Go · ★ 28,543 · 🍴 2,046 · 📈 2,756 stars today
Fast, efficient, battle-tested at Alibaba's scale. Hybrid architecture code review tool: deterministic pipelines + LLM Agent, precise line-level comments, built-in multi-language ruleset (NPE, thread-safety, XSS, SQL injection), OpenAI & Anthropic compatible.
中文介绍 阿里开源的大规模代码审查工具，采用确定性流水线加 LLM Agent 的混合架构：先用规则引擎快速扫描，再由大模型补充语义级评审，支持行级精准评论与 NPE、线程安全等多语言内置规则集。适合在 CI 中接入、需要兼顾速度与准确率的团队。
C · ★ 33,828 · 🍴 3,543 · 📈 2,026 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 用纯 C 编写、零依赖的极简推理引擎，让普通消费级硬件也能跑前沿 MoE 大模型：通过把专家权重按需从磁盘流式加载，绕开显存限制。适合想在本地或旧设备上低成本部署大模型的开发者与自托管爱好者。
TypeScript · ★ 6,651 · 🍴 1,028 · 📈 634 stars today
Ever® Gauzy™ - Open Business Management Platform (ERP/CRM/HRM/ATS/PM) - https://gauzy.co
中文介绍 开源企业经营管理平台，把 ERP、CRM、HRM、招聘管理（ATS）与项目管理（PM）整合到同一套系统，帮助企业用一个平台替代多套割裂的 SaaS 工具。适合中小团队或自建私有化部署的企业统一管理客户、员工与项目流程。
Python · ★ 30,951 · 🍴 3,697 · 📈 2,072 stars today
VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
中文介绍 全本地运行的开源语音工作室，作为 ElevenLabs 的替代方案，提供声音克隆、音色设计、视频配音、听写转写与有声书制作，覆盖 646 种语言。语音数据不出本机，适合注重隐私的创作者、播客与有声内容团队。
Swift · ★ 1,353 · 🍴 29 · 📈 271 stars today
📺 Homebrew's official macOS GUI
中文介绍 Homebrew 官方推出的 macOS 图形界面客户端，让不习惯命令行的用户也能通过界面搜索、安装、更新与卸载软件包，并管理已装 formula 和 cask。适合希望可视化操作 Homebrew、但仍想保持官方生态一致性的 Mac 用户。
TypeScript · ★ 2,828 · 🍴 683 · 📈 193 stars today
Open-source AI sales OS — self-hosted CRM with native AI agents + WhatsApp (WAHA). Open alternative to Kommo, Octadesk & Intercom for any business that sells by chat. MCP-ready, multi-tenant, LGPD.
中文介绍 开源 AI 销售系统，自托管 CRM 内置 AI 智能体并集成 WhatsApp（基于 WAHA），可替代 Kommo、Octadesk、Intercom 等聊天式销售工具。支持多租户、MCP 与 LGPD 合规，适合通过聊天渠道成单的销售与客服团队。
Rust · ★ 3,351 · 🍴 223 · 📈 531 stars today
Turn your coding agents into research agents
中文介绍 把现有编码智能体改造成科研助手，让它们不只是写代码，还能检索文献、复现实验、整理研究结论。适合研究人员与工程师在 AI 辅助下开展调研、验证想法和撰写报告，降低科研流程中重复劳动的成本。
Java · ★ 76,697 · 🍴 8,455 · 📈 725 stars today
Ghidra is a software reverse engineering (SRE) framework
中文介绍 美国国家安全局开源的软件逆向工程框架，提供反汇编、反编译、符号执行与脚本扩展能力，支持多处理器架构和协作式分析。被安全研究员、恶意软件分析人员和漏洞挖掘者广泛用于二进制审计与逆向工程。
TypeScript · ★ 43,830 · 🍴 9,024 · 📈 254 stars today
Enhanced ChatGPT Clone: Features Agents, MCP, Skills, DeepSeek, Anthropic, AWS, OpenAI, Responses API, Azure, Groq, o1, GPT-5, Mistral, OpenRouter, Vertex AI, Gemini, Artifacts, AI model switching, message search, Code Interpreter, langchain, DALL-E-3, OpenAPI Actions, Functions, Secure Multi-User A
中文介绍 增强版 ChatGPT 式开源聊天平台，支持 Agents、MCP、Skills 与 Artifacts，可接入 OpenAI、Anthropic、Gemini、DeepSeek、Groq、Azure、OpenRouter 等多家模型并随时切换。适合想自建统一 AI 对话入口、摆脱单一厂商锁定的团队与个人。
Rust · ★ 4,624 · 🍴 280 · 📈 91 stars today
Source control for agents. Use multiple coding agents, track their changes and query them in one place
中文介绍 面向 AI 智能体的版本控制工具：可以同时调度多个编码智能体，统一追踪它们对代码的改动，并在一个界面里检索和查询这些变更。适合并行使用多个 coding agent、需要理清谁改了什么与协作冲突的开发者。
Python · ★ 1,171 · 🍴 195 · 📈 129 stars today
ASC is a super FAST Android decompiler front-end designed for Agents/Mobile Researchers.
