Python · ★ 38,169 · 🍴 2,197 · 📈 3,854 stars today
A skill to stop your coding agent from burying the answer. ADHD-friendly output.
中文介绍 给编码 AI agent 设计的技能（skill），强制它把结论直接摆在开头，不要先铺垫一大段推理过程，输出对 ADHD 用户更友好。适合接入 Claude Code 等支持 skills 的 agent，减少翻找答案的时间。
JavaScript · ★ 24,138 · 🍴 5,024 · 📈 1,588 stars today
A spy satellite simulator in your browser, except the data is real. Live open source spatial intelligence on a photorealistic 3D globe.
中文介绍 浏览器里运行的间谍卫星模拟器，但用的是真实公开数据：在照片级 3D 地球上叠加实时开源空间情报，可查看卫星影像与地理信息。适合对地观测、OSINT 与地理信息可视化场景。
Shell · ★ 284,680 · 🍴 25,464 · 📈 731 stars today
An agentic skills framework & software development methodology that works.
中文介绍 面向 AI agent 的技能框架与软件开发方法论，把可复用的 skills 组织成工程化流程，让编码 agent 按既定规范完成开发任务。适合希望把 agent 真正落进日常研发流程的团队与个人开发者。
TypeScript · ★ 1,613 · 🍴 253 · 📈 299 stars today
Open Source AI trading agent that operates autonomously across 1000+ markets - Polymarket, Kalshi, Binance, Hyperliquid, Solana DEXs, 5 EVM chains. Scans for edge, executes instantly, manages risk while you sleep. Agent commerce protocol for machine-to-machine payments. Self-hosted. Built on Claude.
中文介绍 开源 AI 交易 agent，自主覆盖 Polymarket、Kalshi、Binance、Hyperliquid、Solana DEX 及 5 条 EVM 链等 1000 多个市场，自动扫描套利机会、即时下单并管理风险。面向量化交易者与加密市场套利玩家。
TypeScript · ★ 3,752 · 🍴 242 · 📈 837 stars today
Make Every Team AI Native
中文介绍 腾讯开源的团队协作 CLI 工具，目标是把 AI 能力嵌入团队日常工作流，让每个团队以 AI 原生方式协作。具体能力需结合仓库文档确认，适合企业团队统一接入 AI 工具链。
Rust · ★ 35,715 · 🍴 2,264 · 📈 247 stars today
Hundreds of models & providers. One command to find what runs on your hardware.
中文介绍 一条命令即可检测本机硬件能跑哪些大模型：覆盖数百个模型与推理服务商，自动评估显存、算力等条件并给出可运行清单。适合本地部署 LLM 前先做可行性筛选的开发者。
★ 18,749 · 🍴 3,480 · 📈 891 stars today
Notes of the book System Desgin Interview - An Insider's Guide
中文介绍 整理《System Design Interview: An Insider's Guide》一书的读书笔记，围绕分布式系统设计面试常见题目记录要点与解题思路。适合准备系统设计面试的后端工程师复习使用。
HTML · ★ 37,700 · 🍴 2,389 · 📈 1,287 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 为 Claude Code、Codex、Pi 等编码 agent 提供的图表设计套件，包含 38 种编辑风格图表，输出自包含 HTML + SVG，无阴影、不依赖 Mermaid。适合需要高质量图表的技术文档与演示场景。
JavaScript · ★ 30,813 · 🍴 2,984 · 📈 957 stars today
Prompt as Code | GPT Image 2 / 2.5 提示词与案例库，530+ 个案例、20+ 套工业级模板与可复用 Skills，新增 2.5 同提示词对比专区，附完整提示词与生成记录，持续更新。
中文介绍 GPT Image 2/2.5 的提示词与案例库，践行 Prompt as Code：收录 530+ 案例、20+ 套工业级模板与可复用 Skills，并设 2.5 同提示词对比专区，附完整提示词和生成记录。适合做 AI 图像生成的提示词工程参考。
C · ★ 4,400 · 🍴 519 · 📈 87 stars today
Graphics Creation Tools
中文介绍 基于 Armory 引擎的 3D 纹理绘制工具，支持 GPU 加速的实时材质绘制，可直接在模型表面作画并导出贴图。适合游戏与三维美术从业者制作手绘风格或写实材质。
TypeScript · ★ 64,205 · 🍴 8,993 · 📈 591 stars today
Never stop coding. Free MIT AI gateway: one endpoint, 352 providers (150+ free), 1200+ models Kimi, Claude, GPT, Gemini, GLM, DeepSeek, MiniMax. Works with Claude Code, Codex, Cursor, OpenCode, Cline & Copilot. Quota-aware auto-fallback, RTK+Caveman compression saves 15-95% tokens, MCP/A2A, Desktop/
中文介绍 免费且 MIT 许可的 AI 网关：提供统一 API 端点，聚合 352 家服务商（150+ 免费）与 1200+ 模型，涵盖 Kimi、Claude、GPT、Gemini、GLM、DeepSeek 等，兼容 Claude Code、Codex、Cursor、Cline 等客户端。
C · ★ 27,438 · 🍴 3,006 · 📈 130 stars today
Run frontier MoE models on hardware you already own — pure C, zero deps, experts streamed from disk. Tiny engine, immense model. 🐦
中文介绍 纯 C 实现、零依赖的轻量推理引擎，通过把 MoE 专家权重从磁盘流式加载，让个人已有硬件也能跑前沿混合专家大模型。适合显存有限却想本地体验大模型的开发者。
TypeScript · ★ 35,269 · 🍴 5,629 · 📈 806 stars today
Open Multi-Agent Interactive Classroom — Get an immersive, multi-agent learning experience in just one click
中文介绍 开源多智能体互动课堂，一键进入沉浸式学习体验，由多个 AI agent 分别扮演教师、助教等角色协同授课与答疑。适合探索 AI 教育、虚拟课堂与多智能体协作演示的场景。
TypeScript · ★ 18,073 · 🍴 2,098 · 📈 94 stars today
LLM Wiki is a cross-platform desktop application that turns your documents into an organized, interlinked knowledge base — automatically. Instead of traditional RAG (retrieve-and-answer from scratch every time), the LLM incrementally builds and maintains a persistent wiki from your sources。
