聚合官方博客 · 权威媒体 · 深度通讯 · 学术信号。
只做索引，不做转载，标题与链接均指向原始发布方。
📝 今日导语 GPT-6 生成界面，ASML CEO 谈工业数据，LlamaIndex 拆解检索：把产品体验与数据方法一起看。 🏷 今日关键词： GPT-6 智能 UI ASML 工业数据 LlamaIndex 检索 OpenClaw 协作 AI 成本优化 📰 今日精选内容 GPT-6 与面向所有人的智能 UI OpenAI News · 文章 · 评分 90 对话 ASML CEO Christophe Fouquet：EUV、供应链
Perplexity's pplx-embed-v2-late comes in 2 sizes: a 0.6B model built to run on edge devices, and a 9B model for building high-quality indexes. Its best score is 92.4% on MADQA, and its weakest is 61.2% on ViDoRe v3 Markd
GPT-6向所有用户开放，ChatGPT交互升级！OpenAI：模型更聪明，也更难测试 新浪网
HN ⭐ 51 · 16 评论
平均运行成本下降75%！Anthropic轻量级模型上新，瞄准智能体任务 ThePaper.cn
Mistral Large 4 来了：一万亿参数、具备原生多模态能力- OSCHINA - 开源 × AI · 开发者生态社区 OSCHINA
OpenAI：青少年日均使用ChatGPT不足15分钟 联合早报
AI“世纪IPO之战”：OpenAI奋起直追，与Anthropic差距正在缩小 华尔街见闻
苹果筑起的权限高墙，被 Meta AI 助手“借道”绕过 InfoQ-CN
刚刚，OpenAI公开722个数学发现！霍奇猜想、黎曼zeta函数等手稿挤爆GitHub 智源社区
【每日瞰AI】OpenAI向Chat用户推送GPT-6，并加入可交互界面；月之暗面估值据报升至约500亿美元，筹备明年赴港上市 电子工程专辑
OpenAI免费开放GPT-6 Luna，AI行业竞赛转向入口与性价比 虎嗅网
一个抢入口一个拼低价 两大模型厂商同日上新 东方财富
近日，由中国科学院香港创新研究院人工智能与机器人创新中心（CAIR）主办、香港科技园公司协办的第四届香港具身智能医疗科技论坛（The 4th Hong Kong Clinical-driven Robotics and Embodied AI TEchnology Symposium，CREATE Symposium），于香港科学园高锟会议中心隆重举办。 作为衔接临床医学与前沿科技的重要国际交流平台，本届论坛以 “医疗 AI 商业化之路
AI价格战升温，OpenAI紧逼Anthropic 华尔街日报中文版
在机器人领域，一个更大的模型并不能自动解决所有问题。 作者丨马晓宁 编辑丨岑 峰 当地时间 9 月 30 日，IROS 2026 主会在美国匹兹堡落下帷幕，本届大会的主要奖项也在最后一天正式揭晓。 AI科技评论在现场获悉，今年的 Best Paper Award 最终颁给了一项关于机器人“长期记忆”的研究。Yumin Lee、Hyoseok Ju 和 Giseop Kim 凭借《LT-Mem: Volatility-Aware Spat
大模型并没有“吃掉”机器人学。 作者丨 马晓宁 编辑丨岑 峰 9月27日，IROS 2026 将在匹兹堡开幕。重隔31年后，IROS 重回这座钢铁之城。 作为全球规模最大的机器人学术会议之一，今年进入正式议程的论文达到 1933 篇。把这近两千篇论文铺开来看，它更像是一张 2026 年机器人研究的横截面：从强化学习、模仿学习，到路径规划、视觉感知、运动控制、灵巧操作，再到人形机器人、触觉和大模型，过去几年机器人领域几乎所有重要技术路线，
5小时动作碎片，最后变成了机器人的身体直觉。 作者丨 吴思梦 编辑丨 岑 峰 马晓宁 如果你玩过《Mario Tennis》或者《TopSpin》，大概不会觉得打网球是一件特别复杂的事。 球飞过来，移动人物，判断落点，在合适的时机按下击球键。 尤其在《Mario Tennis Aces》里，游戏甚至把网球进一步拆成了一套很直观的动作语言：正手、反手、上旋、切削、吊高球，甚至还有需要精确卡点才能完成的特殊击球。 Mario Ten nis
我的工作就是让AI取代自己！前Anthropic研究员警告AI或灭绝人类 驱动之家
NVIDIA 与微软携手 以 RTX Spark 和 AI 智能体开启 Windows PC 新序章 驱动之家
