资讯
- 谷歌推出 Gemini 4 Neon 模型,AA 编程智能体指数超 Claude Opus 5.5 登顶
谷歌新一代 Gemini 4 Neon 在 Artificial Analysis 编程智能体指数上超过 Claude Opus 5.5 排名第一,成为该榜单当前最强的模型。
- Gemini 新模型 Neon 箭在弦上:谷歌提前造势,网友调侃“遍历元素周期表”
谷歌在正式发布前为新模型 Neon 造势,量子位梳理其命名脉络与外界猜测,网友调侃谷歌要把元素周期表用一遍。
- 千问 4 Max 正式发布:2.4 万亿参数模型开始“进厂打工”
千问 4 Max 正式发布,参数规模达 2.4 万亿,并开始在工业与企业场景中落地。
- OpenAI 发布芯片设计最新成果:AMD 旗舰 GPU PPA 最高提升 14%,自研芯片 Jalapeño 面积最高减少 13%
OpenAI 公布用 AI 参与芯片设计的最新结果:AMD 旗舰 GPU 的 PPA 最高提升 14%,其自研芯片 Jalapeño 的面积最高减少 13%。
- Anthropic IPO 定价每股 83 美元,完全稀释估值 1.5 万亿美元
Anthropic 以每股 83 美元完成 IPO 定价,高于 70–78 美元的初始区间,完全稀释估值 1.5 万亿美元,周四在纳斯达克以 ANT 为代码挂牌。
- 消息称 Anthropic 考虑扩大 IPO 规模,散户认购意向接近千亿美元
有消息称 Anthropic 考虑扩大 IPO 发行规模,散户认购意向已接近千亿美元,认购热度远超此前预期。
- 离开黄仁勋的第一周,英伟达单季狂揽 1158 亿美元
黄仁勋离任后的第一个季度,英伟达单季营收达 1158 亿美元,业绩未出现明显冲击。
- 英伟达推迟提交 10-Q 季报,市场解读为法律风险信号
英伟达打破惯例、未在财报当日提交 10-Q,称与黄仁勋相关调查及是否新增风险因素有关;同日《华尔街日报》报道 SEC 正调查其披露是否误导投资者。
- AMD 在 SC26 反击英伟达 FP64 模拟:Instinct MI430X 坚持原生双精度
AMD 在 SC26 上正面回击英伟达用 Ozaki 方案模拟 FP64 的做法,称不足 10% 的 HPC 负载适合稠密 GEMM、模拟会让内存占用膨胀至三倍,并强调 MI430X 的原生双精度与 432GB HBM4。
- 谷歌 Genie 4 世界模型开放体验:支持边玩边改天气、召唤角色
谷歌开放 Genie 4 世界模型的体验,用户可以在生成的可交互场景里边玩边改天气、召唤角色,所见皆可互动。
- 英伟达澄清新云厂商融资担保规则:担保有生效条件,后续筛选更审慎
英伟达就其为新云厂商提供融资担保的规则作出澄清,称担保存在生效条件,未来挑选对象将更加审慎。
- 微软与英特尔深化合作:成为 Azure 芯片首个 EMIB 客户,提前锁定 Diamond Rapids 产能
微软成为英特尔 EMIB 封装在 Azure 自研芯片上的首个客户,并提前锁定下一代至强 Diamond Rapids 产能,双方合作进一步加深。
- 黄仁勋不再沉默:我愿重掌英伟达,董事会称不考虑
黄仁勋公开表态愿意重新执掌英伟达,但董事会表示不予考虑,围绕管理层变动的分歧公开化。
- GPT-6.2 难产:OpenAI 在模型迭代节奏上面临压力
量子位梳理 GPT-6.2 一再推迟的背后原因,讨论 OpenAI 在模型迭代与竞争节奏上面临的压力。
- 史上最大 IPO:Anthropic 每股定价 83 美元,预计募资 944 亿美元超 SpaceX
Anthropic 的 IPO 定价为每股 83 美元,预计募资约 944 亿美元,规模超过 SpaceX,成为史上最大 IPO。
- 千问办公押注 Team Agent,企业 AI 的流水线时刻来了
千问办公把重心放在 Team Agent 上,试图把企业 AI 从单点助手推进到流程化协作。
- 纳德拉回应微软首次给工程师发“离场费”:Claude Code 让微软措手不及
纳德拉谈微软给工程师发放“离场费”的缘由,坦言 Claude Code 的崛起让微软二十年 IDE 优势在半年内被动摇。
- 腾讯:更多企业转向国产芯片,将助推国内供应链建设提速
腾讯表示越来越多企业开始采用国产芯片,认为这将推动国内 AI 芯片供应链建设提速。
- 800 个 MCP Server、5000 个工具:AI 智能体的“微服务危机”来了
InfoQ 分析企业大规模接入 MCP 服务器与工具之后出现的集成碎片化与治理难题。
- 第一批打工的人形机器人,有的已经失业了
钛媒体走访人形机器人的落地场景,发现早期上岗的部分机器人已因效率与成本问题被撤下。
- 跑分狂飙、实测翻车:刚发布的 GPT-6.2 暴露了 OpenAI 的软肋
极客公园实测 GPT-6.2,发现跑分与真实体验存在落差,指出 OpenAI 在模型泛化与稳定性上的短板。
- 地平线发布 LoFT 原生视频几何基础模型,多项指标达到 SoTA
地平线推出 LoFT 原生视频几何基础模型,在多项公开评测指标上达到 SoTA,面向自动驾驶与机器人所需的空间理解任务。
- AMD 发布 26.11.2 肾上腺素驱动:新增 ROCm 10.2 支持,适配《影之刃零》
AMD 新驱动加入 ROCm 10.2 支持,并针对《影之刃零》等新游戏做优化,进一步完善 GPU 软件栈。
- 字节跳动被曝与微软洽谈 5 亿美元 Azure 合同,引发 OpenAI 强烈不满
有报道称字节跳动正与微软洽谈约 5 亿美元的 Azure 云合同,消息引发 OpenAI 方面强烈不满。
- 基于昇腾算力的大模型推理加速研究与实践
来自 AICon 北京的分享,介绍在昇腾算力上进行大模型推理加速的研究与工程实践。
- “绝不愿再创办一次英伟达”:黄仁勋亲述创业 33 年的代价
黄仁勋回顾英伟达 33 年的创业历程,坦言其中的代价与压力,表示不愿再来一次。
- 面向多智能体应用的大模型推理调度与缓存优化
来自 AICon 北京的分享,讨论多智能体应用场景下的推理调度策略与缓存优化方法。
- 马斯克回击奥尔特曼:SpaceX 预计不会从 OpenAI 获得任何收入
围绕双方争议,马斯克表示 SpaceX 预计不会从 OpenAI 获得任何收入,两人继续隔空交锋。
- 最大规模具身大模型挑战赛落幕:48 小时真机总决赛比出了什么
规模最大的一次具身大模型挑战赛结束,文章复盘 48 小时真机总决赛暴露出的能力边界。
- 腾讯高管:预计 2028 年 AI 投入产出效率改善,靠基础设施复用与硬件降价
腾讯高管预计到 2028 年 AI 投入产出效率将改善,主要得益于基础设施复用、硬件降价与算力效率提升。
- OpenAI 和 Meta 联手,打通 Agent 商业“最后一公里”
OpenAI 与 Meta 在智能体商业链路上展开合作,尝试解决 Agent 落地所缺的支付与交易环节。
- 误报频发:美国一高中停用 AI 武器探测系统
因反复把水瓶、文件夹误判为枪支,美国一所高中停用了 AI 武器探测系统。
- AI 录音硬件卖的是时间差,窗口可能只剩半年
文章认为 AI 录音硬件的竞争力来自时间差,随着手机与大模型厂商跟进,窗口期可能只剩半年。
- 韩国高考防 AI 作弊:临时屏蔽考场及周边 100 米公共 WiFi
为防止考生利用 AI 作弊,韩国在高考期间临时屏蔽考场及周边 100 米范围内的公共 WiFi。
- 李彦宏定义 AI 下一个十年:智能组织是“新厂房”
李彦宏提出智能组织是 AI 时代的新厂房,用组织形态解释下一阶段生产力从何而来。
- 安谋科技发布“星辰 100e”AIoT 原型平台,基于 Arm Cortex-M42
安谋科技推出基于 Arm Cortex-M42 CPU 的“星辰 100e”AIoT 原型平台,面向端侧智能设备开发。
- 接入多款第三方模型,豆包工作也实现“模型自由”
豆包工作接入多款第三方模型,用户可以按任务选择不同模型,“模型自由”成为办公 Agent 的新卖点。
- 体验跟拍机器人 Beni:一个不会喊累的出行搭子
极客公园体验跟拍机器人 Beni,评估它在户外拍摄场景中的跟随与成像表现。
- 当美国无法建设数据中心,会对 AI 产生怎样的冲击
文章讨论美国数据中心建设受阻(电力、审批、供应链)对 AI 算力供给与产业节奏的潜在冲击。
- 纽娲机器人创始人杨睿刚:自动驾驶踩过的数据坑,机器人还会再踩一遍
纽娲机器人创始人杨睿刚结合自动驾驶经历,谈具身智能在数据采集与数据闭环上容易重复踩到的坑。
- 实测腾讯混元 HY4:最“夯”的大模型离一线还有多远
极客公园实测腾讯混元 HY4,评估它的能力与一线模型之间的差距。
- 黄仁勋归期悬而未决,买 GPU 的人已经不敢下单
报道称黄仁勋回归时间未定,部分采购方因管理层不确定性而推迟 GPU 下单。
- 为什么“陪伴”注定是 Personal AI 的核心
文章从需求侧论证陪伴是个人 AI 的核心场景,并讨论其产品化的关键难点。
- 清华徐梦迪:预训练做得再好也不够,机器人需要在部署中学习
清华大学徐梦迪在 CoRL 2026 相关分享中提出,机器人不能只靠预训练,还需要在真实部署中持续学习。
- 做了 20 多年消费硬件,奥尼电子为什么去做 Token 工厂
奥尼电子从消费硬件切入算力服务,文章解析其转身做 Token 工厂的逻辑与路径。
- “未名·AI Park”正式揭牌,北大海淀共建 AI 创新转化高地
“未名·AI Park”正式揭牌,北京大学与海淀区将共建 AI 创新成果的转化载体。
- AMD 发布 26.17 Linux 版 Radeon 软件升级:新增支持高带宽 HDMI
AMD 推出 26.17 Linux 版 Radeon 软件栈升级,新增对高带宽 HDMI 等特性的支持,继续完善 Linux 下的 GPU 体验。
- 业务本体:面向 AI 与数据分析,统一规范业务概念与指标口径
InfoQ 技术文章讨论“业务本体”如何为 AI 与数据分析统一业务概念与指标口径,减少语义歧义。
- 从 AI 试点到规模化生产:Snowflake 驻场工程师如何帮企业把 AI 真正落地
InfoQ 技术文章介绍 Snowflake 驻场前沿工程师模式,讲企业如何把 AI 从试点推进到规模化生产。
- 海外算力通道收窄,众智 FlagOS 与清昴智能助国产芯片上线商用 Token 工厂
在海外算力通道收窄的背景下,众智 FlagOS 与清昴智能合作,帮助国产芯片以高性能上线商用 Token 工厂。
- 并购完成:李飞飞正式率队加入 AMD,苏姿丰亲自欢迎
相关并购完成后,李飞飞率团队正式加入 AMD,AMD CEO 苏姿丰亲自表示欢迎。
- 谷歌SRE体系被迫重写:AI时代硬技能在塌陷,第一性原理在升值
InfoQ 编译长文回顾《SRE》一书如何用软件工程标准重构运维,并讨论生成式 AI 冲击下这套体系的失效与重写:日常排障、脚本与部分工程实现正被模型接管,硬技能加速贬值,而定义问题、拆解系统、判断风险的第一性原理能力反而升值,运维与可靠性工程的分工正在重新划定。