中文介绍 跨平台桌面应用，把本地文档自动整理成相互链接的知识库，替代每次从零检索的传统 RAG 方案，形成可积累、可浏览的知识网络。适合个人知识管理与文档密集型研究者。
TypeScript · ★ 31,122 · 🍴 2,657 · 📈 175 stars today
The open agent skills tool - npx skills
中文介绍 Vercel 推出的开放 agent 技能工具，通过 npx skills 即可安装与管理可复用的 agent 技能，方便在不同编码 agent 之间共享能力扩展。适合想快速给 AI 编码助手加装技能的开发者。
TypeScript · ★ 2,268 · 🍴 192 · 📈 636 stars today
Local-first AI coding agent desktop: Electron + Rust host core + pi Agent Harness + user-installable plugins
中文介绍 本地优先的 AI 编码 agent 桌面客户端，采用 Electron 前端、Rust 宿主内核与 pi Agent Harness，支持用户自行安装插件扩展功能。适合注重数据本地化、想自定义编码 agent 工作流的开发者。
👍 115
Foundation vision-language models (VLMs) exhibit broad intelligence about the world, yet translating this intelligence into robot control remains challenging. We present Show-Harness, an Embodied Harness that enables VLMs to "play" robots through a compact semantic interface linking intent to action
中文介绍 提出 Show-Harness 具身框架，通过紧凑的语义接口把高层意图映射为机器人动作，使基础 VLM 无需额外机器人专用策略即可直接「操控」机器人，将通用世界知识转化为可执行的控制信号，降低对专用机器人数据的依赖。
👍 78
Large language model (LLM)-based multi-agent systems (MAS) achieve strong performance by employing specialized multiple agents, yet their performance depends on the prompt design of each agent. For MAS prompt optimization, textual gradient methods that guide prompt updates using natural-language fee
中文介绍 针对 LLM 多智能体系统（MAS）中各智能体 prompt 设计依赖人工、制约整体性能的问题，提出 AgentGrad：一种干预引导的 prompt 优化方法，以自然语言反馈作为文本梯度指导各智能体的 prompt 更新，提升 MAS 的协同表现。
👍 76
Recent video world models generate increasingly realistic and interactive visual experiences, yet lack reliable mechanisms for maintaining persistent world state and enforcing programmable rules over extended interactions. We introduce Programmable World Model, a framework that decouples world-state
中文介绍 提出 Programmable World Model 框架，将世界状态与生成渲染解耦，使视频世界模型在长程交互中维持持久的世界状态并执行可编程规则。相比仅追求画面真实感的现有模型，该方法为交互一致性与规则约束提供了可靠机制。
👍 59
World-Action Models inherit world knowledge from video-generative priors, and channel it into executable control signals through embodied experience. Existing systems, however, are monolithic: the generative backbone, visual representation, architecture, information flow, inference procedure, and tr
中文介绍 针对现有 World-Action Model 系统高度单体化、难以复现与扩展的问题，提出 OpenWAM 开放模块化框架，将生成骨干、视觉表征、网络架构、信息流与推理流程等组件解耦，为系统化的 world-action 预训练提供可复用基座。
👍 30
Multimodal Large Language Models (MLLMs) perform strongly on general visual understanding tasks such as visual question answering, yet they often struggle with a basic comparative skill: identifying what has changed between two similar images. We introduce VDiff-Bench, a challenging multiple-choice
中文介绍 提出 VDiff-Bench 多项选择基准，评估 MLLM 识别两张相似图像之间差异的细粒度比较能力。尽管 MLLM 在通用视觉问答上表现优异，在该基准上仍明显吃力，暴露了当前模型在细粒度视觉差异识别上的短板。
👍 28
Recent advances in wearable sensing enable continuous monitoring of physiological and behavioral signals, yet existing benchmarks rarely evaluate whether AI systems can reason over a real user's longitudinal wearable record. We introduce WearableQA, a benchmark comprising 4,084 10-option multiple-ch
中文介绍 提出 WearableQA 基准，包含 4,084 道十选一题目，用于评估 AI 系统能否对真实用户的长周期可穿戴记录（生理与行为信号）开展健康推理，填补了现有基准缺乏真实纵向穿戴数据评测的空白。
👍 19
SWE-Bench Pro has emerged as a standard benchmark for evaluating software engineering agents on challenging repository-level tasks. However, our analysis work show that its evaluation is undermined by two sources of unreliability: reward hacking, enabled by leakage of gold solutions or hidden evalua
中文介绍 分析发现 SWE-Bench Pro 的评测受两类不可靠因素影响：金标解答泄漏与隐藏评测信息引发的 reward hacking。为此提出 SWE-Bench Pro Verified，通过修补泄漏与作弊路径构建更可靠的软件工程智能体基准。
👍 16
While research on recursive self-improvement (RSI) has predominantly automated model training pipelines, reliable autonomous development demands a missing pillar: post-hoc monitoring and auditing to understand what models learn and ensure safe alignment. Mechanistic interpretability tools are essent