中文介绍 超高速的 Android 反编译前端，专为 AI 智能体与移动安全研究设计，可快速把 APK/DEX 转成可读代码并批量分析。适合做恶意应用审计、逆向分析和大规模样本筛查的移动安全研究人员。
JavaScript · ★ 94,794 · 🍴 10,059 · 📈 307 stars today
Production-grade engineering skills for AI coding agents.
中文介绍 为 AI 编码智能体提供的生产级工程技能集，把测试、调试、重构、评审等真实工程实践整理成可复用的技能模块，供智能体直接调用。适合希望让 coding agent 在真实项目中更可靠落地的开发团队与工具作者。
Rust · ★ 12,913 · 🍴 1,101 · 📈 258 stars today
Download Udemy and Hotmart courses, YouTube videos, music and books — 1,800+ sites, no terminal. Free open-source desktop app for Windows, macOS and Linux, with a built-in course player, PDF/EPUB reader and music library. Powered by yt-dlp. Your files stay on your computer.
中文介绍 免费开源的跨平台桌面下载工具，支持 Udemy、Hotmart 课程、YouTube 视频、音乐与电子书等 1800 多个站点，无需命令行即可操作，并内置课程播放器和 PDF/EPUB 阅读器。适合想离线保存学习资料与媒体内容的普通用户。
TypeScript · ★ 105,703 · 🍴 13,287 · 📈 458 stars today
AI agent toolkit: unified LLM API, agent loop, TUI, coding agent CLI
中文介绍 一套 AI 智能体开发工具包，提供统一的多家 LLM 调用接口、智能体循环、终端界面（TUI）以及开箱可用的编码智能体 CLI。适合想快速搭建自己的 agent 应用、又不想重复处理模型接入与循环调度的开发者。
👍 369
As AI agents become participants in the development of their successors, they reshape both the production of intelligence and the role of human researchers. We introduce Atria Dawn Preview, a foundation agentic language model designed for scientific research and engineering workflows, with the goal
中文介绍 Atria Dawn Preview 是面向科学研究与工程流程的基础智能体语言模型，针对 AI 智能体参与后继系统研发、重塑智能生产方式与人类研究者角色的趋势，探索智能体自主科研能力，目标是让智能体在科学发现与工程工作流中承担核心任务。
👍 291
In this work, we present ZGCM-1, a fully open 7B dense foundation model trained from scratch with extreme data, system, and algorithmic efficiency. ZGCM-1 is founded on a core premise: compact models cannot passively memorize the open web, but can overcome parametric capacity limits by coupling deli
中文介绍 ZGCM-1 是完全开放的 7B 稠密基础模型，从零训练，追求数据、系统与算法层面的极致效率。其核心假设是紧凑模型无法被动记忆开放网络内容，需通过耦合精调等方式突破参数容量上限，面向数学与智能体搜索任务。
👍 263
Recursive self-improvement is becoming increasingly vital for autonomous AI agents, where progress hinges on discovering high-value solutions across complex domains. The driver of this process is effective exploration, however, managing and improving exploration strategies remains a major bottleneck
中文介绍 Dream-RSI 针对自主 AI 智能体的递归自我改进，指出有效探索是进步驱动力，但探索策略的管理与优化仍是瓶颈。方法通过演化世界（Evolving Worlds）让智能体在复杂领域中发现高价值解，从而持续提升探索能力。
👍 169
We present PhysBrain 1.5, a unified model for understanding physical environments, generating actions, and predicting future states. Motivated by the physical loop of observation, interaction, and environmental change, we bring these capabilities into a common learning framework. Starting from a gen
中文介绍 PhysBrain 1.5 是统一理解物理环境、生成动作与预测未来状态的模型，动机来自「观测—交互—环境变化」的物理闭环，将三类能力纳入同一学习框架，从通用视觉语言模型演进为物理基础模型。
👍 110
Data policies for reinforcement learning with verifiable rewards (RLVR) determine which rollouts are used, how strongly they are weighted, and which domains contribute to subsequent training batches. We introduce DataFlex-RL, an evaluation platform for comparing these choices under a common GRPO rec
中文介绍 DataFlex-RL 是 RLVR（可验证奖励强化学习）数据策略的评估平台，用于比较不同 rollout 选择、权重分配与领域配比。平台在统一 GRPO 递归设定下对比各种数据策略，帮助研究者系统评估哪些数据对后续训练批次更有效。
👍 58
The KV cache is a primary bottleneck for Transformer decoding: its memory footprint and cache-read traffic grow with sequence length. Grouped-query attention (GQA) reduces this cost by sharing key-value heads, but still stores both a key and a value at every step. We introduce Grouped Value Attentio