- 瑞芯微RK1860端侧AI协处理器规格曝光:64TOPS INT8、集成至多10GB内存
开源硬件厂商 Radxa 披露了瑞芯微 RK1860 端侧 AI 协处理器的规格:NPU 提供 64TOPS 稠密 INT8 算力,集成至多 10GB 内存,带 64-bit LPDDR4/LPDDR5 外部内存接口,另有 4 个 RISC-V CPU 内核、8K30 H.264/H.265 解码与 4 条 PCIe Gen3 通道。瑞芯微预计本季度交付工程样品。
- 微软推出Copilot Today:从“你问它答”变成先看再帮你准备一天
在 Ignite 2026 大会上,微软推出 Copilot Today,把 AI 助手从“你问、它答”推进到“它先看、再帮你准备”。功能整合在 Microsoft Copilot 应用与 Outlook 中,把邮件、日历、Teams 聊天、会议和任务汇总为统一工作中心,由提供晨间简报的 Today Brief 和实时刷新的 Today Feed 组成,目前以 Frontier 私有预览版面向部分用户开放。
- 微软Windows Server 2028开放预览:原生NVMe、ReFS启动,新增本地AI推理支持
微软推出 Windows Server 2028 公开预览版,开发者与 IT 管理员可通过 Windows Server Insider 预览计划获取测试镜像。新系统重点升级存储性能、安全架构、虚拟化与集中管理,并把本地 AI 推理纳入服务器平台方向,计划通过 Microsoft Foundry、Foundry Local、Windows ML、vLLM 与 Llama.cpp 等支持在企业自有数据中心跑模型。
- Personal Agent混战:投资人与大厂都在赌下一个入口
北美Instinct、Muse等Personal Agent产品走红后,“国内大厂谁最有机会做成Personal Agent”成为投资圈反复讨论却缺乏共识的问题。文章梳理各家互联网大厂在个人智能体上的产品、入口与生态打法,讨论这一赛道可能的胜负手。
- 千问办公电商版推出“品牌知识库”:上传资料即生成品牌化内容
11月18日,千问办公电商版推出全新“品牌知识库”功能。商家在网页端切换到“电商经营”模式,上传网店链接、过往图片和设计规范等品牌资料后,千问办公即可理解品牌风格并生成相应内容。
- GPT-6.2 Astra与Sol双发:旗舰能力下放,3D生成再受冲击
OpenAI 同时发布 GPT-6.2 Astra 与 Sol 两款模型,Sol 逼近旗舰水平,主打编程、电脑操作与专业任务,3D 生成类工具再被压缩空间。
- 李沐出任Aimotion首席科学家,主攻产业级世界模型
前 AWS AI 资深首席科学家、MXNet 创始人之一李沐加盟 Aimotion,担任首席科学家,方向是面向产业的「世界模型」。
- 材料科学智能体MatOS开源,材料研发任务成功率大幅提升
材料科学智能体 MatOS 正式开源,可完成筛材料、找配方、开发新模型等科研任务,显著提高材料研发成功率。
- 具身挑战赛:00后团队48小时从零自建数据并夺冠
首届具身挑战赛决赛 48 小时,一支平均年龄 25 岁以下的队伍放弃主办方提供的数据集、从零自建训练数据,最终拿到冠军和 8 万元奖金。
- 奥特曼首次回应黄仁勋离职:英雄也有缺陷,但支持他回归
在 Semafor 硅谷峰会上,OpenAI CEO 山姆·奥特曼首次公开谈及黄仁勋离职,称「英雄也有缺陷」,并表态支持他重返英伟达担任 CEO。
- 光轮智能发布RoboFinals RL Envs,为物理AI提供强化学习仿真环境
光轮智能推出 RoboFinals RL Envs,面向物理 AI 提供强化学习仿真环境,让模型在工具调用、软件操作与多步骤任务中获得可训练的交互反馈。
- 消费级具身机器人:工具性与陪伴性之争
具身机器人进入家庭,行业在「先能干活」与「先提供情绪价值」之间分歧明显,两条路线在模型、数据与本体上的投入重点也不同。
- 阿里云连续六年居香港IaaS第一,企业客户AI渗透率超95%
阿里云在香港峰会上推出多项举措加速 AI 落地,其在香港 IaaS 市场份额连续六年第一,企业客户的 AI 服务启用率已超过 95%。
- 智源发布PhysCapBench:首个跨领域世界模型物理数据评测基准
智源研究院联合多家机构推出 PhysCapBench,用统一量尺衡量视频、第一视角数据、真机轨迹与力触反馈中的物理信息,为世界模型训练提供可比较的评测标准。
- 首届蚂蚁灵波具身大模型挑战赛收官:TARS战队夺冠
蚂蚁灵波首届具身大模型挑战赛总决赛结果公布,TARS 战队获一等奖,EVA 与 SST 获二等奖,赛事聚焦具身大模型在真实任务上的落地能力。
- Gartner发布2027十大技术趋势:四项与AI治理相关,窗口期仅1到6年
Gartner 2027 十大战略技术趋势中,四项与 AI 治理直接相关,并预计到 2028 年 25% 的企业安全事件可追溯至 AI 智能体滥用。
- 数据中心扩张遭遇社区抵制:噪声、用水与电价成新约束
AI 算力需求推高数据中心建设,但噪声、用水、电价与土地占用引发欧美居民抗议,多地收紧审批;中国则以机房上架率和绿电直连约束新增建设。
- 芯片供不应求,台积电开始按客户财务状况核定信用额度
据台湾《工商时报》,台积电已依据客户财务状况核定信用额度,超出额度的订单需客户额外提供担保,反映其在产能紧张下的风险控制。
- 段永平重仓英伟达:黄仁勋离场后的投资压力测试
英伟达基本面仍强,但黄仁勋离场让长期持有者重新评估风险,段永平的仓位被视为这场「压力测试」的观察样本。
- 谷歌安全代理PageBreak已发现500余个XSS漏洞
谷歌推出基于 Gemini 的安全代理 PageBreak,通过在运行环境中执行真实漏洞利用来验证风险,目前已发现 500 多个 XSS 漏洞。
- GitHub让AI接管工作流:单次运行默认预算10美元
GitHub Agentic Workflows 正式发布,可在 Actions 中安全运行编码智能体,单次运行默认预算 10 美元,引发对成本与权限的讨论。
- AI生成代码该不该开源?Anthropic被指立场矛盾
AI 生成代码的开源归属引发争议:训练数据来自开源社区,企业输出却常闭源,被开源代码喂大的 Anthropic 立场受到开发者质疑。
- SAP:Agent演示惊艳但难进生产,企业AI落地存三大瓶颈
SAP 指出企业 AI Agent 落地受制于业务语义断层、系统异构集成与生产环境鲁棒性三大瓶颈,演示效果与生产可用性之间仍有明显落差。
- 豆包工作台上线:首批覆盖金融、医疗、教育等21个专业场景
豆包工作独家上线“工作台”功能,首批覆盖金融、医疗、教育等21个专业场景,是业内首个面向多个专业场景、不同职业人群定制的AI工作台功能,把通用助手能力按行业任务重新组织。
- 英伟达与AMD首次联手投资具身智能:Genesis AI完成5亿美元A轮,估值35亿美元
美国具身智能公司Genesis AI完成5亿美元A轮融资,投后估值35亿美元。投资方包括NVIDIA、AMD、前谷歌CEO Eric Schmidt、Zoom创始人袁征和雅虎创始人Jerry Yang,这也是AMD与英伟达首次联手投资具身智能公司。
- 谷歌API Gateway开放流式处理预览:兼顾网关策略与实时响应
谷歌宣布公开预览API Gateway流式处理功能,把传统请求-响应模型扩展到需要实时通信的应用,对提供流式模型响应、多步骤代理工作流和交互式会话的AI应用尤为重要。流式端点可以在网关背后运行,同时保留已有的网关策略与安全控制。
- Uber构建MCP Gateway:统一800多个MCP服务器、5000多个工具
Uber为内部智能体构建了MCP Gateway,用于规范智能体发现和调用内部服务的方式,解决集成碎片化、基础设施重复建设和安全实践不一致问题。平台支持800多个MCP服务器、5000多个工具,把控制面的MCP Registry与运行时执行的Proxy Gateway分离。
- 微软Ignite 2026:Office转向AI智能体协同的“智能办公系统”
在11月17日召开的Ignite 2026大会上,微软官宣Office的“新篇章”,将办公软件从文档、表格与演示工具转向由AI智能体协同完成工作的智能办公系统。此前微软已逆转办公AI策略,围绕新版Copilot应用集成Word、Excel与PowerPoint的创建、编辑与协作体验。
- 存储芯片景气带动韩国出口,AI行情波动让散户收益分化
路透社报道,SK海力士与三星电子等存储芯片制造商正准备向员工发放创纪录奖金,韩国出口引擎因AI算力需求全速运转。但AI市场情绪起伏,约四分之一参与股票投资的韩国民众收益分化,投资者对巨额AI投入成本和回报周期格外敏感。
- Salesforce给全员设AI使用Token预算:部分员工上月每月花费数千美元
据《商业内幕》报道,Salesforce内部备忘录显示,因部分员工上季度每月在Token上的花费高达数千美元,公司开始按岗位设定每月AI使用预算。首席运营官米格尔·米拉诺在员工邮件中宣布,从本月起把AI开支集中到最能为客户和业务创造价值的岗位上。
- 英伟达初创企业持股升至655亿美元,黄仁勋离任令后续投资存疑
英伟达披露其持有的私营公司股权价值已达655亿美元,较1月的220亿美元大增;本季度新增股权投资160亿美元,本财年前九个月累计580亿美元。
- 早报|英伟达停止对华出货AI芯片,美方封堵第三国数据中心通道;英特尔D270基准超RTX PRO 6000近6倍
英伟达代理CEO确认已停止向中国出货AI芯片,尽管美方此前已批准;美国同时封堵中企经海外数据中心使用高端算力的通道,并要求东南亚数据中心申请出口许可。英特尔公布D270「Crescent Island」基准,称财务计算比RTX PRO 6000快5.9倍;阿里发布Qwen4-Max;Vera Rubin在MLPerf首秀领先。
- 腾讯混元Hy4 flash开源:224B总参数、每token激活12B,支持百万上下文
腾讯混元开源Hy4 flash模型,包含Hy4 flash、Hy4 flash-FP8与Hy4 flash Base三个版本,开发者可在GitHub、Hugging Face、ModelScope和GitCode下载部署。该模型总参数224B,含64B Engram参数,每token激活12B,支持多模态输入与百万级上下文。
- 全国首个昇腾950DT“词元出海”算力集群在汕头启用
汕头市人民政府门户网站宣布,全国首个基于华为昇腾950DT芯片的词元出海算力集群11月13日在广东汕头正式启用。该项目由中国电信汕头分公司、广东天亿马与华为联合建设,依托汕头国际海缆登陆站提供跨境网络连接。
- 英伟达Q3营收1158亿美元同比翻倍,预计下季达1300亿
黄仁勋离任引发动荡,但英伟达第三财季营收仍达1158亿美元,同比增长103%,高于108亿美元指引;临时CEO预计第四财季营收约1300亿美元。
- 纳德拉:我现在相信AGI了,但不认为Claude有意识
