package”, Duniam says. He says they’re willing to work with the government and that illegal overstayers must
中文摘要 澳大利亚反对党提出立法禁止焚烧国旗，并将1月26日固定为澳大利亚日。泰然党参议员邓尼亚姆表示，反对党接近形成完整的移民方案，愿意与政府合作。
Aviation faces hotter, stormier skies – and passengers might have to accept more disruption
Are people going to have to increasingly accept more delays and diversions as the effects of climate change intensify?
中文摘要 报道称，气候变化加剧，航空业面临更炎热、更多风暴的天空，乘客可能不得不接受更多延误和改道，需适应气候变化带来的影响。
Fuel tanker blast in western Iran kills at least 10
At least six other people were injured in the tanker blast on Iran's Hamedan-Sanandaj highway, officials say.
中文摘要 伊朗西部哈马丹-萨南达杰高速公路上发生油罐车爆炸，造成至少10人死亡，另有至少6人受伤。官员证实了上述消息。
Trump-backed group announces $10m ad spend for midterm Senate race in Texas
US president has pledged more than $400m through his super PAC to boost Republican candidates in the upcoming midterms.
中文摘要 特朗普支持的团体宣布在得克萨斯州中期选举参议院竞选中投入1000万美元广告费。总统通过其超级政治行动委员会承诺提供超过4亿美元，助力共和党候选人。
US and Iran trade retaliatory attacks on ships as conflict flares
The US says it hit three Iranian oil tankers, while Iran says it targeted three US-linked ships.
中文摘要 美国与伊朗在海上发生报复性袭击，美国称击中三艘伊朗油轮，伊朗则称袭击了三艘与美国有关的船只，冲突升级。
Huckabee Condemns ‘Terrorist’ Attacks on Palestinians in West Bank Town Hall
Mike Huckabee, the U.S. ambassador to Israel, made a rare trip to meet with Palestinians after international backlash over attacks by Israeli settlers.
中文摘要 美国驻以色列大使赫卡比罕见访问约旦河西岸，与巴勒斯坦人会面，谴责以色列定居者对巴勒斯坦人的“恐怖”袭击，此前国际社会对此类袭击强烈反弹。
Trump’s pique with UK could lead to favor to acolyte Milei over Falklands
US president suggests he might support his Argentinian ally’s drive to reclaim sovereignty of disputed islands Donald Trump has long embraced vengeance. “I am your retribution,” he declared to an admiring audience at the 2023 Conservative Political Action Conference (CPAC) as he commenced his march
中文摘要 美国总统特朗普暗示可能支持阿根廷总统米莱对福克兰群岛主权的要求，此举或与对英国不满有关，可能给米莱带来外交助力。
Drone attack blamed on guerrillas kills at least three soldiers in Colombia
Officials accuse National Liberation Army of launching strike followed by clash inside military compound An attack involving explosives dropped from drones followed by a clash inside a military compound in north-eastern Colombia has killed at least three soldiers and wounded four, Colombia’s army sa
中文摘要 哥伦比亚东北部一座军事基地遭无人机投掷爆炸物袭击，随后发生交火，造成至少3名士兵死亡、4人受伤，当局指责民族解放军所为。
Putin begins talks with Steve Witkoff and Jared Kushner on ending war with Ukraine
Moscow and Kyiv vow not to strike each other’s capitals for three days during negotiations Russian president Vladimir Putin began talks with US envoys Jared Kushner and Steve Witkoff in the Kremlin on Saturday – aimed at ending more than four years of war in Ukraine, ahead of the pair’s first visit
中文摘要 俄罗斯总统普京在克里姆林宫与美国特使威特科夫和库什纳会谈，旨在结束持续四年多的乌克兰战争。莫斯科和基辅承诺谈判期间三天内不打击对方首都。
Prince William to represent King Charles at funeral of Norway’s King Harald V
Duty falls by precedent to William to attend service of late monarch, alongside Princess Royal, who is godmother to new king Haakon VIII The Prince of Wales is to represent his father at the funeral of King Harald V of Norway next week, Buckingham Palace has announced. William, the heir to the thron
中文摘要 白金汉宫宣布，威尔士亲王将代表其父查理三世出席下周挪威国王哈拉尔五世的葬礼。安妮公主作为新国王哈康八世的教母也将出席。
US envoys meet Putin in Moscow for Ukraine talks
Footage shows Vladimir Putin greeting Steve Witkoff and Jared Kushner, before saying the current situation "is not so easy".