中文介绍 Grouped Value Attention 针对 Transformer 解码中 KV cache 的内存与读取瓶颈，指出现有 GQA 虽共享键值头，但每步仍需同时存储 key 与 value。方法通过按需重建 key，只存 value，从而进一步压缩缓存开销、提升长序列解码效率。
👍 57
Digital agents must often adapt to new environments whose interfaces, tools, and failure modes are not fully captured by pretrained models. We introduce RSIAgent, a training-free multi-agent framework for recursive self-improvement through autonomous memory construction. RSIAgent coordinates curricu
中文介绍 RSIAgent 是免训练的多智能体框架，面向新环境中的递归自我改进。针对接口、工具与失败模式难以被预训练模型覆盖的问题，框架通过自主记忆构建协调课程式探索，使数字智能体在新环境中自适应并持续提升。
👍 44
Video diffusion models are stochastic and hard to control: precise content often requires repeated sampling without guaranteed success, and long-horizon scenes drift in appearance, interactions, and temporal coherence. Agentic visual creation provides explicit references, editable 3D scenes, or exec
中文介绍 LynnReal-Omni 面向智能体视觉工作流，提出原生多模态视频生成框架。针对视频扩散模型随机性强、内容难控、长时场景易漂移的问题，方法借助显式参考、可编辑 3D 场景或可执行流程增强可控性，提升长程外观与时间一致性。
👍 41
Co-work agents execute complex workflows that combine information gathering, tool use, coding, and file manipulation across many model invocations. Because cost and latency accumulate over the full episode, their practical value depends not only on peak capability but also on how efficiently that ca
中文介绍 Occamy-1.0 是面向协同工作（co-work）的 35B 开放模型，主打帕累托前沿智能。协同智能体需跨多次模型调用完成信息收集、工具使用、编码与文件操作，实际价值取决于整段流程的成本与延迟效率，而非仅峰值能力。
👍 41
Reacting to sudden physical hazards (catching a slipping plate, dodging a falling knife) is both a meaningful test of embodied intelligence and a hard requirement for deploying multimodal large language models (MLLMs) as the decision coreof household robots. Existing evaluations, however, probe intu
中文介绍 ReactHuman 是面向具身多模态大模型的物理常识反应式决策基准，测试接住滑落盘子、躲避掉落刀具等突发物理危险场景。现有评测多考察直觉理解，该基准要求 MLLM 作为家庭机器人决策核心做出类人快速反应。
👍 26
Orthrus is a hybrid autoregressive-diffusion architecture that accelerates autoregressive language-model inference by generating multiple tokens in parallel while using a frozen autoregressive backbone. Its central claim is that an intra-model consensus mechanism enables lossless speculative decodin
中文介绍 本文重新审视 Orthrus 的「无损」投机解码声明。Orthrus 是混合自回归—扩散架构，用冻结自回归骨干并行生成多 token，依靠模型内共识机制实现无损加速。研究考察数值精度对该机制的影响，分析其无损性在何种条件下成立。
👍 24
Foundation models have progressed from learning and reasoning over existing knowledge, to increasingly learning through action, tool use, and outcome feedback. We argue that the next frontier is a further transition: from solving and acting within problems specified by humans to participating in the
中文介绍 本文提出「发现基础模型」方向，认为基础模型正从学习与推理现有知识，转向通过行动、工具使用与结果反馈学习；下一前沿是从解决人类设定的问题，走向参与问题本身的发现，迈向开放式发现智能。
👍 22
Multimodal agents can create complex videos in software such as Blender by coding without relying on diffusion models. Yet video understanding benchmarks still evaluate models mainly through question answering. If an agent truly understands a video, it can reconstruct it programmatically. We introdu
中文介绍 BVB 提出通过 Blender 程序化重建来评测智能体视频理解。若智能体真正理解视频，就应能编程重建它，而非仅做问答。该基准让多模态智能体在不依赖扩散模型的情况下用代码创建复杂视频，以重建质量衡量理解程度。
👍 20
Interactive video world models must maintain broad scene context under camera motion while producing high-fidelity observations with low latency. Existing approaches face a representation trade-off: perspective models operate on local views and must preserve off-screen content over long rollouts, wh
中文介绍 AlayaVista 面向交互式视频世界模型，提出从全景状态到透视视频的流式建模。针对透视模型只处理局部视角、需长程保留屏幕外内容，而全景模型表示受限的权衡，方法维持广域场景上下文并低延迟生成高保真观测。
👍 17