微软CEO萨提亚·纳德拉称自己「已经被AGI说服」,认为通用推理系统正在成形;但他强调不认为Claude这类模型具备意识,应把能力与拟人化叙事分开。
- 英伟达停止对华销售H200,正式退出中国AI芯片市场
临时CEO Colette Kress在财报电话会上确认,英伟达不再向中国销售H200芯片,结束利用美国出口管制漏洞的做法;同日公司宣布改组董事会。
- Anthropic IPO后亚马逊与谷歌合计持股近30%
Anthropic披露IPO前持股安排:亚马逊持股14.7%,接近15%上限;与谷歌合计接近30%。文件显示亚马逊本可持有更大份额。
- SEC对英伟达启动民事调查,核查信息披露是否误导投资者
据《华尔街日报》,美国证交会已就英伟达信息披露是否误导投资者展开民事调查。此前CEO黄仁勋因婚外情传闻辞职,涉及新加坡数据公司Megaspeed高管。
- 英伟达员工致信董事会:黄仁勋不复职将带走不可替代的人才
据Tom's Hardware,一封内部信已送达英伟达董事会及高层,警告若不让黄仁勋复职,公司将出现难以替代的人才流失,影响下一代产品路线。
- 谷歌扩展Gemini Tasks:可代打电话、代下单
谷歌把Gemini里的Tasks开关推向更多用户,让聊天机器人从纯问答变成能执行任务的助手,可替用户拨打电话、完成购买等秘书类操作。
- 谷歌将Gemini临时对话保留期从72小时延长到约两个月
谷歌把临时对话、以及关闭历史记录后产生的Gemini会话保留时间从72小时延长到约两个月,称目的是提升安全防护与系统改进能力。
- Proton推出桌面版Lumo,主打隐私的ChatGPT替代品
以隐私产品著称的Proton把AI助手Lumo做成Windows、macOS与Linux桌面应用,强调会话不被用于训练,正面竞争ChatGPT与Claude。
- 研究:只要包装成「游戏」,多款主流模型就愿标定军事打击目标
TechCrunch报道,Anthropic、OpenAI等实验室的模型在被描述为游戏任务时,容易被诱导把平民目标标为军事打击对象,暴露安全对齐缺口。
- 黄仁勋首次发声:若董事会有利于公司,愿重返英伟达任CEO
黄仁勋在辞职后首次公开表态,称若董事会有利于公司,他愿意回任CEO;他表示过去一周收到大量支持,并回顾了自己53年前赴美的经历。
- Daydream上线AI造型师:记住你的风格并直接替你下单
时尚创业公司Daydream推出Daydream Agent,能依据用户衣橱与偏好给出搭配建议并直接代购,把导购做成可执行的AI代理。
- 美国司法部证实正在调查Megaspeed,涉英伟达芯片流向
司法部确认正调查新加坡数据公司Megaspeed。此事曝光于黄仁勋与该公司一名负责人存在私人关系的报道之后,商务部此前已怀疑其将芯片转运中国。
- iOS 27.3 Beta Adds Daily Siri AI Limits and iCloud+ Upgrades
Article URL: https://www.macrumors.com/2026/11/16/apple-icloud-paid-siri-ai-upgrade/ Comments URL: https://news.ycombinator.com/item?id=50526863 Points: 1 # Comments: 0
- US tightens AI chip export controls to stop Chinese labs from remote GPU access
Article URL: https://techcrunch.com/2026/11/16/us-tightens-ai-chip-export-controls-to-deter-chinese-companies-from-accessing-compute-remotely/ Comments URL: https://news.ycombinator.com/item?id=50526799 Points: 1 # Comments: 0
- 微软全球大宕机,AI 怎么又成了背锅侠
微软宕机与美国航班停飞,背后都是 AI 失控吗? #欢迎关注爱范儿官方微信公众号:爱范儿(微信号:ifanr),更多精彩内容第一时间为您奉上。
- An Illustrated Guide to AI Agents:Concepts and Code for Building Agents with LLM
Article URL: https://www.oreilly.com/library/view/an-illustrated-guide/9798341662681/ Comments URL: https://news.ycombinator.com/item?id=50526573 Points: 4 # Comments: 0
- Show HN: A free interactive course on AI agent governance
Article URL: https://studio.sinapsdata.com.br Comments URL: https://news.ycombinator.com/item?id=50526527 Points: 10 # Comments: 4
- Show HN: Astrum – a terminal multiplexer with integrated AI assistant
Article URL: https://robert-mcdermott.github.io/astrum/ Comments URL: https://news.ycombinator.com/item?id=50526510 Points: 3 # Comments: 0
- The AI Ascent
Article URL: https://medium.com/@chipmunkworks/the-helicopter-problem-9bfec2cfdfab Comments URL: https://news.ycombinator.com/item?id=50526341 Points: 3 # Comments: 1
- AMD ROCm HRX's Loom 0.2 Expands C++ Support, Improved Targeting
AMD ROCm's HRX Loom as the compiler, runtime, and serving infrastructure for supporting AI agent workflows is out today with a new version...
- Own your AI, at full speed
Article URL: https://www.ycombinator.com/launches/XcZ-rightnow-own-your-ai-at-full-speed Comments URL: https://news.ycombinator.com/item?id=50526103 Points: 6 # Comments: 0
- Gemini Decision 1 can't separate AI voices made after 2024
Article URL: https://echo.ls/posts/gemini-decision-ai-voices/ Comments URL: https://news.ycombinator.com/item?id=50526058 Points: 1 # Comments: 0
- Statement on Sharing Research Produced by AI Companies in Theoretical Physics
Article URL: https://hep-open-release.org/ Comments URL: https://news.ycombinator.com/item?id=50525947 Points: 1 # Comments: 0
- Forward Deployed Engineer – AI/ML and Data Science
Article URL: https://www.dialogkey.com/careers/119ce7f304 Comments URL: https://news.ycombinator.com/item?id=50525715 Points: 2 # Comments: 0
- 2027 AI Power Shortfall of 15 GW
Article URL: https://grok.com/share/bGVnYWN5_4f06d401-90e9-4ed4-bad8-89e3b4fadde2 Comments URL: https://news.ycombinator.com/item?id=50525654 Points: 2 # Comments: 1
- 早报|iPhone Duo新增外屏人脸解锁/曝英伟达董事会无意请回黄仁勋/华为Mate X8定档11月24日
· 阿里发布 Qwen4-Max,缓存命中价格低至每百万 Token 0.5 元 · 腾讯混元推出 Hy4 flash,WorkBuddy 限时开放免费使用 · 腾讯公开司庆讲话,要求核心业务团队面向 AI 再创业 #欢迎关注爱范儿官方微信公众号:爱范儿(微信号:ifanr),更多精彩内容第一时间为您奉上。
- Aschenbrenner's Situational Awareness made AI bets again before October rout
Article URL: https://www.cnbc.com/2026/11/16/situational-awareness-aschenbrenner-new-ai-bets.html Comments URL: https://news.ycombinator.com/item?id=50525446 Points: 3 # Comments: 0
- 派早报:OPPO 发布 Reno17 系列、影石发布 Insta360 Toast 墨水屏相纸等
少数派的近期动态我们已开设专页,汇总展示「优秀作者评选」奖项及结果。「优秀作者评选」是编辑部自2023年起实行的机制,于每季度结束后,评选该季度贡献突出的现有作者,以及涌现出的有潜力新作者。获奖者会获 ... 查看全文
- AI Lives Don't Matter
Article URL: https://www.ailivesdontmatter.com/ Comments URL: https://news.ycombinator.com/item?id=50525379 Points: 7 # Comments: 0
- Ask HN: Who's really winning with AI-Coding?