从结构化表征到无演示学习之路。 作者丨 吴思梦 编辑丨岑 峰 一个盘子，一只水龙头，一小团蓝色的洗洁精。 吴佳俊用这样一个几乎没有技术含量的厨房场景，开始了他在 IROS 2026 的演讲。 一个学生挤了洗洁精，打开水龙头，洗了一半，临时离开。现在，换成机器人接手。第一眼看过去，任务直接明确：把盘子洗干净，再放回去。但机器人真正需要处理的，并不是“洗盘子”这三个字。 它得先知道，眼前哪个东西是盘子；得知道这个盘子到底脏不脏；得注意到盘子
过去一年，“物理AI”几乎成了自动驾驶、车企共同争夺的下一张船票。 世界模型、具身大模型、机器人数据……众多公司从不同切入口布局物理AI，试图成为新风口之下的主角，相比之下，如祺出行选择了一条不那么显眼的路径，不自己训练大模型，而是尝试成为模型背后的数据基础设施。 实际上，如祺出行在2023年就开始思考，除了提供包括Robotaxi等出行服务，一家出行平台还能在AI产业里扮演什么角色。随后两年，如祺出行逐步完成数据闭环，一端连接驾驶员、
警告“AI或灭绝人类”的前Anthropic研究员：某种意义上，我曾研究让AI取代自己 凤凰网科技
从鄂尔多斯市区一路向西抵达边陲小镇棋盘井，车窗外，密集的商业住宅渐渐被平原和山体取代，路上的乘用车越来越少，拖着长车厢的重卡却越来越多。 煤矿、电厂、工业园区之间，重卡拖着大量货物沿着固定路线流动。这里，也是卡尔动力验证无人货运的主要场景之一。卡尔动力五年前来到内蒙古，尝试编队无人驾驶。如今，它已经拥有百台量级的无人运输车队，业务从内蒙古延伸至新疆、甘肃、陕西、山西等能源运输区域，并开始测试没有驾驶舱的运输机器人。 9月22日，卡尔动力
AI价格战升温：OpenAI借GPT-5.6逆袭，Anthropic优势不再？ 凤凰网科技
“AI央行”英伟达的“千亿资本局” 华尔街见闻
Unsloth's October 6 security overview explains how Studio checks code, weights, packages and tools before anything runs. Custom model code is scanned and approval is bound to its fingerprint. Flagged weight files are blo
ChatGPT聊天框长出界面
小模型新守门员
纽约时报被指在OpenAI版权案中漂白证据 资助研究未披露 虎嗅网
Anthropic发布Claude Haiku 5.5 平均运行成本较4.5版本低约75% 观点网
Anthropic发布最新AI模型ClaudeHaiku5.5 新浪财经
Anthropic估值剑指2万亿美元分析师泼冷水：这是“2026年最荒谬IPO”_全球市场_财经 证券之星
“两万亿Anthropic”要翻车？机构：2026年最荒谬IPO，只值1500亿美元！ 财联社
Get ready for more AI in your Windows and more AI on the desktop.
Anthropic发布Claude Haiku 5.5，运行成本较前代降75%；三星电子Q3营业利润预计超100万亿韩元，同比大增近8倍｜全球科技早参 www.mrjjxw.com
I've always been kind of into computers since I was young. And then when film started to move from analog film to digital, I became more interested in that aspect of it. And the visual effects workflow for many years has
博通为OpenAI定制AI芯片寻求逾500亿美元融资 观点网
OpenAI全面上线GPT6！ChatGPT迎智能UI重磅升级：免费用户也能用 驱动之家
While much of the automotive world sits dumbfounded as China gobbles up all its customers, BMW continues to roll out extremely well-crafted, technologically advanced electric vehicles that impress in both drive quality a
Opus-built Creative Cloud alternatives are ambitious, free, and nowhere near finished.