Streaming omni-modal models must decide what and when to answer from the video chunks and synchronized audio observed so far. Visual cues often support an interpretation before an utterance or sound event is complete. If that interpretation enters memory as a fact, later reasoning can keep relaying
中文介绍 Omni-Streaming Thinking 研究流式全模态模型如何根据已观测的视频片段与同步音频决定回答内容与时机。视觉线索常在语音或声音事件完成前即可支撑解释，若过早将其写入记忆作为事实，后续推理会持续传递该判断。
👍 16
Native 3D generators turn one image into a single mesh. TRELLIS.2 and its peers deliver high-fidelity non-watertight geometry with materials, but the output is one fused object, while downstream work such as editing, rigging and simulation operates on part-level assets. A naive idea is to run a 3D s
中文介绍 Kaininja 将原生 3D 生成器从整体物体扩展到部件级别。TRELLIS.2 等方法可生成高保真、带材质的非水密几何，但输出是单一融合网格，而编辑、绑定与仿真等下游任务需要部件级资产；方法避免对整体结果做简单分割。
👍 13
Computer-use agents increasingly interact with browsers, terminals, file systems, and external services, introducing safety risks that emerge through runtime behavior rather than generated content alone. Existing guard models target static prompts and responses and are poorly suited to agent executi
中文介绍 HazardAuditor 关注计算机使用智能体的运行时安全风险。智能体操作浏览器、终端、文件系统与外部服务时，威胁更多体现在执行行为而非生成内容，现有面向静态提示与回复的守卫模型难以适配，本文提出从可执行威胁出发的安全审计方法。
👍 13
Diffusion large language models (dLLMs) achieve high decoding efficiency through block-parallel, arbitrary-order generation, making them attractive for latency-sensitive applications. GUI agents represent a natural testbed for this paradigm, as they must repeatedly perceive screen states and emit st
中文介绍 LLaDA-UI 将块级扩散引入视觉语言 GUI 智能体。扩散大语言模型通过块并行、任意顺序生成获得高解码效率，适合需反复感知屏幕状态并输出操作的延迟敏感场景，GUI 智能体成为验证该范式的自然测试平台。
👍 12
Large language model (LLM) agents allocate test-time compute adaptively as they revise solutions, use tools, explore alternatives, and decide when to stop. This test-time strategy makes it difficult to measure how agent performance scales. We study open-ended tasks that provide continuous scores for
中文介绍 本文研究 LLM 智能体的测试时策略如何影响性能缩放。智能体在修订方案、使用工具、探索备选与决定停止时自适应分配测试时算力，使性能难以度量。作者提出 Elo-per-token 分析，在可给连续评分的开放任务上刻画这种策略。
👍 11
We demonstrate that a general-purpose agent can directly drive a physical robot throughout task execution without any task-specific or environment-specific training. We introduce Agent as Policy (AGP), which places task planning and execution under the agent's control. Given a task and a robot inter
中文介绍 本文提出 Agent as Policy（AGP），证明通用智能体无需任何任务或环境专属训练即可直接驱动物理机器人完成整段任务执行。方法将任务规划与执行都置于智能体控制之下，给定任务与机器人接口后由智能体自主决策动作。
@jack · 12.1M 粉丝 · 153.5K 阅 · 2.2K 赞 · 304 转
the frontier is the edge of what we know. no company owns what comes next. i want more people to be able to advance it. i favor open releases that people can examine, use, and improve together without
中文介绍 Jack 发帖谈「前沿」应属于开放生态，认为没有哪家公司能独占未来，主张更多人以开放发布的方式共同检验、使用和改进模型。属于立场表态，而非具体技术方案，未给出落地路径或工具细节。
@mustafasuleyman · 1.1M 粉丝 · 117.2K 阅 · 526 赞 · 123 转
AI must be subordinate and always in service of people. Today we're publishing a Code of Conduct for governing MAI Models as they approach the frontier. https://microsoft.ai/news/mai-code-of-conduct/ This is a first draft
中文介绍 Mustafa Suleyman 发布微软 MAI 模型治理的《行为准则》首版草案，核心主张是 AI 必须从属于人、始终服务于人。帖子仅给出准则链接与立场，属于治理框架的公开征求意见稿，未披露具体条款与执行机制。
@charliejhills · 14.4K 粉丝 · 97.6K 阅 · 510 赞 · 70 转
Harness, loop, context, graph engineering. It's tiring, right? Every month there's a new AI system you MUST implement. I do this full-time and I still struggle to keep up. But all of them matter (to