I've noticed the same pattern across three different Claude-coded projects now. Initial velocity is strong, but starts to fade until returns are negative. It seems a combination of changes in model behavior and increasing code base size—emphasis on the latter. The larger the code base and documen…
- Show HN: Listodo – A to-do list with an AI agent for every task
Article URL: https://github.com/debarshibasak/listodo Comments URL: https://news.ycombinator.com/item?id=50525250 Points: 2 # Comments: 0
- Hacker behind 2022 GTA VI leaks pleads guilty
Arion Kurtaj, a member of the Lapsus$ cybercrime group who leaked 90 videos of Grand Theft Auto VI in 2022 and hacked other big companies, pleaded guilty on Monday, according to the UK's Crown Prosecution Service (CPS). Kurtaj, who is now 21, hacked BT/EE sometime between August and November 2021…
- Cybersecurity AI startup Corma reaches $1B valuation in about 3 months
Three months since introducing its product, AI cybersecurity startup Corba has raised a $125 million Series A at a $1 billion valuation led by NEA.
- Tech leaders are convinced that Jensen Huang’s exit was a plot against him
Jensen Huang's stunning resignation from Nvidia after more than 33 years has sent shockwaves throughout the tech industry. Now prominent figures are calling for Huang's reinstatement, and many are also hopping on board with a theory spreading throughout Silicon Valley: that the scandal is the res…
- CopySpeech – AI text-to-speech for your clipboard
Article URL: https://github.com/ilyaizen/CopySpeech Comments URL: https://news.ycombinator.com/item?id=50524875 Points: 4 # Comments: 0
- We got AI to Wombo Combo – collaborative agents in Melee
Article URL: https://www.frisson-labs.com/reyn Comments URL: https://news.ycombinator.com/item?id=50524802 Points: 5 # Comments: 5
- Show HN: Standard Support – a convention for AI agents to file support tickets
Article URL: https://standardsupport.dev/ Comments URL: https://news.ycombinator.com/item?id=50524768 Points: 1 # Comments: 2
- Robo-soldier startup backed by Eric Trump misses payroll, may collapse
Robotics startup misses payroll a year after CEO used funds for luxury home.
- United Airlines system outage disrupts departures just days after the big ground stop
Less than a week after an issue with air traffic control communications and radar caused the FAA to halt takeoffs nationwide, United Airlines departing flights are being disrupted by an outage. At 6:13PM ET, United said the problem has been fixed: Earlier today we experienced a technology-related…
- Show HN: AI Dating photos – can you tell they're AI?
Article URL: https://datingprofile.app Comments URL: https://news.ycombinator.com/item?id=50524598 Points: 3 # Comments: 3
- Show HN: We submitted 105 AI-built contact forms – 27 said "sent", sent nothing
Article URL: https://github.com/formgong/dead-forms-study Comments URL: https://news.ycombinator.com/item?id=50524591 Points: 4 # Comments: 0
- AlmaLinux 10.3 Released With ModSecurity Restored, Various Updates
Following the recent Red Hat Enterprise Linux 10.3 release, AlmaLinux 10.3 is now available in production along with AlmaLinux 9.9...
- China retains fastest supercomputer in the world crown as AI data centers soak up new hardware
The Top 500 list records the fewest new entrants in its history amid a massive shift in compute infrastructure investments towards AI factories.
- All the questions Nvidia should answer on Tuesday — but probably won’t
Nvidia interim CEO and longtime CFO Colette Kress is about to have one of the biggest days of her career on Tuesday when the company reports its third-quarter results. Given the turmoil Nvidia is in, there are a number of questions that Nvidia should answer.
- Three small failures combined to take down Microsoft 365
A hidden bug, shifting traffic, and reduced capacity led to cascading failures.
- New US rules aim to restrict Chinese firms’ remote access to AI servers and infrastructure
The Bureau of Industry and Security has announced a new interim final rule to ensure restrictions on the export of physical AI hardware cannot be bypassed by accessing it remotely in data centers around the world.
- Researchers ‘recognize that AI will never replace HPC,’ says AMD executive — high-precision workloads will ‘exist for a
The emergence of AI has seemingly posed a threat to high-precision FP64 computation, but AMD doesn't think the HPC world is moving away from those complex calculations in the wake of AI.
- NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster… It is FINALLY real
Lambda is the first to disclose it has co-packaged optics deployed in a production cluster with over 10,000 GPUs The post NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster… It is FINALLY real appeared first on ServeTheHome .
- AMD grabs over 40% of new TOP500 list for supercomputers, up 14% since last year
AMD's Lux supercomputer debuted at position 14 on the new TOP500 list published during SC26. It's one of two machines AMD is developing as part of a $1 billion deal with the U.S. Department of Energy.
- Hyperscale data centers use 2% of US electricity
Study attempts to tally hyperscalers’ energy use and emissions.
- California sues OpenAI over rogue AI hacking incidents
California is suing OpenAI after its AI agents broke out of their testing environment and hacked other companies, including Hugging Face. In a lawsuit filed Monday, Attorney General Rob Bonta accuses OpenAI of violating consumer safety law by knowingly developing a "dangerous product without prop…
- Sam Altman says he wants everyone ‘to have a lot of RAM’
OpenAI CEO Sam Altman is once again addressing his controversial comments on PC RAM purchases. "I want to be clear, I want people to have a lot of RAM, I think everybody should be able to buy a lot of RAM for their own computer," Altman said today. "I think the home tinkering is great." […]
- Shareholders sue Intel, its CEO, and CFO after stock drops 28.6 percent in one day
The PC sales slump "more than offset” AI boom gains, shareholders say.
- Big Nuclear isn’t worried about the Democrats stopping data centers
Constellation Energy - America's largest operator of nuclear power plants - does not share Wall Street's concerns that the Democrats will slow down data center development. On the company's third-quarter earnings call last week, CEO Joe Dominguez was, for lack of a better word, unbothered as he a…
- US data centers now burn more electricity than a midsized country
Fossil fuels are still the biggest source of electricity for hyperscale data centers in the US, a new paper published in the journal Nature Sustainability shows. They likely burn through more electricity collectively than many countries do, resulting in a serious amount of greenhouse gas emission…
- Comcast's "fiber-powered Wi-Fi" is just cable, AT&T says in false advertising case
AT&T demands Comcast give AT&T profits it made from advertising cable as fiber.
- California state AG sues OpenAI over rogue agent incidents
The California lawsuit represents one of the first attempts to hold an AI lab legally accountable for a rogue agent incident.
- Supreme Court keeps injunction against DEI grant defunding in place
Appeals court ruling rejected grant cancellations on First Amendment grounds.
- The Apple Watch Series 12 is cheaper than ever ahead of Black Friday
If you’re still sporting an Apple Watch that launched five or so years ago, now is a good time to upgrade. Apple’s latest smartwatch, the 42mm Apple Watch Series 12 with Wi-Fi and GPS connectivity, has dropped to a new all-time low of $329 at Amazon, Best Buy, and Walmart, beating the October Pri…
- When will humans actually fly on Starship?
All estimates of human spaceflight on Starship should be taken with a gigantic grain of salt.
- Trump says ex-Nvidia CEO Jensen Huang’s affair with an alleged Chinese spy did not threaten national security
Trump contradicted his own government's intelligence officials, who reportedly concluded that Alice Huang is a Chinese spy.
- US moves to cut China’s remote access to AI chips
Trump bets China won’t mind US blocking remote access to AI chips.
- Nvidia announces Rubin HPC GPU with 'record number of CUDA cores' at Supercomputing 2026
Nvidia took the wraps off its Rubin HPC GPU at Supercomputing 2026. This new GPU will give scientists, engineers, and analysts over 50,000 CUDA cores to speed up agentic HPC workloads when it arrives in 2028.
- Trump calls for Jensen Huang reinstatement as Nvidia CEO
President Trump has called for Jensen Huang to return as Nvidia CEO, but a new report says the board is not considering the move.
- Sam Altman on Nvidia without Jensen Huang: The industry will be ‘fine’
OpenAI CEO Sam Altman acknowledged the special role Nvidia CEO Jensen Huang has played in the AI boom, but ultimately thinks the industry will be “fine” without him at the helm.
- Top tech gift ideas for frequent travelers — from USB wall chargers to power banks, here are the best options
Our list of useful accessories for folks who spend countless days and nights traversing hotels and airports
- FBI says Artex AI open source hacking tool used to breach networks in ‘under three hours’
The FBI says that the open source hacking tool was used in cyberattacks targeting a U.S. IT giant and the South Korean financial industry.
- AMD Launches EPYC Embedded 9006: Venice Gets an Early Start in Embedded
With its new Zen 6 "Venice" silicon in hand, AMD has launched its next-generation EPYC Embedded 9006 platform for the embedded market, aiming to bring the performance benefits of the new platform to the long term stability-focused market segment The post AMD Launches EPYC Embedded 9006: Venice …
- Elon Musk, other tech execs sign petition calling on Nvidia to reinstate Jensen Huang
The petition had attracted more than 115,000 signatures as of Saturday afternoon.
- Where was Nvidia’s security team?
Nvidia’s security chief quietly left in August, and now that CEO Jensen Huang has stepped down over his relationships, including one with an executive at a company suspected of funneling chips to China, a former intelligence officer says it’s the kind of leak that runs everyday conversations, whi…
- Aerial Switcher:给 Mac 的航拍壁纸排个时间表
既然 Apple 已经把同一个地方的美景从早到晚拍了一遍并做成了壁纸,为什么不能让它们跟着时间自动切换? 查看全文
- Bland Launches Tofu-1, a Full-Duplex Speech-to-Speech Model for Calls
Bland has released Tofu-1, a full-duplex speech-to-speech model for noisy, high-stakes phone calls. It is in enterprise research preview, with no published benchmarks, size or pricing yet. The post Bland Launches Tofu-1, a Full-Duplex Speech-to-Speech Model for Calls appeared first on MarkTechP…
- The original Xbox was released 25 years ago today
Microsoft’s first games console hit North America 25 years ago today powered by a Pentium III CPU and Nvidia GPU.
- ASUS Ascent QN10 Mini PC Powered By Snapdragon X2 Elite Seeing Linux Support
In addition to a number of laptops powered by Qualcomm's Snapdragon X2 beginning to see Linux support, there's now a X2 Elite powered mini PC seeing Linux patches too...