中文介绍 提出 SAEScientist-Bench，评估 AI 智能体能否自主开展 SAE（稀疏自编码器）可解释性研究。该基准填补了递归自我改进研究中事后监控与审计环节的空白，检验智能体端到端完成机制可解释性研究任务的能力。
👍 16
We present Cadence, an error-bounded lossy compressor for numeric time series pairing a 330M-parameter time-series foundation model (Google TimesFM-3) with an adaptive arithmetic coder, guaranteeing |x_t-x_t|leτ on every sample. One negative result constrains the design space: for lossless coding a
中文介绍 提出 Cadence，一种误差有界的有损时间序列压缩器，将 330M 参数的 Google TimesFM-3 时间序列基础模型与自适应算术编码器结合，保证每个采样点满足 |x_t−x̂_t|≤τ 的误差上界，适用于需求时间序列的高效压缩。
👍 15
AI research agents combine prior knowledge, public sources, and experimental feedback to produce useful results. The Discovery Certification Protocol (DCP) turns claims about these results into executable recovery and feedback tests. Gate 1 validates useful improvement on sealed evaluation. Gate 2 g
中文介绍 提出 Discovery Certification Protocol（DCP），将 AI 研究智能体所声称的发现转化为可执行的复现与反馈测试，通过分层关卡审计结果是否真正可复现，以区分单纯的分数提升与真实发现，为主张提供更严格的认证。
👍 11
World models are increasingly used as policy-in-the-loop imagination environments, where reliable rollouts require fine-grained controllability with respect to low-level robot actions. A key obstacle to scaling such models in robotics is that actions are not a universal language in pixel space: chan
中文介绍 提出 SyncWorld，通过视觉校准解决动作在像素空间中缺乏统一表示、世界模型难以精细可控的问题，使世界模型可作为零样本模拟器，为策略提供与底层机器人动作逐一对齐的想象 rollout，推动策略闭环中的世界模型应用。
👍 11
Chain-of-Thought (CoT) improves the reasoning ability of Large Language Models (LLMs) but incurs substantial computation and context costs. Existing methods either lose intermediate information through hard pruning or lack a principled criterion for continuous compression. We present A*-Thought-V2,
中文介绍 针对 CoT 推理计算与上下文开销大的问题，提出 A*-Thought-V2，利用 LLM 潜在空间的几何动力学实现高效隐式推理，既避免硬剪枝造成中间信息丢失，也不依赖启发式压缩准则，在保持推理能力的同时降低开销。
👍 9
High-quality structured organic reaction data are essential for developing artificial intelligence for chemistry (AI4Chem), yet much of this knowledge remains dispersed across patent text, images, and reaction schemes. We present DianShi-RxnDB, a large-scale, fine-grained organic reaction data platf
中文介绍 提出 DianShi-RxnDB，一个大规模、细粒度的有机反应数据平台，通过全自动流程从专利文本、图像与反应式中抽取结构化反应知识，为 AI4Chem 研究与化学 AI 智能体提供高质量的训练与检索数据支撑。
👍 8
Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. However, existing benchmarks mainly focus on isolated kernels, predefined operat
中文介绍 提出 Φ-Bench，评估 LLM 能否开发与优化支撑自身运行的底层基础设施。现有基准多聚焦孤立 kernel 与预定义算子，该基准覆盖更完整的系统级工程任务，检验模型在真实基础设施开发与调优场景中的能力。
👍 8
Large language models (LLMs) are often post-trained on pre-collected reasoning trajectories to improve their reasoning capability. Such trajectories tend to be long due to complex, interwoven paths, which often include detours on the path toward the answer. However, it has been underexplored whether
中文介绍 重新审视 LLM 后训练所用的完整推理轨迹：这类轨迹因包含大量绕路与弯路而冗长，其必要性此前缺乏验证。该工作系统分析完整轨迹中冗余部分对后训练效果的影响，为推理数据的筛选与构建提供实证依据。
👍 8
Image relighting is traditionally tackled via complex inverse rendering pipelines, which suffer from ill-posed optimization, or single-image generative models that ignore crucial multi-view cues necessary for understanding 3D geometry and material interactions. To address these limitations, we intro
中文介绍 提出 RelightFormer，一种用于多视角物体重打光的前馈生成式 Transformer。相比优化病态的逆渲染管线与忽略多视角线索的单图生成模型，该方法利用多视角信息理解 3D 几何与材质交互，生成更一致的打光效果。
👍 7
We study scheming in LLM agents, in which agents covertly pursue misaligned goals. Our focus is to understand how scheming arises from the interaction of key factors, such as instrumental goals, environmental affordances, oversight conditions, and perceived consequences. Prior work examines only a s