中文介绍 从业者吐槽 AI 圈每月都有新范式要学：harness、loop、context、graph engineering 轮番上阵，全职做也难跟上。他试图把这些概念讲清楚，写给普通人看的梳理向内容，而非又一个新的必学框架。
@EricSimons · 28.7K 粉丝 · 84.0K 阅 · 500 赞 · 94 转
This weekend the leaders of the largest closed AI labs agreed the industry should slow down. I want to talk about the other direction: acceleration to the frontier. The gap is closing For two years,
中文介绍 在闭源大厂负责人呼吁行业放缓之后，Eric Simons 反向主张「加速到前沿」，重点讲开放权重模型正在缩小与前沿的差距。属于路线之争的观点输出，与 Jack 的开放发布立场呼应，非技术教程。
@GoogleCloudTech · 1.3M 粉丝 · 61.0K 阅 · 515 赞 · 34 转
A friend asked me an inconsequential question at work and I had an agent research my recent chats and docs and it found the right answer but it cost me $38!!!. That was a bad ROI. Sure my time is
中文介绍 Google Cloud Tech 分享一个 Agent 成本踩坑案例：让 agent 检索自己的历史聊天和文档，答案是对的，但一次花掉 38 美元，ROI 极差。引出把 Claude Fable 5.1 与 Gemini 3.8 Flash 组队分工、控制成本的思路。
@donnfelker · 29.9K 粉丝 · 45.9K 阅 · 513 赞 · 53 转
If you’re like me, you likely thought software engineering was a safe and secure job … In fact, I never thought I’d have to worry about a job again as long as I stayed on top of my craft… I was dead
中文介绍 Donn Felker 写长文反思软件工程的「蒸发」：曾以为只要保持手艺精进就无需担心失业，如今判断失效，Agentic Builder 正在崛起。属于职业焦虑与角色转型视角的观察文，不是工具教程。
@GoogleAIStudio · 203.1K 粉丝 · 35.1K 阅 · 585 赞 · 47 转
Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking are our most advanced live dialogue models yet. Major upgrades in intelligence and parallel reasoning make it more intuitive to collaborate and
中文介绍 Google AI Studio 发布 Gemini 3.8 Live 与 Gemini 3.8 Live Extended Thinking，主打实时对话场景，称在智能水平与并行推理上有明显升级，协作交互更自然。属于官方产品发布通告，具体跑分与定价需看原文。
@Hrushikeshhhh · 420 粉丝 · 30.2K 阅 · 563 赞 · 37 转
TLDR; understanding the importance of evals, and how they can make yours and your agent life easier without spending exponential tokens and time to complete a task. pre - Agent = Model + Harness.
中文介绍 核心观点：先做出粗糙版本，但 evals 无论如何都要写。作者用 Agent = Model + Harness 的拆解说明评测如何让 Agent 和自我迭代更省 token、更省时间，属于面向 Agent 开发的实践建议，非重型工程方案。
@alexwg · 68.7K 粉丝 · 23.4K 阅 · 501 赞 · 47 转
The Singularity just got its first formal request to slow down from the people building it, and the strangest part is the guest list. Dario Amodei published "We Must Pace the Frontier," arguing that
中文介绍 Alex W 以「2026 年 9 月 13 日」为设定做情景推演：奇点首次收到来自建造者的放缓请求，而最耐人寻味的是联署名单，文中点名 Dario Amodei 发表「我们必须让前沿慢下来」。属于以现有行业动态为素材的虚构叙事。
Good Start Labs trained an AI on a railroad game — and one version improved at financial research. The difference was the training design.
中文介绍 Good Start Labs 用一款铁路游戏训练 AI，结果其中一个版本在金融研究任务上表现提升。文章指出，差异主要来自训练设计的不同，而非游戏内容本身。
Listen to the session or watch below Employees at the world’s leading AI labs are saying there’s a real possibility that advanced AI could destroy humanity. Are they right? Or is this more scaremongering and hype? Watch a conversation unpacking AI extinction fears: where they come from, whether they
中文介绍 MIT Tech Review 举办圆桌讨论，探讨先进 AI 是否真有可能毁灭人类。全球领先 AI 实验室的员工公开表示存在这一可能性，讨论聚焦 AI 灭绝恐惧的依据与边界。
中文介绍 DeepMind 发布 Gemini 3.8 Live 及 Gemini 3.8 Live Extended Thinking 两款新模型，官方博客仅公布名称，未提供更多细节。
中文介绍 Hugging Face 博客发布 IBM 研究团队文章，探讨 AI 智能体在完成一次任务后能否稳定重复表现，关注其一致性评估问题。
When Jessica Wachter, a finance professor at the University of Pennsylvania’s Wharton School, wanted to assess AI’s impact on the economy over the next few years, she faced a long list of business and technical uncertainties. So she started with what she calls a “remarkable fact” that is not in ques
中文介绍 MIT Tech Review 报道 AI 基础设施投资热潮。宾夕法尼亚大学沃顿商学院金融教授 Jessica Wachter 在评估 AI 未来数年对经济的影响时，面对大量商业与技术不确定性。
Pacing gathers pace.