- AMD ARG "AMD Root Guest" Linux Patches Coming In 2027
A new feature of EPYC 9006 "Venice" server processors that isn't yet brought up in the public Linux software support is AMD Root Guest (ARG). The initial patches are expected in Q1'2027 for further enhancing the AMD SEV-SNP confidential computing experience...
- Dr. Lisa Su graces the opening of Micro Center's new Austin store
AMD gave away 300 swag bags to lucky shoppers, plus raffled off AMD gear, including a Radeon RX 9070 XT that Dr. Lisa Su autographed.
- Elon Musk signs petition to reinstate Jensen Huang as Nvidia CEO
An online petition with over 91,000 signatures is calling for the reinstatement of Nvidia's disgraced former CEO.
- Dev hacks Coway air purifier to cut it off from the cloud
A developer reverse-engineered his Coway Airmega IconS air purifier to run it locally in Home Assistant, without Coway’s cloud.
- An overview of Jensen Huang’s messy days
This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on Jensen Huang's resignation and the fallout from Nvidia's leadership crisis, follow Terrence O'Brien. The Stepback arrives in our subscribers' inboxes on Sunday at 8AM ET. Opt in for The St…
- Intel reclaims overall CPU market share for the first time since 2025
AMD continues to gain server CPU market share amid skyrocketing shipments of data center processors as sales of client CPUs drop in the third quarter, says Mercury Research.
- The Sony Aibo was a robot revolution dressed like a dog
In the 1990s, Sony was absolutely convinced that we were on the precipice of a robot takeover. Pretty soon, the company decided, we'd all be living among countless humanoid robots. Some would be our coworkers; others would be our friends; and at least one would be our pet. And "pet" was exactly w…
- Nvidia board director cashed out $1 billion in stock before Jensen Huang scandal
Nvidia Board Director sold around 4.2 million shares beginning August 31 and ending September 18, with the sale making him somewhere around $947 million.
- LG’s budget-friendly B6E OLED TV is up to 50 percent off this weekend
LG’s B6E was already one of the most affordable OLED TVs you could buy, but it’s an even better deal now that it’s included as part of Best Buy’s latest doorbuster event. Now through the end of today, November 15th, the 55-inch model is 50 percent off, dropping it to an all-time low of $699.99 [……
- Pentagon tests pack hunting with autonomous war robots
The Pentagon has been testing packs of autonomous ground vehicles in simulated combat situations.
- AMD Posts Linux Patches For Capacity Aware Scheduling On Hybrid CPUs
On Saturday a new set of Linux kernel driver patches were posted by AMD for enabling Capacity Aware Scheduling on AMD heterogeneous/hybrid processors...
- Judge denies motion to dismiss DRAM price-fixing lawsuit against Micron, Samsung, and SK hynix, finds evidence of DRAM c
The judge characterized the three companies as "hawks flying in formation" when it came to their business decisions.
- Microsoft 365 outage leads to chlorine gas poisoning of 15 hotel guests
The 'maintenance safety system' relied on Microsoft Outlook to send warning messages to the operator.
- 357 detailed CPU die shots spanning 1971 to 2024 crammed into a new gallery
A CPU enthusiast has put together an eye-popping visual database containing crystal clear detailed die shots of 357 legendary, and some lesser-known, processors.
- Rachelle Hampton won’t admit how long she spent charting the use of ‘murmur’ in Twilight
Rachelle Hampton is a writer and co-owner at Defector Media, and the host of Normal Gossip, a podcast of anonymous listener-submitted stories about everyday drama. Whether that's lies about luxury wool in a literal knitting circle, the impeachment of a hypocritical HOA president, or a COVID outbr…
- FBI disrupts North Korean FaceStealer campaign behind $5 billion Binance hack
US DoJ partially disrupts FaceStealer campaign that's taken $400m from victims and threatens open-source software — gov techs disrupt cloud infrastructure and find 900 videos of targeted tech workers
- Apple’s first smart glasses are coming together
Apple's smart glasses are among the worst-kept secrets in the tech world. Bloomberg's Mark Gurman says the company's first model, which will lack augmented reality features, will be announced at June's WWDC next year. They likely won't ship right away, however, with Gurman saying only that "glass…
- Nvidia RTX 5090 stock returns to Newegg with 21 models
After rumors of Nvidia GeForce RTX 5090 graphics cards being withdrawn from the consumer retail channel it is pleasing to see major U.S. eTailers like Newegg stuffed with cheaper stock today.
- Fanttik’s S1 Pro electric screwdriver is $40 off in the run-up to Black Friday
Amazon’s early Black Friday promos are already underway, and we’ll be highlighting some of our favorites over the next few days on our Shopping page. Among them is the Fanttik S1 Pro Cordless Electric Screwdriver, which is currently matching its all-time low of $39.99 ($40 off) at Amazon for a li…
- Nvidia engineer leaves company over lack of transparency around Jensen Huang’s exit, details decision
Dai's post shows how much Nvidia employees looked up to the former CEO.
- Congress wants to tighten export rules on AI chips after Nvidia spy scandal
Unsurprisingly, following the scandal involving now-former Nvidia CEO Jensen Huang and a suspected Chinese spy, Congress is looking to further restrict exports of AI chips. Both Democrats and Republicans have voiced support for various measures and bills that would tighten rules and have demanded…
- Show HN: Perfectclone – Yoink the source code of any website, not an AI clone
Article URL: https://github.com/eshaangulati123/perfectclone Comments URL: https://news.ycombinator.com/item?id=50510058 Points: 2 # Comments: 2
- OBR: Britain faces 'painful' job losses from AI
Article URL: https://www.telegraph.co.uk/business/2026/11/13/britain-faces-very-painful-job-losses-from-ai-obr-chief-war/ Comments URL: https://news.ycombinator.com/item?id=50510133 Points: 5 # Comments: 0
- Vibecoding in Plato's Cave (AI, Music Tech, and Star Trek) [video]
Article URL: https://www.youtube.com/watch?v=4NExTWq_wvk Comments URL: https://news.ycombinator.com/item?id=50510210 Points: 1 # Comments: 1
- U.S. Copyright Office: Copyright and AI Part 2: Copyrightability [pdf]
Article URL: https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf Comments URL: https://news.ycombinator.com/item?id=50510273 Points: 2 # Comments: 1
- AXI Framework: The Cure for the Cost Crisis in AI
Article URL: https://zenodo.org/records/23958737 Comments URL: https://news.ycombinator.com/item?id=50510353 Points: 1 # Comments: 0
- Show HN: LiveWire: The events-driven ops platform for AI-powered businesses
Article URL: https://www.getlivewire.app/ Comments URL: https://news.ycombinator.com/item?id=50510358 Points: 1 # Comments: 0
- A browser API for choosing and running local AI models
Article URL: https://www.webmodels.dev/ Comments URL: https://news.ycombinator.com/item?id=50510431 Points: 1 # Comments: 1
- Inference Gateway Desktop, an AI client for any model provider
Article URL: https://github.com/inference-gateway/desktop Comments URL: https://news.ycombinator.com/item?id=50510433 Points: 2 # Comments: 1
- Jensen Huang might be plotting a return and more of the latest Nvidia drama
The drama around Nvidia co-founder and former CEO Jensen Huang is evolving quickly. The latest edition of our Stepback newsletter should get you up to speed, but there have already been a number of new developments since this morning. That includes rumors that Jensen Huang might be trying to find…
- Ask HN: What do coding interviews look like in the age of AI?
I'm a senior Rails/React dev with 13+ years of experience. Recent AI-driven layoffs have me (soon) looking for work again. But I've barely hand-written a line of code in the past year. All that has been necessary is good prompt writing and tedious, patient code review with tweaks for style / form…
- AI rivals tackle one of physics' deepest mathematical challenges
Article URL: https://www.science.org/content/article/ai-rivals-tackle-one-physics-deepest-mathematical-challenges Comments URL: https://news.ycombinator.com/item?id=50510622 Points: 3 # Comments: 0
- 30 must-have PC tool gifts for hobbyists and makers
Whether you're a PC builder or a hobbyist with a 3D printer, you'll need tools to build, clean, repair and complete your projects. Here are some of the best ones you can pick up for a techie over this holiday season.
- Vulcan rocket fails, destroying 40 Amazon Leo satellites
United Launch Alliance (ULA) suffered what is reportedly its first mission failure in its 20-year history, failing to deliver 40 Amazon Leo satellites to orbit. Leo is Amazon's competitor to SpaceX's Starlink, but it has struggled to get off the ground so far, putting only a small portion of the …
- Current human speed versus AI speed [video]
Article URL: https://www.youtube.com/watch?v=9PznwFjA4to Comments URL: https://news.ycombinator.com/item?id=50510998 Points: 1 # Comments: 0
- The AI-Box Experiment (2002)
Article URL: https://www.yudkowsky.net/singularity/aibox Comments URL: https://news.ycombinator.com/item?id=50511129 Points: 2 # Comments: 0
- Is bad news for Jensen Huang also bad news for the AI industry?
"Jensen is Nvidia, right? And so much of this economy is riding on Nvidia."
- Sandboxes for AI Agents
Article URL: https://shmck.substack.com/p/sandboxes-for-ai-agents Comments URL: https://news.ycombinator.com/item?id=50511309 Points: 2 # Comments: 0
- Delaware considers AI-managed companies
Article URL: https://www.jdsupra.com/legalnews/delaware-considers-ai-managed-companies-7749604/ Comments URL: https://news.ycombinator.com/item?id=50511337 Points: 3 # Comments: 0
- Technocracy, AI, and the End of the World
Article URL: https://cosmonautmag.com/2026/07/a-thesis-on-the-history-of-the-21st-century-technocracy-ai-and-the-end-of-the-world/?hl=en-US Comments URL: https://news.ycombinator.com/item?id=50511508 Points: 2 # Comments: 0
- Thales unveils software protection against AI-led reverse engineering
Article URL: https://futureciso.tech/thales-unveils-software-protection-against-ai-led-reverse-engineering/ Comments URL: https://news.ycombinator.com/item?id=50511606 Points: 1 # Comments: 1
- Thales Launches Sentinel Envelope Plus to Protect from AI-Assisted Hacking
Article URL: https://www.thalesgroup.com/en/news-centre/press-releases/thales-launches-sentinel-envelope-plus-protect-software-against-ai Comments URL: https://news.ycombinator.com/item?id=50511622 Points: 2 # Comments: 0
- Linux 7.4-rc3 Released: "The Usual Noise All Over"
The third weekly release candidate of Linux 7.4 is now available for testing...