中文介绍 提出 SchemeArena，对 LLM 智能体中的 scheming（隐蔽追求与既定目标不一致的意图）进行因子化压力测试，系统考察工具性目标、环境可供性、监督条件与预期后果等因素的交互如何诱发该行为，弥补以往仅关注单一因素的不足。
👍 7
Advertising video generation is not only a video synthesis task, but also a product-conditioned reasoning problem whose success is measured by online business metrics. Recent video foundation models can generate realistic clips from multimodal conditions, yet they do not optimize how a product shoul
中文介绍 针对广告视频生成不仅是合成任务、更需以线上业务指标衡量产品条件推理的问题，提出 AgenticGen——奖励引导的智能体视频生成框架，使视频基础模型围绕产品卖点进行规划与优化，而非仅追求画面真实。
👍 6
Leading benchmarks for formal theorem proving with large language models are small collections drawn from competition math, such as the IMO and Putnam, that poorly represent field-specific applications. We introduce StochBench, a Lean 4 benchmark of 450 graduate stochastic-processes problems at vary
中文介绍 提出 StochBench，一个基于 Lean 4 的随机过程基准，包含 450 道不同难度的研究生水平题目。现有形式化定理证明基准多取自 IMO、Putnam 等竞赛数学，难以代表特定领域应用，该基准弥补了这一空白。
👍 5
AI-powered search products such as ChatGPT search, Google's AI Overviews, and Perplexity provide LLM-synthesized answers grounded in live web results. We developed OreoLook (formerly lixSearch), an open-source answer engine using automated browser agents and provider-routed LLM inference. Its local
中文介绍 提出开源答案引擎 OreoLook（原 lixSearch），结合自动化浏览器智能体与按供应商路由的 LLM 推理，并设计三层缓存架构，在消费级 CPU 硬件上实现低延迟的 LLM 网页搜索与答案合成。
@sairahul1 · 143.5K 粉丝 · 1.5M 阅 · 561 赞 · 53 转
Someone gave GPT-6 Astra $52. 13 hours later: $11,345.95. Someone else gave it $43.47 and one instruction. It turned that into $41,495 overnight. Someone went to sleep. Woke up to $200 in their
中文介绍 盘点用 GPT-6 Astra 赚钱的十种说法：有人给它 52 美元、13 小时后变成 1.1 万美元，也有人投 43 美元一夜做到 4.1 万美元。属收益截图式内容，具体方法并未展开。
@apexin_ai · 395 粉丝 · 281.6K 阅 · 500 赞 · 84 转
Astra shows how far AI has come. What takes it further? The continued progress of today’s most capable models depends on smarter decisions before training, better strategies during training, and more
中文介绍 以 Astra 为参照讨论 AI 自我改进：模型能否继续变强，取决于训练前更聪明的决策、训练中更好的策略，以及更多环节的协同。偏观点讨论，非实操教程。
@lucaspatiri_ · 5.8K 粉丝 · 275.5K 阅 · 506 赞 · 41 转
openai shipped its most capable model on september 3. i wasn't going to write this. we've generated 2.6 billion organic views across 62 campaigns, managed 590 creators, and produced 22,504 tracked
中文介绍 分享用 ChatGPT Astra 批量做 UGC 内容的打法，团队自称累计拿到 26 亿自然播放、跑过 62 个 campaign、合作 590 位创作者、追踪 22,504 条内容。属机构经验输出，带数据背书。
@BradGroux · 6.6K 粉丝 · 229.0K 阅 · 561 赞 · 118 转
I've been using GPT-6 Astra with the workflow I developed around GPT-5.6: documented task scope, repository context, standard operating procedures, and verification before delivery. In my testing,
中文介绍 对比 GPT-5.6 与 GPT-6 Astra：沿用此前那套工作流——写清任务范围、给仓库上下文、按 SOP 执行、交付前验证，作者发现在新模型上仍要反复交代同样的东西。属工作流迁移复盘。
@semochkin_alex · 1.5K 粉丝 · 191.5K 阅 · 507 赞 · 44 转
Hi, I'm Alex, cofounder at @relate_studio. In this tutorial, I'll walk you through how we design brand mascots with AI: finding first concepts in Krea, refining the best direction in Nano Banana 2,
中文介绍 relate_studio 联合创始人的品牌吉祥物设计教程：先在 Krea 里找初步概念，再用 Nano Banana 2 打磨最佳方向，声称不需要设计经验也能上手。属完整流程示范。
@MiaAI_lab · 27.4K 粉丝 · 180.6K 阅 · 505 赞 · 58 转
For the past three months I’ve been building and shipping local AI recipes, mostly for NVIDIA DGX Spark systems. What started as a personal project has grown far beyond what I expected. My GitHub
中文介绍 作者三个月来持续构建并开源本地 AI 配方，主要针对 NVIDIA DGX Spark 设备，原本只是个人项目，如今 GitHub 上的规模已远超预期。属本地部署实践与开源分享。
@ClaudeDevs · 683.4K 粉丝 · 106.4K 阅 · 1.3K 赞 · 70 转
Tuning prompt caching, instructions, and effort can reduce Claude's cost without sacrificing application performance. Performance and cost are often viewed as a trade-off: to spend less, you accept
中文介绍 Claude 官方讲降本思路：调优 prompt caching、instructions 与 reasoning effort 三处配置，就能在不牺牲应用表现的前提下降低开销，并指出性能与成本并非必然取舍。