中文摘要 美国特使威特科夫和库什纳在莫斯科与普京会晤，讨论乌克兰问题。视频显示普京迎接二人，并表示当前局势“不那么容易”。
TV presenter among 11 sentenced to death in Egypt drugs case
Sarah Khalifa is best known for her TV programme Mission Impossible which dealt with crime issues.
中文摘要 埃及一毒品案中，11人被判处死刑，其中包括知名电视主持人萨拉·哈利法，她以主持犯罪类节目《不可能的任务》闻名。
US envoys Witkoff, Kushner meet Putin for talks on Ukraine war
US envoys Steve Witkoff and Jared Kushner met Russian President Vladimir Putin at the Kremlin on Saturday.
中文摘要 美国特使威特科夫和库什纳于周六在克里姆林宫与俄罗斯总统普京会晤，就乌克兰战争进行会谈。
Kenyans Made a Living Writing College Essays. Then A.I. Arrived.
Thousands of Kenyans made a living writing essays for overseas students. With A.I., the work has dried up, a warning for online gig work that has been a global lifeline.
中文摘要 大量肯尼亚人曾以代写留学申请文书为生，但人工智能的出现使此类工作枯竭，凸显在线零工经济面临的挑战，这种工作曾是全球生计来源。
Building collapses after heavy rain in northern India
A building collapsed in Muradabad, northern India, tearing through power lines and sending sparks into the street.
中文摘要 印度北部穆拉达巴德一栋建筑因暴雨倒塌，损坏电线并导致火花四溅，暂无人员伤亡报道。
US regulator sues ISS as it steps up scrutiny of proxy advisers
The SEC accuses the firm of ignoring a subpoena as part of its ‘fact-finding investigation’
中文摘要 美国证交会起诉代理顾问公司ISS，指控其忽视传票，作为其「事实调查」的一部分。此举反映美国监管机构加大对代理顾问的审查力度。
US strikes three Iranian oil tankers in response to attacks on warships
Military action over Strait of Hormuz continues to escalate after pause in six-month-old conflict
中文摘要 美国对三艘伊朗油轮发动军事打击，以回应其对美军舰的袭击。霍尔木兹海峡相关军事行动在持续六个月的冲突暂停后再度升级。
Trump Renames Lake Ontario, Canada Responds
President Trump’s decision to rename Lake Ontario to Lake America is making waves on both sides of the border. Bloomberg News Shaquille Omari looks at how Canada and tech companies are responding. (Source: Bloomberg)
中文摘要 美国总统特朗普决定将安大略湖更名为「美国湖」，引发美加两国关注。彭博新闻记者Shaquille Omari报道加拿大及科技公司的反应。
Bloomberg This Weekend 09/05/2026
The news doesn’t stop when markets close. Hosts David Gura, Christina Ruffini and Lisa Mateo bring clarity, context and a bit of humor to the weekend’s biggest headlines, LIVE from New York. (Source: Bloomberg)
中文摘要 彭博周末节目于2026年9月5日播出，主持人大卫·古拉、克里斯蒂娜·鲁菲尼和丽莎·马特奥在纽约现场解读本周末重大新闻，提供清晰背景和幽默视角。
Jobs Surge Raises Odds of September Fed Hike
On Bloomberg This Weekend, Seaport Research Partner Managing Director & Chief Equity Strategist Jonathan Golub says a stronger-than-expected US jobs report reinforces the case for the Federal Reserve to raise interest rates, despite President Donald Trump’s calls for lower borrowing costs. Golub als
中文摘要 美国就业数据超预期，提升美联储9月加息可能性。Seaport Research董事总经理乔纳森·戈卢布称，尽管特朗普呼吁降低借贷成本，强劲就业报告仍支持加息。
Masked protesters block Dover port roads
Port operator says services back to normal after ‘public order incident’ affected ‘all routes in and out’
中文摘要 蒙面抗议者堵塞多佛港道路，影响所有进出路线。港口运营商表示「公共秩序事件」后服务已恢复正常。
The Hidden Cost of America’s Bureaucracy