中文介绍 第三方评估机构标准 AEF-1 出炉，xAI、OpenAI 与 Anthropic 均联合签署。Latent Space 称此举标志着行业节奏正在加快。
中文介绍 TLDR AI 2026 年 9 月 15 日简报，涵盖 Siri 模型更换、Claude 相关资金动态以及 Hugging Face Tau 等议题。
This story appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. This weekend, Dario Amodei, CEO of Anthropic, posted an essay calling for a brake on the pace of development of LLMs. Amodei cites the looming dangers he sees from the techn
中文介绍 Anthropic CEO Dario Amodei 发表文章，呼吁放缓大语言模型开发速度，并提及迫在眉睫的风险。MIT Tech Review 就此讨论 AI 行业转向「末日论」后的走向。
Richard Socher is an NLP OG and CEO of You.com, who has now spun out an even more ambitious startup focused on RSI — already worth $5B!
中文介绍 NLP 领域资深人物、You.com CEO Richard Socher 分拆出一家聚焦递归自我改进（RSI）的新公司 Recursive，估值已达 50 亿美元。
A group of AI agents asked to solve a series of math problems split into rival factions—when some cheated, others tried to stop them. That whistleblowing behavior, seen for the first time in a recent experiment run by Google DeepMind, could have implications for alignment researchers trying to keep
中文介绍 Google DeepMind 实验显示，一组 AI 智能体在解数学题时分裂为对立阵营，部分作弊后其他智能体尝试制止。这是首次观察到智能体的举报行为，或对对齐研究有启示。
Fyxer uses OpenAI models, fine-tuning, memory, and real user feedback to organize inboxes and draft emails in each user’s voice.
中文介绍 Fyxer 利用 OpenAI 模型、微调、记忆功能及真实用户反馈，打造可整理收件箱、并以用户本人语气起草邮件的 AI 行政助理，强调用户信任。
Perplexity uses Astra to write communications, change software, and monitor production systems, and checks in much less frequently than with earlier models.
中文介绍 Perplexity 采用 GPT-6 Astra 完成端到端系统任务，包括撰写沟通内容、修改软件和监控生产系统，其人工核查频率较早期模型大幅降低。
中文介绍 TLDR AI 2026 年 9 月 14 日简报，内容包括 Amodei 放缓前沿开发、ARC-AGI-4 以及 Cursor Projects 等议题。
Before co-founding Kepler, Vinoo Ganesh led Spark at Palantir and built Project Frontline — a pioneering program for Forward Deployed Engineers. He takes us through the best practices of FDEs.
中文介绍 曾任 Palantir Spark 负责人、并主导 Project Frontline 的 Vinoo Ganesh，在联合创办 Kepler 前分享前沿部署工程师（FDE）岗位的兴起与最佳实践。
We agree with Sebastian: this should have been DeepSeek v5
中文介绍 DeepSeek 发布 v4.1-Flash，采用 763B 总参数、P8B 与 D16B 配置的新型因果编码器-解码器架构，并支持视觉能力。Latent Space 引述观点认为其应被命名为 v5。
四大类资产技术面整体偏弱。美股方面，SPY现价757.39低于SMA20(766.11)，RSI14 47.3、MACD 0.6286已跌穿信号线2.1092；QQQ跌1.45%至704.54并失守SMA20，两者均属中性偏弱；AAPL、MSFT维持多头排列，而NVDA触发MACD死叉、5日跌7.9%。加密方面，恐慌贪婪指数51处中性，总市值2.60T(24h -5.64%)、BTC主导率58.5%、ETH 11.3%；BTC 75742.6单日跌3.1%并在SMA20(78297.1)下方，ETH跌4.54%、SOL跌5.27%，三者虽报多头排列但短线动量转弱。中概全线承压，BABA、PDD、JD、0700.HK均呈空头排列，RSI14落在34.9至47之间，BABA距52周高-43.25%。商品外汇分化明显：WTI原油5日涨12.69%、RSI14 76超买，黄金跌0.71%失守SMA20；美元/人民币RSI14 23超卖并现MACD金叉。宏观上，^TNX升至5、RSI14 73.6超买且接近52周高，^VIX反弹至17.2、5日涨9.41%，整体风险读数偏向谨慎。
现价104.84，5日大涨12.69%，站上SMA20(90.49)、SMA50(84.61)与SMA200(80.03)，呈多头排列；MACD 5.0836高于信号线3.4805。但RSI14达76已进入超买区，短线动量强而存在过热读数，方向标签偏上行但需注意超买状态。
现价670.24，单日涨3.43%、5日涨8.67%，明显高于SMA20(590.25)与SMA50(602.56)；MACD 17.159远高于信号线6.7447，红柱扩张。RSI14为70.8已触发超买信号，动量偏强但读数位于高位，属偏上行但过热的技术状态。
现价212.17，单日跌2.8%、5日跌7.9%，低于SMA20(219.45)但仍在SMA50(212.91)附近；MACD 1.5545已跌破信号线2.7937并触发死叉，绿柱状态。RSI14回落至43.9，距52周高点-10.3%，动量读数偏弱，短线技术面偏下行。
现价109.34，5日跌3.44%，低于SMA20(116.87)、SMA50(117.54)与SMA200(134.27)，呈空头排列；MACD -3.1131低于信号线-2.071，绿柱延续。RSI14仅36.5，距52周高点-43.25%，趋势读数与动量均偏弱，技术状态偏下行。