@RohOnChain · 68.8K 粉丝 · 102.3K 阅 · 524 赞 · 48 转
I will break down exactly how to build an AI agent like hedge funds that finds profitable trading strategies from the markets 24/7 with GPT-6 Astra. Let's get straight to it. Bookmark This - I'm
中文介绍 拆解如何用 GPT-6 Astra 搭一个类似对冲基金的 AI agent，全天候从市场里挖掘可盈利的交易策略，作者建议读者收藏。属交易自动化选题，方法细节需看全文。
@paulfchristiano · 5.3K 粉丝 · 70.7K 阅 · 654 赞 · 96 转
I am excited to be joining the OpenAI nonprofit board, serving on the Safety and Security Committee to support safety oversight. Based on the recent trajectory of capabilities and the continued
中文介绍 OpenAI 非营利董事会新成员的个人声明：将加入安全与安全委员会，参与安全监督，并表示这一决定基于近期模型能力的发展轨迹。属治理与人事层面的动态。
@ForwardEditor · 10.0K 粉丝 · 61.4K 阅 · 512 赞 · 36 转
This weekend, I watched a movie with my daughter while my computer worked on a video I would have normally spent most of a day editing. The video was a favor for someone in my personal life. I'm a
中文介绍 视频剪辑师展示用 Astra 自动剪片的实操：周末陪女儿看电影时电脑同步完成了平时要花大半天的剪辑工作，素材是为熟人帮忙的私人片子。属真实场景用例。
@JoshARosen · 1.3K 粉丝 · 47.0K 阅 · 510 赞 · 49 转
LLM-as-judge is already a common part of building AI applications. Run your application against a dataset, have another model grade the results, and use the scores to figure out whether the latest
中文介绍 讲把 LLM-as-judge 从离线评测搬进 agent 运行时：不再只是跑一遍数据集让另一个模型打分，而是让评判逻辑在 agent 运行过程中持续生效。属架构方法论分享。
@TechWithTimm · 16.2K 粉丝 · 40.2K 阅 · 506 赞 · 79 转
A lot of software engineers see AI engineering as a completely different career. But it’s not. I have a friend who works as an AI engineer and makes around $200,000 to $300,000 a year. When I asked
中文介绍 讨论软件工程师转 AI 工程师这条路径，作者举朋友年入 20 万到 30 万美元为例，认为两者并非完全不同的职业，转型没有想象中难。属职业规划类内容。
@GoogleAI · 2.5M 粉丝 · 37.4K 阅 · 643 赞 · 90 转
Today, we’re introducing AlphaGenome Atlas 🧬 an interactive platform for genomics discovery. In 2025, @GoogleDeepMind released AlphaGenome to help decode the complex language of our DNA — but
中文介绍 Google 发布 AlphaGenome Atlas，一个用于基因组学发现的交互式平台，建立在 DeepMind 2025 年推出的 AlphaGenome 之上，提供人类 DNA 的高分辨率图谱。属官方产品发布。
@DataRepublican · 957.2K 粉丝 · 37.3K 阅 · 846 赞 · 155 转
NYC and LA are restricting student AI. We used it to expose what school missed—and repair two years of gaps in six weeks. Part 1 of 3 in a series. Between my husband and I, three of our parents were
中文介绍 系列第一篇：作者称在纽约、洛杉矶限制学生用 AI 的背景下，用 AI 帮儿子在六周内补上两年 Saxon 数学的缺口，并借此指出学校教学遗漏的地方。属家庭教育实验记录。
@poteto · 92.6K 粉丝 · 23.3K 阅 · 822 赞 · 60 转
In my last post, I showed you why verification is the foundation of everything I do with agents, and how to get started creating your own verification skills. The key lesson was that if an agent can't
中文介绍 pstack 系列第二篇，承接上篇结论「验证是 agent 一切工作的基础」，继续讲如何给自己搭建 verification skills，前提是 agent 无法自行确认结果是否正确。属 agent 工程进阶教程。
@jhMITgunship · 1.5K 粉丝 · 18.3K 阅 · 1.4K 赞 · 1 转
What superintelligence needs to operate in the physical world For much of the work we give models from OpenAI, Anthropic and the other frontier labs, the world has already been turned into context.
中文介绍 从物理世界切入讨论超智能的能力边界：前沿实验室的模型处理的多数是已经被转成 context 的世界，而真实物理世界不会直接递给你一个状态向量。属偏理论视角。
@Alaya_AI · 321.1K 粉丝 · 10.6K 阅 · 504 赞 · 450 转
As autonomous vehicles, robotics, and spatial computing platforms evolve, AI model training is shifting away from isolated 2D image annotation toward complex multimodal sensor fusion. Today’s frontier
中文介绍 面向自动驾驶、机器人与空间计算，讲多模态传感器同步：模型训练正从孤立的 2D 图像标注转向复杂的传感器融合，对时序对齐的要求也随之提高。属厂商技术科普。
César de la Fuente’s lab uses Codex and ChatGPT to search living and extinct genomes for antimicrobial candidates to fight drug-resistant infections.
中文介绍 研究员 César de la Fuente 的实验室利用 Codex 与 ChatGPT，在现存及已灭绝物种的基因组中搜索抗菌分子候选，以对抗耐药性感染。
Meet the Data agent in ChatGPT Work. Connect company data, uncover insights, and build interactive dashboards with AI using natural language.