OpenAI正式发布GPT-6，抵御“越狱攻击”能力更强，但均未触及“RSI”门槛 华尔街见闻
Teenage Engineering founder and CEO Jesper Kouthoofd told Highsnobiety that it plans to stop making synths. And yes, that includes the iconic OP-1, which put the company on the map. TE isn't just a music gear maker; it's
📌 One-Sentence Summary OpenAI 部署工程师建议以成功完成任务的成本为目标，通过代表性评测选择模型配置，再用缓存、程序化工具调用、推理控制和延迟处理减少工作量。 📝 Summary OpenAI 部署工程师 Mandeep 和 Sapto 解释，token 标价无法完整反映 AI 应用成本。便宜模型可能消耗更多 token、无法完成任务，或需要人工介入，因此应关注在达到准确率要求、延迟可接受时，成功完成一项任务
📌 One-Sentence Summary ClawLabs 演示多人共享的持久智能体工作流，并介绍结合协作、维护循环与可执行部署边界的开源控制平面 OpenClaw Enterprise。 📝 Summary ClawLabs 介绍团队如何从终端编码会话迁移到 Codex Desktop、分布式测试机器，最终转向共享的持久智能体服务器。更多算力并未消除由人替同事和智能体传话的瓶颈；团队可见的会话让代码变更背后的意图得以共享，也让贡献
📌 One-Sentence Summary Qodo CEO Itamar Friedman 认为，AI 代码评审需要组织经验、有证据支撑的智能体群、事故学习，以及超越单个 PR 的任务级治理。 📝 Summary Qodo CEO Itamar Friedman 将公司描述为编码智能体周围的上下文与治理层：收集工程团队的隐性知识，既指导代码生成，也用于审核变更。他区分智能与基于经验的智慧：局部看似合理的代码，仍可能因忽略数据库行为或
📌 One-Sentence Summary OpenAI 的 Jason 主张用持久智能体、共享上下文、页面指令和可复用插件扩展团队协作，减少个人逐项协调工作的负担。 📝 Summary OpenAI 开发者体验团队的 Jason 描述了从个人 AI 辅助，走向让同事和智能体共同推进工作的系统的变化。工具和并行任务增多后，他起初仍要协调每次上下文传递、跟进和自动化。他以一场假设的用量限制问题调查为例，把解决方法分为收集上下文、组织和推
📌 One-Sentence Summary OpenAI 展示三个内部 ChatGPT 工作流，通过插件、Spaces 和持续运行的 Dots 委托监测、重复研究与知识服务，同时保留人工审核。 📝 Summary OpenAI 开发者体验团队介绍了三位内部重度用户如何把 ChatGPT 从一次性助手变成持续协作的队友。Dani 委托 Dot 监测产品发布：收集社交帖子与互动数据，维护报告站点和内容表格，并提醒她关注紧急 Slack 对
As previously promised , here's Anthropic's new fast, low cost model: Introducing Claude Haiku 5.5 . The previous Haiku, 4.5, was very much showing its age. It came out almost a year ago , and was priced at $1/million in
The developer of Hermes Agent raised a $90 million Series B.
A common misconception I've seen with Googlebooks is that the Quick Insert and Google logo keys are new. They're not, as they first debuted a couple of years ago on some Chromebooks. But there is something new for Google
Anthropic's Claude Haiku 5.5 starts at $0.10 per million input tokens, keeps 1M context, scores 72.4% on OSWorld. The post Anthropic Releases Claude Haiku 5.5: A Small Model With 1M Context Priced at $0.10 per Million In
Getting in on the Prime Day action doesn’t have to result in an empty wallet. Just as we found a couple handfuls of goodies that are under $25, there are plenty of deals during Amazon’s October sale that are $50 or less,
Microsoft revealed the specs and price for its Surface Laptop Ultra, AI PCs that run on Nvidia chips that are designed to run AI models and agents.
太炸裂！Anthropic二季度营收达逾115亿美元 同比实现14倍暴涨 财联社
Just over a week ago, we covered a deal at Amazon that knocked around 30 percent off Roku’s new (and first-ever) OLED TVs. The deal expired, but has returned for the final day of October Prime Day. The 55-inch Pro Series
OpenAI is launching a new Intelligent UI feature in ChatGPT that allows the chatbot to answer your questions with interactive visuals. The update, which is rolling out to all users alongside GPT-6, gives ChatGPT the abil
OpenAI is rolling out GPT-6 with "Intelligent UI," a feature that turns answers into interactive interfaces with charts, buttons, and forms. The model can now respond while still thinking, cutting wait times by 44 percen
The first array of laptops powered by Nvidia's new RTX Spark chip are designed to compete with high-end MacBook Pros - and they've got some high-end prices to match. The flagship Surface Laptop Ultra and RTX Spark laptop
Claude Haiku 5.5 is now available on Amazon Bedrock and Claude Platform on AWS. According to Anthropic, it is the fastest, most efficient model in the Claude 5.5 family, built for subagents and high-volume, cost-sensitiv
HN ⭐ 330 · 316 评论
Anthropic's new Claude Haiku 5.5 crushes its predecessor in benchmarks, jumping from 15.7 to 72.4 percent on the OSWorld computer use test. Token prices drop by up to 90 percent, though a new tokenizer eats into some of
For the first time in years, the Internet Corporation for Assigned Names and Numbers - better known as ICANN - is accepting applications for new top-level domains. These are the suffixes at the end of all URLs, and you m
At a Microsoft event in San Francisco on Wednesday, NVIDIA founder and CEO Jensen Huang and Microsoft CEO Satya Nadella outlined how NVIDIA and Microsoft are co-engineering hardware and software for AI agents to run on W
Three engineers put GPT, Claude, and Grok in charge of a real car. Only one of them was successful.