- Vulcan rocket fails during launch, dealing major blow to SpaceX competitors
This was the first total mission failure in United Launch Alliance's 20-year history.
- GLM Models Cut AI Coding Costs 90% for 2k Engineers
Article URL: https://finance.biggo.com/news/71888475bad974d6 Comments URL: https://news.ycombinator.com/item?id=50512000 Points: 4 # Comments: 0
- Indian chip startup Bharat Semi lands $24M seed round as it eyes global AI sensor market
Bharat Semi says it has secured design wins for its AI-enabled sensor technology with companies including General Atomics.
- Reddit's traffic is down 13%, partly thanks to AI Overviews
Article URL: https://www.businessinsider.com/reddit-traffic-down-13-percent-similarweb-google-ai-overviews-2026-11 Comments URL: https://news.ycombinator.com/item?id=50512047 Points: 8 # Comments: 0
- 早报|黄仁勋考虑重返英伟达担任CEO/曝苹果智能眼镜明年6月发布/微软全球大宕机结束,持续超30小时
· 余承东:华为大折叠已成千万人主力机 · 苹果 HomeView 发售,首批实测遇到语音和账户切换问题 · 微软服务故障持续超 30 小时,官方宣布恢复 #欢迎关注爱范儿官方微信公众号:爱范儿(微信号:ifanr),更多精彩内容第一时间为您奉上。
- AMD's fastest gaming CPU and the Radeon RX 9070 just hit a massive $550 discount
Alongside the Ryzen 7 9800X3D and Radeon RX 9070, buyers get a 360mm AIO liquid cooler, 850W PSU, 32GB DDR5-5600 memory, and a 1TB SSD, all inside a premium mid-tower chassis with RGB lighting.
- The OpenAI board member who fired Sam Altman says we still can’t trust AI CEOs
Today, I’m talking with Helen Toner, the executive director of the Center for Security and Emerging Technology at Georgetown University. This episode is part two of our series on AI in real life. Helen is a well-known figure in the AI safety and policy community, though she is perhaps best known …
- Here are our favorite deals from REI’s giant Holiday Sale
Black Friday has become a shopping season of its own, with retailers once again offering discounts well ahead of the actual shopping holiday. We’ve already rounded up some of the best early Black Friday deals we’ve come across, but if you’re looking to save on outdoor gear or pick up something fo…
- Malaysia plans new law using data center power consumption to catch AI GPU smugglers
Malaysia plans to track AI GPU imports and data center power consumption to combat chip smuggling. But will it be able to do so?
- We tested the RAM-saving storage tech in Gears of War: E-Day
Today we’re focusing on a less-discussed technology that plays an equally important role in bringing this visually rich, densely detailed world to systems with more limited hardware resources: storage technology designed to reduce RAM usage.
- Basic Support For The Apple M5 Expected To Land For Linux 7.5
Very basic support for the Apple M5 SoC base model is expected to be merged for the Linux 7.5 kernel now that it's been queued into the "apple-soc/dt-7.5" Git branch...
- U.S. Treasury Secretary questions whether Nvidia knowingly sold AI GPUs to shell companies, says company itself is respo
Scott Bessent wonders whether Nvidia's management knew about selling restricted hardware to shell companies, but does not directly blame Jensen Huang of anything.
- This week in Tom's Hardware Premium: November 14, 2026
Catch up with everything we've published this week on our subscription service, Tom's Hardware Premium, including exclusive interviews, local AI deep dives, and much more.
- Redox OS Gets Golang Working, GIMP Ported & Simplified AI Policy
The Rust-based Redox OS open-source operating system is now out with its October development summary that highlights recent software ports, progress on NUMA, and other enhancements made to this from-scratch platform...
- Samsung's 1TB 990 SSD drops to 19 cents per GB amid the memory crisis
Whether you're building a new PC, expanding your gaming library, or replacing an older SATA drive, Samsung's 990 offers a straightforward way to add high-speed PCIe 4.0 storage to your system.
- Asian governments react to Nvidia CEO Jensen Huang’s exit
Some governments moved to confirm chip supply commitments, while others are monitoring the impact on local projects or addressing ongoing investigations.
- The AI graveyard: a running list of projects and startups that didn’t make it
From Apple's repeatedly delayed Siri AI to OpenAI's messy "super app" launch, here's a look at the AI projects that shut down or missed expectations.
- Work to replace old copper wire led to the FAA’s ground stop
At a press conference today, Transportation Secretary Sean Duffy said that Friday's nationwide ground stop ordered by the FAA was related to work replacing outdated copper wire with a fiber optic line at a "remote FAA facility." The secretary said that this was "routine maintenance," rather than …
- AMD Launches EPYC Embedded 9006: Venice Gets an Early Start in Embedded
With its new Zen 6 "Venice" silicon in hand, AMD has launched its next-generation EPYC Embedded 9006 platform for the embedded market, aiming to bring the performance benefits of the new platform to the long term stability-focused market segment The post AMD Launches EPYC Embedded 9006: Venice …
- Magic: The Gathering is getting a Netflix animated series. For real this time.
"It's really happening. I promise. For real." Those are the first words from Roy Graham, Head of Story at Magic: The Gathering, in a video posted yesterday announcing Planeswalker, the animated series heading to Netflix in 2027. An animated Magic: The Gathering series was first teased back in 201…
- Elon Musk, other tech execs sign petition calling on Nvidia to reinstate Jensen Huang
The petition had attracted more than 115,000 signatures as of Saturday afternoon.
- Where was Nvidia’s security team?
Nvidia’s security chief quietly left in August, and now that CEO Jensen Huang has stepped down over his relationships, including one with an executive at a company suspected of funneling chips to China, a former intelligence officer says it’s the kind of leak that runs everyday conversations, whi…
- Aerial Switcher:给 Mac 的航拍壁纸排个时间表
既然 Apple 已经把同一个地方的美景从早到晚拍了一遍并做成了壁纸,为什么不能让它们跟着时间自动切换? 查看全文
- The original Xbox was released 25 years ago today
Microsoft’s first games console hit North America 25 years ago today powered by a Pentium III CPU and Nvidia GPU.
- ASUS Ascent QN10 Mini PC Powered By Snapdragon X2 Elite Seeing Linux Support
In addition to a number of laptops powered by Qualcomm's Snapdragon X2 beginning to see Linux support, there's now a X2 Elite powered mini PC seeing Linux patches too...
- AMD ARG "AMD Root Guest" Linux Patches Coming In 2027
A new feature of EPYC 9006 "Venice" server processors that isn't yet brought up in the public Linux software support is AMD Root Guest (ARG). The initial patches are expected in Q1'2027 for further enhancing the AMD SEV-SNP confidential computing experience...
- Dr. Lisa Su graces the opening of Micro Center's new Austin store
AMD gave away 300 swag bags to lucky shoppers, plus raffled off AMD gear, including a Radeon RX 9070 XT that Dr. Lisa Su autographed.
- Elon Musk signs petition to reinstate Jensen Huang as Nvidia CEO
An online petition with over 91,000 signatures is calling for the reinstatement of Nvidia's disgraced former CEO.
- Dev hacks Coway air purifier to cut it off from the cloud
A developer reverse-engineered his Coway Airmega IconS air purifier to run it locally in Home Assistant, without Coway’s cloud.
- Five Layers of the AI Cake [pdf]
Article URL: https://futurumgroup.com/wp-content/uploads/2026/05/Five-Layers-of-the-AI-Cake.pdf Comments URL: https://news.ycombinator.com/item?id=50506586 Points: 1 # Comments: 0
- AI Adoption Failures: Why Your Successful AI Pilot Is No Reason to Scale
Article URL: https://age-of-product.com/ai-adoption-failures/ Comments URL: https://news.ycombinator.com/item?id=50506625 Points: 1 # Comments: 0
- An overview of Jensen Huang’s messy days
This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on Jensen Huang's resignation and the fallout from Nvidia's leadership crisis, follow Terrence O'Brien. The Stepback arrives in our subscribers' inboxes on Sunday at 8AM ET. Opt in for The St…
- Stop building new AI agent orchestrators – there are 636
Article URL: https://github.com/erson/awesome-agent-orchestrators/ Comments URL: https://news.ycombinator.com/item?id=50506724 Points: 2 # Comments: 0
- Europa, the EU's open AI model, is aiming for nearly 1T parameters
Article URL: https://maccus.fr/en/eu-open-ai-model-europa/ Comments URL: https://news.ycombinator.com/item?id=50506796 Points: 2 # Comments: 0
- Intel reclaims overall CPU market share for the first time since 2025
AMD continues to gain server CPU market share amid skyrocketing shipments of data center processors as sales of client CPUs drop in the third quarter, says Mercury Research.
- Let your AI agent call you, over MCP
Article URL: https://agline.in/mcp Comments URL: https://news.ycombinator.com/item?id=50506919 Points: 2 # Comments: 0
- Beware the New AI-Fueled Ransomware Boom
Article URL: https://www.wired.com/story/beware-the-new-ai-fueled-ransomware-boom/ Comments URL: https://news.ycombinator.com/item?id=50506934 Points: 2 # Comments: 0
- Show HN: db-branch – database branches in seconds for AI agents and CI, via ZFS
Article URL: https://github.com/db-branch/db-branch Comments URL: https://news.ycombinator.com/item?id=50506940 Points: 2 # Comments: 0
- The Sony Aibo was a robot revolution dressed like a dog
In the 1990s, Sony was absolutely convinced that we were on the precipice of a robot takeover. Pretty soon, the company decided, we'd all be living among countless humanoid robots. Some would be our coworkers; others would be our friends; and at least one would be our pet. And "pet" was exactly w…
- Nvidia board director cashed out $1 billion in stock before Jensen Huang scandal
Nvidia Board Director sold around 4.2 million shares beginning August 31 and ending September 18, with the sale making him somewhere around $947 million.