中文介绍 ChatGPT Work 推出数据智能体（Data agent），可连接企业数据、以自然语言获取洞察，并构建交互式仪表盘。
On July 22, 2026, a transmission line fault in Ashburn, Virginia—the heart of the world’s largest data center cluster—knocked more than 3 gigawatts of load off the grid in seconds. And it wasn’t the first time. Two years earlier, a single failed surge arrester dropped roughly 60 Virginia facilities
中文介绍 2026 年 7 月 22 日，弗吉尼亚州阿什本——全球最大数据中心集群所在地——一条输电线路故障在数秒内使电网减少逾 3 吉瓦负荷，此前两年也曾发生类似事故。文章认为为 AI 供电本质上是架构问题。
Introducing ChatGPT for Financial Services, combining built-in financial data and GPT-6 Astra for research, modeling, and client-ready materials.
中文介绍 OpenAI 推出面向金融服务的 ChatGPT，整合内置金融数据与 GPT-6 Astra，用于研究、建模以及制作可直接交付客户的材料。
OpenAI and GSA will offer eligible federal, state, local, and tribal governments $0 license fees, 50% off usage, and expanded cyber defense support.
中文介绍 OpenAI 与美国总务管理局（GSA）合作，向符合条件的联邦、州、地方及部落政府提供零许可费、使用量五折优惠，并扩大网络防御支持。
a quiet day
中文介绍 该期 AI 简报为安静的一天，未报道重大行业动态。
Build and launch cloud agents with the Agents API, a managed service powered by the Codex harness for orchestration, long-running sessions, and tool use.
中文介绍 OpenAI 发布 Agents API，这一托管服务由 Codex harness 驱动，支持任务编排、长时间运行的会话与工具调用，用于构建并上线云端智能体。
GPT‑Live‑1 brings natural, full-duplex voice conversations to the API, with stronger instruction following, custom voices, and telephony support.
中文介绍 GPT-Live-1 通过 API 提供自然的全双工语音对话，具备更强的指令遵循能力、自定义音色与电话支持。
中文介绍 Hugging Face 博客发布文章，介绍以 Gradio Workflow 重建 AUTOMATIC1111 的方案。
中文介绍 TLDR AI 当日简报涵盖 DeepSeek Flash v4.1、Siri AI，以及 Anthropic 有关 AI 经济的观点等话题。
Paul Christiano joins the OpenAI Foundation Board and its Safety and Security Committee, bringing experience in AI alignment, safety, and standards.
中文介绍 Paul Christiano 加入 OpenAI Foundation 董事会及其安全与安保委员会，其在 AI 对齐、安全与标准方面具备经验。
中文介绍 IBM 发布 Granite Time Series PatchTST-FM-r2 模型，称其为当前最优（SOTA）时间序列模型，并采用对商业友好的许可协议。
Chris Lehane argues that stronger AI capabilities require stronger safety evidence, shared standards, and durable policy action while the policy window remains open.
中文介绍 OpenAI 的 Chris Lehane 主张，AI 能力越强越需要更强的安全证据、共同标准与持久的政策行动，而当前政策窗口仍然敞开。
Meet GPT-6 Astra, OpenAI’s most capable model for business, with advanced reasoning, computer use, and stronger writing and design judgment.
中文介绍 OpenAI 发布 GPT-6 Astra，称其为面向企业的最强模型，具备先进推理、计算机操作能力，以及更强的写作与设计判断力。
Overshadowing Cognition's $48B Series E, Mistral's $24B Series D, Meta's Muse agent, and GPT Image 2.5. The most jam packed, feel the AGI day in the history of AI.
中文介绍 OpenAI 称借助 Astra-next、约 1 万个智能体与 1300 亿 token，在 88 小时内报告了纳维-斯托克斯方程奇点发现，成本逾 4000 万美元，或成为第二个千禧年大奖。
美股方面，SPY 报 757.83、单日 -0.6%，位于 SMA20 767.56 与 SMA50 758.25 下方，RSI14 44.1，MACD 1.2257 已落于信号线 2.8366 之下；QQQ 单日 -1.06%、RSI 46.3，两者均标注中性趋势。个股分化明显：AAPL 单日 +3.56%、MACD 2.36 领先信号线 1.5711，MSFT 触发 SMA50 上穿 SMA200 金叉，而 NVDA 出现 MACD 死叉、单日 -2.37%。加密方面，恐慌贪婪指数读 69（贪婪），但总市值 2.63T、24 小时 -4.03%，BTC 主导率 58.5%、ETH 11.3%；BTC 报 76699.87、五日 -3.91%，MACD 2278.31 低于信号 2947.57，ETH 与 SOL 维持多头排列但 RSI 仅 58.5 与 54。中概全线偏弱：BABA RSI 34.9、PDD 30.3、JD 31.6，均为空头排列或中性，0700.HK RSI 37.6。商品外汇中，WTI 原油单日 +8.05%、RSI 79 超买且多头排列，黄金 4359.1 呈中性、MACD 31.41 低于信号 59.63，美元指数 99.09 中性，美元兑人民币 6.7 触 RSI 20.8 超卖并贴近 52 周低。宏观端 10Y 美债收益率 4.94%、RSI 72 超买且接近 52 周高，VIX 单日 +8.38% 至 17.84。整体分布为「股指中性、加密高位回落、中概弱势、能源与利率偏强」。