In her new book The Time Tax, Atlantic staff writer Annie Lowrey examines how paperwork, delays and administrative hurdles consume Americans’ time and can make government benefits and services harder to access. Joining Christina Ruffini and Jeff Mason on Bloomberg This Weekend, Lowrey argues those b
中文摘要 大西洋月刊撰稿人安妮·洛瑞在新书《时间税》中探讨文书、延误和行政障碍如何消耗美国人时间，使政府福利和服务更难以获取。她做客彭博周末节目。
US Strikes Iranian Tankers After Attack on US Warships
Bloomberg This Weekend has breaking news that US forces struck three Iranian crude carriers after the Islamic Revolutionary Guard Corps launched ballistic missiles toward two US Navy warships, according to US Central Command. The renewed fighting around the Strait of Hormuz adds pressure to shipping
中文摘要 美国中央司令部称，美军袭击三艘伊朗原油运输船，此前伊斯兰革命卫队向两艘美军舰发射弹道导弹。霍尔木兹海峡冲突再起。
Esper Raises Concern Over Pentagon Leadership Loss
On Bloomberg This Weekend Former Defense Secretary Mark Esper calls the Pentagon’s polygraph testing of roughly 50 Joint Staff members unprecedented in scale as officials investigate leaks about US munitions stocks. Esper tells hosts Christina Ruffini and Jeff Mason that he is less concerned about i
中文摘要 前国防部长马克·埃斯珀对五角大楼约50名联合参谋部成员进行测谎测试表示担忧，称其规模空前。此举系调查美国弹药库存泄密事件。
EU Envoy Urges Maximum Pressure on Russia
EU Ambassador to the United States Jovita Neliupšienė joins Bloomberg This Weekend and discusses with hosts Christina Ruffini and Jeff Mason that Europe must keep supporting Ukraine while applying “maximum pressure” on Moscow because Russia has not changed its war aims. The comments come as US envoy
中文摘要 欧盟驻美大使约维塔·内柳普希耶涅表示，欧洲必须继续支持乌克兰，并对莫斯科施加「最大压力」，因为俄罗斯未改变其战争目标。
By The Way: Headlines You Should See
Canned cocktails, the science behind the Gen Z stare and the US Open becoming a food fest: Join Bloomberg's Lisa Mateo, Christina Ruffini and Jeff Mason to hear about the headlines from this week you may have missed but need to check out. Watch, on Bloomberg This Weekend! Live Saturday and Sunday, 7
中文摘要 彭博周末节目探讨本周你可能错过的头条：罐装鸡尾酒、Z世代凝视背后的科学，以及美网成为美食节。主持人为丽莎·马特奥、克里斯蒂娜·鲁菲尼和杰夫·梅森。
AI May Transform How Humans Understand Animals
Researchers are using AI to analyze vast collections of animal vocalizations, looking for patterns that could help humans better understand how species such as crows communicate. Bloomberg Weekend Senior Writer Morgan Meaker is on Bloomberg This Weekend and says the technology could reshape humans’
中文摘要 研究人员利用人工智能分析大量动物发声，寻找模式以帮助人类理解乌鸦等物种的交流。彭博周末高级撰稿人摩根·米克称该技术可能改变人类对动物的理解。
难绷，太相信 ai 了😭
8 回复 · 程序员 节点
一个 MonoRepo 复杂系统的部署问题
20 回复 · 程序员 节点
给我整乐呵了，商汤上了 deepseek v4 flash 和 pro
24 回复 · 程序员 节点
认真请教一个问题，关于菲区 GPT PRO
12 回复 · 程序员 节点
我又又又重构了我的站点
30 回复 · 程序员 节点
求教贴， MBP 和 Mini 哪个更好一些
27 回复 · Apple 节点
想入一台港版或者澳版台版的 iPhone1 卡槽+esim
11 回复 · Apple 节点
想入一台 美版有锁 iPhone air
23 回复 · Apple 节点
电子产品大涨价，大家还会正常更新数码设备吗？
49 回复 · Apple 节点
MAC Studio M5 Ultra 被延期发货
30 回复 · Apple 节点
该源今日无内容。