Microsoft just wrapped up a big Windows and Surface-focused keynote in San Francisco. The biggest announcement was arguably the release details about the Surface Laptop Ultra, its new laptop that’s powered by Nvidia’s RT
Enterprise RAG unlocks insights from knowledge sources like SharePoint, Google Drive, and Confluence, but those sources carry complex permissions. Learn how Amazon Quick and Amazon Bedrock Knowledge Bases enforce documen
Meta’s AI agent Muse is now available on iPad, just a month after its mobile debut, as the company rapidly expands the assistant’s reach and integrations.
Liquid AI has released Open d1, two open-weight multimodal models in its d1 decision model family. d1-3B reads text and images. d1-omni-600M reads text with an image, or text with audio. Neither model writes text. Each r
Anthropic发布最新AI模型Claude Haiku 5.5，用于执行那些对成本敏感的任务 东方财富
ChatGPT’s teen safeguards are meant to protect vulnerable users, but new testing found the chatbot continues encouraging engagement during crises and potentially encourages unhealthy relationships with the AI itself.
📌 One-Sentence Summary OpenAI 宣布在 ChatGPT 中全面推出 GPT-6 及「智能界面」（Intelligent UI），重点优化视觉化回答与交互工具体验。 📝 Summary OpenAI 已开始向所有 ChatGPT 用户推送 GPT-6，并同步上线名为「智能界面」的新功能。此次更新旨在为日常问题提供更快速、更具视觉感的响应，并通过按需出现的交互式元素简化复杂话题的理解过程。 📊 Article M
HN ⭐ 788 · 389 评论
At today's Windows and Surface event, Microsoft showed off an upgrade to its Copilot AI system that will give it access to local files on your PC and the ability to take actions across the OS. It's part of an idea Micros
HN ⭐ 569 · 296 评论
OpenAI is launching a new user interface that will bring interactive visuals to ChatGPT.
OpenAI is updating ChatGPT for all users with an “Intelligent UI” that’s more visual—generating interactive elements as part of the chatbot’s outputs.
作者: Shravan Chaudhari, William Paul, Suchi Saria | As we move through the world and carry out everyday tasks, we encounter objects that may become relevant only later. We are capable of recalling where we left something
作者: Saif Punjwani, Micah Goldblum | Modern language models undergo reinforcement learning with verifiable rewards (RLVR) on top of already-trained checkpoints. A key promise of RLVR is the discovery of new reasoning stra
作者: Hongru Cai, Ran Wei, Wenjie Wang | Conditional memory architectures such as DeepSeek Engram use input n-grams to look up learned embeddings, expanding the capacity of large language models (LLMs) with limited additio
Biohub, the research organization backed by Mark Zuckerberg and Priscilla Chan, is coordinating a $1.8 billion initiative to train AI models that predict cell behavior. Meta, Google DeepMind, Isomorphic Labs, and the US
作者: Wei Huang, Bohan Zhang, Chenzhi Liu | Real-time robot control demands enough visual history to infer motion and task progress, but processing that history can delay action. We present Long-WAM, a model-system framewo
作者: Aleksandar Armacki, Haoyuan Cai, Ali H. Sayed | Heavy-tailed noise has been widely observed in modern machine learning, motivating the use of methods like gradient clipping and normalization. While these methods are
Man gets 18 months for AI scheme stealing $8M in music streaming royalties.
作者: Mikey Watts, Yuchen Cui | Vision-language-action models (VLAs) are strikingly sensitive to instruction phrasing and do not inherit the language robustness of the vision-language models they are built on. A one-word e
作者: Zhewei Chen, Hao Zhu, Jiaojiao Jiang | GNN-to-MLP distillation aims to retain the predictive accuracy of a message-passing teacher while deploying a graph-free MLP at inference. Existing methods mainly transfer node-
作者: Luka Radić, Vikrant Singhal, Amartya Sanyal | Machine unlearning asks for a deletion algorithm whose output is close to retraining from scratch without the selected forget examples. In this work, we study forget-only
作者: Artem Zholus, Nicolas Beltran-Velez, Jianhao Yuan | Latent world models have shown a remarkable ability to predict future states and to plan in the real world. In practice, however, we lack a principled way to estima
作者: Yinling Zhang, Langchen Liu, Dongbin Xiu | Language-model agents are increasingly asked to carry out open-ended scientific research, yet their results are usually graded against a known answer, a rubric, or a languag
Google's AI watermark detector SynthID is now public. Anyone can check whether images, videos, or audio were created by Google's AI or partners like OpenAI and Nvidia. More than 180 billion pieces of content already carr
HN ⭐ 210 · 97 评论
HN ⭐ 52 · 44 评论
Meta launches new AI tools after discovering ads on its platforms that may look normal but direct users to harmful content elsewhere online.
OpenAI称面向青少年的ChatGPT即将上线大学规划工具 东方财富
Training AI agents with reinforcement learning can be challenging because their tools, context, and decision-making are managed by complex frameworks. Agent Lightning connects existing agents to RL training, making it ea