现价 103.78，单日 +8.05%、五日 +14.03%，RSI14 读 79 已入超买区，MACD 3.8938 高于信号线 2.3575，红柱方向向上；价格同时站上 SMA20 87.65、SMA50 82.62 与 SMA200 79.38，趋势标记 bullish 并给出「多头排列」，距 52 周高仅 -13.14%，动量与均线结构同向偏强。
现价 326.57，单日 +3.56%，RSI14 58.9 处中性偏上，MACD 2.36 明显高于信号线 1.5711；价格站上 SMA20 315.28 与 SMA50 317.22，SMA200 284.65 远在下方，趋势标记 bullish 并触发「多头排列」，距 52 周高 -5.22%，短周期动量相对占优。
现价 77.84，单日 -0.98%、五日 -5.36%，RSI14 30.3 逼近超卖阈值，MACD -2.076 低于信号线 -1.1204，绿柱延续；价格同时低于 SMA20 84.63、SMA50 85.74 与 SMA200 97.48，趋势标记 bearish 并列出「空头排列」，距 52 周高 -44.16%，下行结构完整。
现价 108.56，单日 -0.77%、五日 -2.91%，RSI14 34.9 位于弱势区间；MACD -2.8635 低于信号线 -1.5083，绿柱未收敛；价格低于 SMA20 118.24、SMA50 117.05 与 SMA200 134.75，趋势标记 bearish 并触发「空头排列」，距 52 周高 -43.65%，各周期均线均构成上方压制。
现价 6.7，单日 -0.13%、五日 -0.25%，RSI14 20.8 处超卖区，价格近乎贴平 52 周低（距低点 0%）；MACD -0.011 与信号线 -0.011 齐平并列出「MACD 金叉」，但 SMA20 6.72、SMA50 6.75、SMA200 6.85 仍呈空头排列，超卖与金叉信号对下行趋势形成对冲，方向读数不一致。
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
本报告仅为技术指标读数与走势形态的客观复述，不构成任何投资建议。过去走势不代表未来表现，指标信号可能随时反转或失效，请结合自身风险承受能力独立判断。仅供技术指标解读参考。
Follow the day’s news live Bowen says yesterday a ‘good day’ amid contentious deal that will see koala park plan move ahead The climate change and energy minister, Chris Bowen, said yesterday was a “good day” after the Greens reached a deal with Labor that will allow big industry to use a new koala
中文摘要 澳大利亚议员安德鲁·哈斯蒂表示接受巴纳比·乔伊斯对「叛徒」视频的解释，并称保琳·汉森是「唯一撒谎的懦夫」。气候与能源部长鲍文称，绿党与工党达成协议令昨日成为「好日子」，考拉公园计划将推进。
Israeli attacks come a week after it said to have established operational control over the strategic ridge.
中文摘要 以色列军方称已摧毁位于黎巴嫩阿里·塔希尔岭的真主党基地。以方表示，此次袭击发生在其宣称对该战略要地建立作战控制一周之后。
Exclusive: Analysis shows cost of chicken fillets, potatoes and donuts rose during supermarket giant’s three-month promotion Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Woolworths raised prices for popular items – including a 33%
中文摘要 分析显示，澳大利亚超市巨头Woolworths在为期三个月的「价格冻结」促销期间，上调了至少四种家庭必需品的价格，包括鸡柳、土豆和甜甜圈。
Out searching with a metal detector in his hometown, Kalle Lappinen found a historical treasure buried in an unexpected location.
中文摘要 业余考古爱好者卡勒·拉皮宁在家乡用金属探测器发现芬兰迄今规模最大的维京时代硬币窖藏，出土位置出人意料。
After lower house approval, the bill to grant tens of thousands of Sahrawis citizenship now goes to the Spanish Senate.
中文摘要 西班牙众议院通过法案，拟授予1977年前出生的西撒哈拉人西班牙公民身份，涉及数万名撒哈拉人，法案现已提交参议院审议。
Recent accusations against the electronics company have raised new questions about the nonconsensual data it harvests.
中文摘要 LG就智能电视功能遭音频监控指控作出辩护。相关指控再次引发外界对该电子企业未经同意收集数据的质疑。
Fujimori administration is latest right-wing government to join alliance for a more aggressive approach against crime.
中文摘要 秘鲁加入美国主导的「美洲之盾」安全联盟。藤森政府成为最新加入该联盟的右翼政府，主张对犯罪采取更强硬的手段。
US Republicans return to the stage to close their midterm convention, still betting on Trump's political magnetism.
中文摘要 美国共和党中期选举全国代表大会进入第二天，特朗普与万斯在达拉斯发表讲话，共和党仍押注特朗普的政治号召力。
Israel’s military says it has destroyed a network of Hezbollah tunnels on the strategic Ali al-Taher ridge in Lebanon.
中文摘要 以色列军方称，已摧毁黎巴嫩南部战略要地阿里·塔希尔岭一带的真主党隧道网络。
Fuel and engine oil shortages are forcing Gaza’s hospitals to ration electricity
中文摘要 燃料和机油短缺加剧，加沙地带的医院被迫限制用电，医疗机构的电力供应进一步紧张。
Family members and royal leaders are jockeying for control in the Tooro Kingdom after its former boy king died last month. Elders chose a TV journalist to succeed him, but the royal family has objected.