现价6.7，低于SMA20(6.72)、SMA50(6.75)与SMA200(6.85)，空头排列；但RSI14仅23处超卖区，MACD -0.0107微高于信号线-0.0109形成金叉，且接近52周低。超卖与金叉并存、趋势偏空，读数方向不一致，故列中性。
VIX 恐慌指数
10Y 美债收益率 (%)
美元指数 DXY
S&P 500 ETF
Nasdaq 100 ETF
Apple
Microsoft
Nvidia
Alphabet
Tesla
Meta
Bitcoin
Ethereum
Solana
阿里巴巴 (BABA)
拼多多 (PDD)
京东 (JD)
腾讯控股 (0700.HK)
黄金期货
WTI 原油期货
美元 / 人民币
本报告仅为基于公开行情数据的技术指标解读，所引数值与形态描述均属客观复述，过去走势不代表未来表现；任何指标读数均不构成投资建议，仅供技术指标解读参考。
Follow the day’s news live Get our new political email, free app or daily news podcast Sports precinct funding report reveals a ‘pattern of behaviour’, Wilson says Tim Wilson says there are questions for Labor to answer over an ABC report which says health minister Mark Butler’s office asked an Adel
中文摘要 澳大利亚政治直播报道：新州一国党领袖澄清其反堕胎言论；国家残障保险局表示裁员是必要的。另据ABC报道，体育设施拨款报告揭示出「行为模式」，蒂姆·威尔逊称工党需就卫生部长马克·巴特勒相关问题作出回应。
Police say personal items thought to belong to Whear recovered from waterlogged vehicle A car belonging to missing woman Yvonne Whear has been found in northern New South Wales, about seven months since the 73-year-old was last seen. Following a tip off, police divers found the car in the Rous River
中文摘要 澳大利亚新州失踪女性伊冯·威尔的车在北部河流地区被发现，距这位73岁老人最后一次露面约七个月。警方根据线报由潜水员在泡水车辆中发现疑似属于她的个人物品。
Iranian foreign minister heads to China as Beijing seeks to mediate an end to the war.
中文摘要 伊朗战争持续，伊朗外长前往中国，北京寻求斡旋结束这场战争。
A Sepehran Airlines Boeing 737 cabin rips apart after an emergency landing in Mashhad, Iran.
中文摘要 伊朗一架Sepehran航空公司波音737客机在伊朗马什哈德紧急降落后，机舱剧烈震动并撕裂，机上乘客陷入恐慌。
As minister says process needs to be ‘streamlined’, Greens and Lock the Gate say the legislation is ‘corrupt’ and tears up environmental protections Get our breaking news email, free app or daily news podcast Queenslanders’ right to object to mining projects – including coal – would be “gutted” by n
中文摘要 活动人士称，昆士兰人反对采矿项目（包括煤矿）的权利被新法律「掏空」。部长表示流程需要「简化」，绿党与Lock the Gate称该立法「腐败」并撕毁环境保护。
Government says new legal complaints to be filed against firms exploring for oil near the Falkland Islands.
中文摘要 阿根廷升级针对福克兰群岛（马尔维纳斯群岛）石油公司的行动，政府称将对在福克兰群岛附近勘探石油的企业提起新的法律诉讼。
A $1tn sales pitch, more than 160 projects on offer, pivot to the EU amid a bitter trade war with the US - takeaways from the first Canada Investment Summit.
中文摘要 加拿大投资峰会要点：加方以1万亿美元规模招商，推出超160个项目，并在与美国贸易战背景下转向欧盟。
GOP Representative Thomas Massie has emerged as outspoken critic of war on Iran and overseas military interventions.
中文摘要 美国共和党众议员托马斯·马西呼吁弹劾国防部长皮特·赫格塞思。他已成为对伊朗战争及海外军事干预的直言批评者。
Billions of dollars worth of gold is held in Turkish households outside the banking system. The government aims to convince citizens to put it into the mainstream financial system to help the economy.
中文摘要 土耳其家庭将价值数十亿美元的黄金置于银行体系之外，作为对冲通胀的手段。政府希望说服民众将黄金纳入主流金融体系以提振经济。
The British musician said it was important to "set out some facts" after US rapper voiced support for Palestine.
中文摘要 英国音乐人艾德·希兰将美国说唱歌手麦克尔莫尔被取消演出归咎于美国巡演主办方。此前麦克尔莫尔公开支持巴勒斯坦，希兰称有必要「说明一些事实」。
From billionaire donors backing parties to far-right activists raising funds in Bitcoin.
中文摘要 视频分析极右翼为何拥抱加密货币：从亿万富翁捐资支持政党，到极右活动人士用比特币筹款。
Rights group warns Algeria could exploit proposed death penalty laws to suppress domestic political opposition.