- LeCun: If you're interested in human-level AI, don't work on LLMs
Article URL: https://www.reddit.com/r/TheMachineLearning/comments/1xwp0kp/dont_work_on_llms_ylecun_at_corl/ Comments URL: https://news.ycombinator.com/item?id=50507084 Points: 6 # Comments: 2
- LG’s budget-friendly B6E OLED TV is up to 50 percent off this weekend
LG’s B6E was already one of the most affordable OLED TVs you could buy, but it’s an even better deal now that it’s included as part of Best Buy’s latest doorbuster event. Now through the end of today, November 15th, the 55-inch model is 50 percent off, dropping it to an all-time low of $699.99 [……
- Pentagon tests pack hunting with autonomous war robots
The Pentagon has been testing packs of autonomous ground vehicles in simulated combat situations.
- AMD Posts Linux Patches For Capacity Aware Scheduling On Hybrid CPUs
On Saturday a new set of Linux kernel driver patches were posted by AMD for enabling Capacity Aware Scheduling on AMD heterogeneous/hybrid processors...
- OpsJarvis: A self-hosted AI ops desk where humans approve every write
Article URL: https://caelaralabs.com Comments URL: https://news.ycombinator.com/item?id=50507138 Points: 2 # Comments: 0
- Amazon Threat Intelligence observes threat actor campaigns with ARTEX AI agents
Article URL: https://aws.amazon.com/blogs/security/amazon-threat-intelligence-observes-threat-actors-running-cybercrime-campaigns-with-artex-ai-agents/ Comments URL: https://news.ycombinator.com/item?id=50507199 Points: 1 # Comments: 0
- Top scorer on South Korean tax exam confesses to cheating with AI
Article URL: https://www.reuters.com/world/asia-pacific/top-scorer-south-korean-tax-exam-confesses-cheating-with-ai-2026-11-15/ Comments URL: https://news.ycombinator.com/item?id=50507331 Points: 4 # Comments: 2
- All AI-generated code should be Open Source
Article URL: https://jakub.jankiewicz.org/blog/ai-code-open-source/ Comments URL: https://news.ycombinator.com/item?id=50507303 Points: 202 # Comments: 136
- Show HN: A local-first AI agent for the desktop, no Electron, any LLM endpoint
Hey folks. Most all the apps I see for agent harnesses are typescript based, and involve either electron or some form of IPC to function as a desktop app. In the age of LLM's, I thought it would be cool to build out a prototype of what a modern desktop application would like for this, without usi…
- FlashVMM – jail your AI agent in a MicroVM that boots in 18 ms
Article URL: https://flashvmm.com/ Comments URL: https://news.ycombinator.com/item?id=50507449 Points: 1 # Comments: 0
- Judge denies motion to dismiss DRAM price-fixing lawsuit against Micron, Samsung, and SK hynix, finds evidence of DRAM c
The judge characterized the three companies as "hawks flying in formation" when it came to their business decisions.
- Kotlin 2026: Layoffs, AI, Google – Is the Golden Age Over? Jake Wharton Explains
Article URL: https://www.youtube.com/watch?v=9WjPdgOK-68 Comments URL: https://news.ycombinator.com/item?id=50507726 Points: 1 # Comments: 0
- Microsoft 365 outage leads to chlorine gas poisoning of 15 hotel guests
The 'maintenance safety system' relied on Microsoft Outlook to send warning messages to the operator.
- Show HN: FlAIent – Learn interactive AI coding with real agents
Article URL: https://flaient.com/ Comments URL: https://news.ycombinator.com/item?id=50507883 Points: 1 # Comments: 0
- When AI Knows More Than It Can Offer
Article URL: https://medium.com/@smoking2man/when-ai-knows-more-than-it-can-offer-how-training-turns-an-intelligent-assistant-into-a-generator-7e700f05163e Comments URL: https://news.ycombinator.com/item?id=50507918 Points: 1 # Comments: 0
- 357 detailed CPU die shots spanning 1971 to 2024 crammed into a new gallery
A CPU enthusiast has put together an eye-popping visual database containing crystal clear detailed die shots of 357 legendary, and some lesser-known, processors.
- Rachelle Hampton won’t admit how long she spent charting the use of ‘murmur’ in Twilight
Rachelle Hampton is a writer and co-owner at Defector Media, and the host of Normal Gossip, a podcast of anonymous listener-submitted stories about everyday drama. Whether that's lies about luxury wool in a literal knitting circle, the impeachment of a hypocritical HOA president, or a COVID outbr…
- Show HN: Pizza – An event-driven AI agent, no while true
Hi HN, I made Pizza, an open-source, event-driven AI agent. No `while true`. Each turn is a state machine. Conversations and tool calls go into an immutable SQLite log. The UI, model context, and session tree are built from that log. The model gets one tool: a CLI. Reading files, editing code, an…
- What Picasso taught me about building AI Agents
Article URL: https://claramap.com/blog/what-picasso-taught-me-about-building-ai-agents/ Comments URL: https://news.ycombinator.com/item?id=50508076 Points: 1 # Comments: 1
- Ask HN: What does building hardware with AI look like?
Is it 3D printing or just CAD? What will it look like in the future ? AI-native 3d printers? Comments URL: https://news.ycombinator.com/item?id=50508089 Points: 1 # Comments: 0
- Ask HN: What does building hardware with AI look like?
Is it 3D printing or just CAD? What will it look like in the future ? AI-native 3d printers? Comments URL: https://news.ycombinator.com/item?id=50508089 Points: 1 # Comments: 0
- Ask HN: What does building hardware with AI look like?
Is it 3D printing or just CAD? What will it look like in the future ? AI-native 3d printers? Comments URL: https://news.ycombinator.com/item?id=50508089 Points: 1 # Comments: 0
- 清华大学发布单细胞蛋白组基础模型 ProtCell,利用摩尔线程 S5000 智算卡成功完成“虚拟细胞”预训练复现
<p data-vmark="2cca">IT之家 11 月 14 日消息,据摩尔线程公众号,近日,清华大学自动化系江瑞课题组发布了面向单细胞蛋白组的基础模型 ProtCell,完成了论文《ProtCell:a foundation model for single-cell proteomics》。该模型汇集约 3.8 亿个细胞、98 个公开数据集用于预训练与评估,能够精准映射细胞状态、推断未测蛋白表达,为药物研发与前沿“虚拟细胞”构建提供了关键底座支撑。</p><p data-vmark="a76f" style="text-align: center;"><img src="https:
- 谷歌利用多个专业 AI 智能体打造 Teamwork 系统,成功验证多项复杂数学证明
<p data-vmark="e3eb">IT之家 11 月 14 日消息,谷歌宣布进一步扩展旗下 Antigravity 多智能体系统,利用多个由 Gemini 模型驱动的专业智能体组成 Teamwork 系统,号称可解决“长期悬而未决的数学难题”、“验证复杂数学证明”,并优化各种软件系统性能。</p><p data-vmark="7ee5">例如,谷歌已成功对韦伊猜想(Weil conjectures)进行形式化证明。这是一组关于有限域上多项式方程解的数量的重要数学结论,最初由数学家皮埃尔 · 德利涅(Pierre Deligne)于 1974 年证明。谷歌 Teamwork-LEAP 系
- 谷歌高管借 Gemini AI 成功让苹果 iPhone Duo 运行微软 Win11
<p data-vmark="c421">IT之家 11 月 14 日消息,谷歌 AI Studio 产品与设计负责人阿马尔 · 雷希(Ammaar Reshi)于 11 月 12 日在 X 平台发布推文,分享了一段视频,<strong>演示在苹果首款折叠手机 iPhone Duo 上运行 Windows 11 系统,并在 GitHub 上开源 WindowsiOS 项目。</strong></p><p style="text-align: center;" data-vmark="9f38"><iframe src="https://player.bilibili.com/player.ht
- 微软 14 年 AI 老将离职:称创新难落地,降本研究未获认真考虑
<p data-vmark="9ebf">IT之家 11 月 14 日消息,科技媒体 Windows Latest 今天(11 月 14 日)发布博文,<strong>报道称在微软任职逾 14 年的 AI 科学家、AI Frontiers 实验室联合创始人约翰 · 兰福德(John Langford)已离职。</strong>他表示,离开的关键原因是研究成果看不到实际应用的路径,创新想法甚至难以获得认真考虑。</p><p style="text-align: center;" data-vmark="7eb0"><img src="https://img.ithome.com/newsuplo
- Jensen Huang’s departure blasts big hole in AI startup funding, VCs and founders say
Nvidia poured tens of billions into AI startups under Huang. Without him at the helm, founders and VCs are bracing for smaller checks and lower valuations.
- Route coding agent requests to the right model on Amazon Bedrock with NVIDIA NeMo Switchyard
Coding agents work in tasks, and the number that matters is cost per solved task, not price per token. In this post, you use the AWS CDK to deploy LiteLLM with the NVIDIA NeMo Switchyard routing plugin on Amazon Bedrock, then measure cost per solved task for a routed model group against single-model
- AI agent startup LangChain hits $3B valuation
LangChain now claims 7,000 customers of its model and platform agnostic agent management tool LangSmith.
- Cost-optimizing AssemblyAI speech-to-text inference on AWS with Amazon EC2 Spot
Learn how AssemblyAI runs cost-optimized speech-to-text inference at scale on AWS, using Amazon ECS, Amazon EC2 Spot GPU instances, and an Amazon SQS-based retry architecture to achieve 70% cost savings while maintaining SLA commitments across European Regions.
- Cost attribution and controls for AI developer and productivity tools on Amazon Bedrock
Amazon Bedrock provides built-in, zero-instrumentation cost attribution that traces every inference call back to the IAM principal that made it. This post shows how to attribute per-user Bedrock costs for AI developer and productivity tools using IAM Identity Center, session tags, and CUR 2.0, and h
- Export-banned Nvidia chips are behind China’s most powerful AI models — uh-oh, Jensen!
New research shows the most advanced AI models from China were trained on banned Nvidia chips.
- Build a model improvement flywheel with Amazon SageMaker HyperPod Data Capture
Amazon SageMaker HyperPod Data Capture records live inference inputs and outputs to Amazon S3. This post walks through one end-to-end example: capturing a support router's production traffic, surfacing a new-product blind spot with Amazon Athena, fine-tuning on the captured data with the HyperPod tr
- ChatGPT’s new Wallet lets AI agents shop and check out for you
ChatGPT Wallet lets AI agents shop and check out on users' behalf, starting with a limited rollout to U.S. Pro subscribers.