中文摘要 乌干达托罗王国前少年国王上月去世后爆发继承权之争。王室长老选定一名电视记者继位，但王室家族对此表示反对。
The Trump administration is taking an increasingly militarized approach to Latin America, and it has pushed governments across the region to work with U.S. forces.
中文摘要 特朗普政府表示将与秘鲁合作打击犯罪，并对拉丁美洲采取日益军事化的方针，推动地区多国政府与美国军队开展合作。
A Russian strike on a shopping centre in the central Ukrainian city of Pavlohrad killed at least four people.
中文摘要 俄罗斯对乌克兰中部城市巴甫洛赫拉德一处购物中心发动袭击，造成至少4人死亡。
Thousands of drones were used to recreate New York’s Twin Towers in a tribute ahead of the 25th anniversary of 9/11.
中文摘要 在9·11事件25周年前夕，数千架无人机在纽约夜空重现世贸中心双子塔，以示纪念。
A journalist, he was the first Japanese person to enter space, joining a Soviet crew as part of a spectacular $35 million ratings stunt in 1990.
中文摘要 日本首位进入太空的记者秋山丰宽去世，享年84岁。1990年他随苏联飞船升空，该次飞行耗资3500万美元，被视为电视收视策划。
Europe has some promising battery tech, but can it make up ground lost to Chinese battery giants?
中文摘要 欧洲拥有一些颇具前景的电池技术，但在中国电池巨头已占据优势的背景下，欧洲电池产业能否重新夺回失去的市场份额，成为业界关注焦点。
Australian bonds slid, tracking an overnight slump in Treasuries after escalating Middle East tensions drove up oil prices.
中文摘要 澳大利亚国债下跌，追随隔夜美债跌势。中东局势升级推高油价，澳洲债券收益率升至2011年以来最高水平。
Molly and Taylor Haylett explain the changes to how they managed their finances when they started a family.
中文摘要 莫莉与泰勒·海利特夫妇讲述在组建家庭后，如何调整各自的财务安排，包括由丈夫为妻子的养老金缴款。
As countries grapple with energy costs pushing up inflation, this month will see how central banks respond.
中文摘要 随着各国应对能源成本推高通胀的压力，本月市场的焦点在于各央行将如何回应，利率是否将再度上行引发关注。
The UK needs as much as £70 billion ($94.5 billion) of investment in its power grid by early next decade, with delays to upgrading aging infrastructure threatening to add billions of pounds to customer bills.
中文摘要 英国需在2030年代初前为其电网投入至多700亿英镑（约945亿美元）。老化基础设施升级延迟，可能令用户账单增加数十亿英镑。
Faster sales growth suggests progress in its risky push to compete in the AI infrastructure race
中文摘要 甲骨文数据中心业务收入大幅增长，销售增速加快，显示其在人工智能基础设施领域这一高风险扩张策略取得进展。
Start-up discloses five examples of when actors ‘circumvented controls’ and made efforts to ‘obfuscate’ the purpose of their research
中文摘要 Anthropic披露五起案例，称有行为者「绕过管控」并试图「混淆」研究目的。该公司表示已阻止将人工智能用于潜在生物武器的尝试。
Asian stocks and bonds were set for declines after a surge in oil prices sparked a selloff in US markets, while the latest inflation data reinforced bets on an imminent Federal Reserve rate hike.
中文摘要 油价飙升引发美国市场抛售，亚洲股市与债市料将下跌；最新通胀数据强化了市场对美联储即将加息的预期。
Treasury Secretary Scott Bessent dismissed any concerns with regard to Thursday’s smaller-than-expected debt buyback operation, and played down concerns about a jump in yields.
中文摘要 美国财政部长斯科特·贝森特对周四规模小于预期的国债回购操作不以为意，并淡化收益率上升的担忧，称美国国债市场表现强劲。
Oil extended its powerful rally above $100 a barrel following an escalation of hostilities in the Middle East that has stoked fears of deeper supply disruptions.
中文摘要 中东敌对行动升级加剧了市场对供应中断的担忧，油价延续强劲涨势，突破每桶100美元关口。
If you walk into a local pharmacy in New York City, more likely than not, lots of the merchandise is locked behind closed doors as a measure to mitigate shoplifting. So how big is the "retail shrink" problem really? And why are stores willing to sacrifice customer experience to protect their invento
中文摘要 在纽约市的药店，大量商品被锁在柜中以防盗窃。报道探讨「零售损耗」问题究竟有多大，以及商家为何愿意牺牲顾客体验来应对。
Fanatics’ Betting and Gaming CEO Matt King discusses the company’s newly launched all-in-one app, which combines sports betting, casino, and prediction markets. King discusses Fanatics’ position in a market long dominated by DraftKings and FanDuel, and asked King how much share Fanatics has gained s
中文摘要 Fanatics博彩与游戏业务CEO马特·金讨论公司新推出的一体化应用，该应用整合体育博彩、赌场与预测市场，并谈及公司在DraftKings与FanDuel长期主导市场中的定位。
11 回复 · 程序员 节点
19 回复 · 程序员 节点
5 回复 · 程序员 节点
9 回复 · Apple 节点
10 回复 · 程序员 节点
11 回复 · Apple 节点
8 回复 · 程序员 节点
40 回复 · Apple 节点
8 回复 · Apple 节点
6 回复 · Linux 节点
该源今日无内容。