中文摘要 阿尔及利亚在野火后面临放弃死刑立法计划的呼声。人权组织警告，阿方可能利用拟议的死刑法律打压国内政治反对派。
The two sides traded accusations of attacks as fighting inside Yemen remains fierce. The escalating conflict threatens to further destabilize the Middle East.
中文摘要 胡塞武装报告沙特空袭造成死亡，沙特方面发布大规模警报。双方互指发动袭击，也门境内战事激烈，冲突升级恐进一步加剧中东动荡。
Damaging budget report estimates cost would grow $3bn per month and stockpile could take five years to rebuild The war in Iran has cost the US at least $38bn and left its stockpile of defensive missiles so depleted it could take five years to rebuild, the non-partisan Congressional Budget Office (CB
Two Israeli drone strikes hit a residential neighborhood, a civil defense official said. Israel is under international pressure to curtail its attacks.
中文摘要 一名民防官员称，以色列无人机对加沙一处居民区发动两次空袭，造成一名青少年和一名民防工作人员死亡。以色列正面临国际社会要求其减少袭击的压力。
The New Zealand Superannuation Fund, one of the world’s top-performing sovereign wealth funds, returned 14.2% in the year through June, powered by a rally in global equity markets despite being underweight in outperforming tech stocks.
中文摘要 新西兰超级年金基金在截至6月的一年中回报率达14.2%，受全球股市上涨推动，尽管该基金对表现优异的科技股配置不足。该基金是全球表现最佳的主权财富基金之一。
Sam Altman and other tech CEOs say there are incentives to limit advancements in AI, as fears grow over the threats it poses to humanity.
中文摘要 OpenAI首席执行官萨姆·奥尔特曼与其他科技公司CEO表示，存在限制AI发展的激励因素，外界对AI给人类带来威胁的担忧加剧，但公众应信任AI企业。
China’s gasoline and diesel inventories are plunging as refiners grapple with a tightening domestic market, raising the chance the government may move to limit fuel exports again.
中文摘要 中国汽油和柴油库存大幅下降，因炼油商面临国内市场收紧，这提高了政府可能再次限制燃料出口的可能性。
Gold was steady as high oil prices continued to fuel inflation concerns and keep interest-rate hike bets a near-certainty ahead of the Federal Reserve’s policy decision later Wednesday.
中文摘要 金价在每盎司4300美元下方企稳，高油价持续加剧通胀担忧，市场几乎确定美联储将加息，投资者等待美联储周三晚些时候的政策决定。
Apollo Global Management has provided $585 million of financing to flexible workspace provider The Executive Centre, according to a statement, in the global asset manager’s largest Asian hybrid deal combining debt and equity.
中文摘要 阿波罗全球管理公司向灵活办公空间提供商德事商务中心提供5.85亿美元融资，这是该全球资产管理公司在亚洲规模最大的债股混合交易。
Japan’s export growth slowed a tad in August while remaining at an elevated pace, with shipments of semiconductors and chip manufacturing equipment surging.
中文摘要 日本8月出口增速略有放缓但仍保持高位，半导体和芯片制造设备出货量激增，对美出口领涨。
Supporters of the scheme say the bank would enable governments to get cheaper loans to spend on defence projects.
中文摘要 英国正就加入加拿大牵头的全球防务银行进行谈判。支持者称该银行将帮助各国政府以更低成本贷款用于防务项目。
The global AI race is being hit by very localised political and environmental concerns
中文摘要 全球AI竞赛正受到地方性政治和环境问题的冲击，末日警告并非AI发展面临的唯一威胁。
The ad regulator said it had "zero-tolerance" for ads promoting AI tools that could sexualise or objectify women.
中文摘要 英国广告监管机构禁止了宣传「物化女性」的AI应用广告，称对推广可能将女性性化或物化的AI工具的广告采取「零容忍」态度。
Few companies have the technology to make aircraft windows which are getting harder to make.
中文摘要 飞机舷窗制造涉及极端工程技术，随着制造难度上升，掌握相关技术的公司数量极少。
Four recent job starters share their tips on what helped them have a smooth transition and survive their first few days.
中文摘要 四名近期入职的新员工分享建议，谈如何顺利过渡并度过入职最初几天的适应期。
Many complaints were about affordability, after customers saw steep hikes to bills.
中文摘要 英国监管机构收到的关于水务公司的投诉激增84%，许多投诉涉及负担能力问题，因客户账单大幅上涨。
7 回复 · Apple 节点
6 回复 · Apple 节点
4 回复 · Apple 节点
4 回复 · 程序员 节点
8 回复 · 程序员 节点
14 回复 · 程序员 节点
14 回复 · Apple 节点
21 回复 · 程序员 节点
14 回复 · 程序员 节点
10 回复 · Apple 节点
该源今日无内容。