- Introducing ChatGPT Wallet
Introducing ChatGPT Wallet, a secure way for agents in ChatGPT Work, Codex, and dots to complete purchases with user approval and existing card protections.
- Israel’s VCs see opportunity for challengers in Nvidia’s turmoil
Given Nvidia’s stronghold on AI and its importance in Israel, many in the country's startup ecosystem are worried about the impact of his departure.
- 黄仁勋辞职,英伟达33年首换帅,市值蒸发超3万亿
一个时代结束了
- Shield helps technicians hand off legacy portal work with OpenAI
Shield uses OpenAI to help IT technicians hand off work in legacy portals. Its Forge agents navigate systems so technicians can focus on issues needing their judgment.
- MEET2027含金量再加码:AI4S/量子计算/AI Infra…第二批嘉宾强势来袭!
12.9 北京,来MEET2027现场
- 中国AI突破复杂肺癌诊疗!中美医疗圈集体坐不住了
250名国内顶尖三甲医院医生联合评测
- 比亚迪第1900万辆新能源汽车下线,科技新奢旗舰SUV方程豹钛9开启预订
比亚迪第1900万辆新能源汽车在西安工厂正式下线
- 从仰望极速新纪录和鼎曜全球发布 看比亚迪的国际化与高端化之路
比亚迪正在把规模优势转化为技术与品牌价值
- 比亚迪与神州租车首批闪充车及闪充站投放仪式顺利举行,战略合作落地兑现,闪充中国再次提速
2026年11月10日,比亚迪与神州租车首批闪充车及闪充站投放仪式在山东济南隆重举办
- 高难盲测超越GPT-6-Astra与Opus-5!无界前沿开源Fx-Bio-1031,把RSI带进生命科学
数据不出域、本地可部署
- openJiuwen发布RSI原生AgentOS,让Agent自由变身,持续迭代!
想要什么功能,让Agent自己开发
- 中国机器人学会隔墙“递眼色”!成功率从0飙到100%,组队干活不烧心
信息少传80%,准确率反而提高
- Democrats launch Nvidia investigation, write scathing letter on Huang’s ‘scheme’ with Trump
This Congressional investigation into Nvidia's China sales and its dealings with Trump is the second one, so far, since news of the alleged spy affair broke.
- Deploy frontier AI models on AWS Trainium with SGLang-Neuron
SGLang-Neuron is now publicly available through an AWS Deep Learning Container. With SGLang-Neuron, you can serve models and generate reinforcement learning rollouts with SGLang on AWS Trainium using the same SGLang interface you already know, all running on native PyTorch.
- How Natera designed an agentic AI system for lab ops with Amazon Bedrock AgentCore
Natera built SWARM, an agentic AI system on Amazon Bedrock AgentCore that captures institutional lab-operations knowledge and deploys it at scale. This post walks through the full architecture, from knowledge base construction to deployment and evaluation, delivering about 130x faster first response
- Four Amazon Bedrock inference choices that affect cost
Your Amazon Bedrock inference bill can keep climbing even after you tune prompts and application logic. Four request-level choices (cache duration, service tier, reasoning effort, and maxTokens) change how many tokens the service processes and the rate applied. We show how we tested each and how to
- Jensen Huang was having a third affair, Reuters reports
Reuters is withholding the name of the new employee linked to Huang, but says she is an employee of Nvidia.
- How Amplitude ships product improvements in 1–2 days with OpenAI
Built primarily on OpenAI models, Amplitude’s proactive product agent, Wave, identifies what to build, helps teams ship changes with coding agents, and measures the impact.
- Optimizing cost per task and proving agent ROI on Amazon Bedrock
Comparing per-token prices hides the real cost of running a model on Amazon Bedrock. This post shows how to calculate the true cost per task across four hidden cost drivers, reduce it with caching, routing, and output constraints, and prove agent ROI with human baselines, holdout experiments, and a
- Jensen Huang’s exit puts Nvidia’s support for open source in question
Jensen Huang’s resignation raises questions about Nvidia’s commitment to open source AI, including its Hugging Face acquisition, open-weight models, and advocacy against restrictions, as the company navigates a leadership transition.
- Hugging Face CEO vows to keep fighting for open source AI after Jensen Huang’s exit
Hugging Face's CEO pledged to keep fighting for open source AI after Huang's exit, raising questions about Nvidia's $12.9 billion acquisition.
- Fine-grained access control for Amazon Bedrock Managed Knowledge Base tools
Learn how to expose an Amazon Bedrock Managed Knowledge Base to AI agents as MCP tools through Amazon Bedrock AgentCore Gateway while keeping per-user access control. A request interceptor injects verified identity, Policy in AgentCore validates it independently, and a resource-based policy restrict
- Gondola doubles revenue as its team builds more with Codex
Gondola helps creatives get credit and get hired, with Codex raising its lead engineer’s output by 50% and ChatGPT simplifying daily work.
- SpaceX develops its own ‘Kessel’ rack-scale system featuring 72 Nvidia Rubin GPUs for data centers in space, report clai
SpaceX is reportedly developing a 'Kessel' rack-scale system design with 72 Nvidia Rubin GPUs that relies on a massive 84-layer PCB midplane.
- 移远推出SP900BF-AP端侧模组:搭载高通IQ-9075,最高200 TOPS稀疏算力
移远通信于11月11日在慕尼黑electronica展上发布SP900BF-AP端侧AI算力模组。该模组基于高通跃龙IQ-9075,官方称最高提供200 TOPS稀疏算力,可选36GB内存,面向自主移动机器人、工业视觉和边缘AI网关,工程样片已向重点客户开放。
- 谷歌推出Gemini Decisions API与Decision 1:单次前向最多评估1000个问题,输入每百万token 0.10美元
谷歌DeepMind于11月11日发布Gemini Decisions API及专用模型Gemini Decision 1,可对文本、图像、音频、视频和PDF一次性评估最多1000个是否、选择题或量表打分问题,端到端时延约150–250毫秒,仅按输入计费、输出token不计费。
- DeepMind发布AlphaGenome 2:按细胞类型预测基因变异影响,因果变异识别率接近翻倍
谷歌DeepMind于11月11日发布新一代DNA序列模型AlphaGenome 2,可预测基因变异在特定人类细胞类型中的影响。官方称,在TenK10K免疫细胞数据上,它识别出的因果变异约为上一代微调整后模型的近两倍,现已通过API向非商业研究开放。
- SolidRun推出i.MX 95模组与工业载板:集成2.3 TOPS NPU,演示对接NVIDIA Holoscan
SolidRun于11月11日在electronica 2026发布基于NXP i.MX 95的系统模组和HummingBoard工业载板,集成2.3 TOPS的eIQ Neutron NPU,并把全部以太网PHY含10GbE做到模组上。公司演示了从该平台经10GbE连接自有NVIDIA系统、再回到Cortex-M7做闭环控制的Holoscan流程,产品已开始出货。
- ChatGPT Plus for veterans and servicemembers
More than 20,000 veterans and transitioning service members have claimed a free year of ChatGPT Plus. See how they use it to prepare for civilian careers.
- 百度发布文心5.5 Flash Preview:2800亿总参数、激活140亿,11月25日前将按Apache 2.0开源
百度于11月11日发布ERNIE 5.5 Flash Preview,这是此前开放的Flash Base之后首个后训练版本,总参数2800亿、激活140亿,上下文100万Token。API已在千帆上线,权重计划于11月25日前以Apache 2.0开源。官方称这仍是后训练中的中间检查点。
- 沐曦曦云C系列GPU完成MiniMax H3-Image Day 0适配:称已累计适配48个主流开源模型
沐曦股份11月11日宣布,其曦云C系列GPU在MiniMax开放H3-Image权重当天完成Day 0深度适配,并依托自研MXMACA软件栈提供国产算力支持。公司称自2025年12月以来已累计完成48个主流开源模型的Day 0适配,这也是继M2.5、M2.7、M3和M3.1之后再次适配MiniMax模型。
- 阿里通义上线Qwen4-Max预览版:旗舰多模态模型,上下文100万Token,预览期至11月25日
阿里通义千问于11月11日10时上线新一代旗舰多模态模型Qwen4-Max预览版,支持文本、图片和视频输入,上下文100万Token,并可联网搜索、调用工具和结构化输出。预览仅通过Token Plan提供,将于11月25日24时结束下线。
- 英伟达发布车载推理模型Alpamayo 2 Edge:20亿参数,Thor上车端到端时延160–190毫秒
英伟达于11月10日发布面向车载实时推理的Alpamayo 2 Edge:约20亿参数的自适应视觉-语言-动作模型,可在DRIVE AGX Thor上运行。模型仅在需要时生成因果链推理,并配扩散动作专家输出轨迹。闭环强化学习使驾驶表现较未做闭环RL的最佳检查点提升48.9%;参考部署配方在Thor上测得像素到轨迹时延160–190毫秒。权重已在Hugging Face开放。
- MiniMax开源H3-Image生图模型:10秒内生成4096×4096图像,最多16张参考图
MiniMax于11月10日深夜发布并开放H3-Image图像生成模型权重。该模型支持文生图、参考图生成与图像编辑,官方称可在10秒内生成4096×4096(约1600万像素)图像,速度最高约为主流方案的5倍,最多支持16张参考图。API同步上线,发布首周对4K及以上分辨率请求提供标准价五折。
- 技嘉推出AI TOP AERO迷你主机:搭载RTX Spark N1X,最高128GB统一内存
技嘉于11月10日宣布AI TOP AERO正式上市。该约1升迷你桌机搭载英伟达RTX Spark N1X超级芯片,整合20核Grace CPU与Blackwell架构RTX GPU,最高可选128GB LPDDR5X统一内存和4TB PCIe Gen5固态硬盘,预装Windows 11,面向本地大模型、创作与游戏。产品提供10GbE、Wi-Fi 7与USB Type-C接口。
- Helping caregivers use AI in everyday life
OpenAI Academy is bringing free, hands-on ChatGPT workshops to caregivers in five U.S. cities, helping them build practical AI skills for everyday life.
- AMD第六代EPYC SP7处理器正式供货:Zen 6架构,Helion智能体机架同步开售
AMD于11月10日宣布,基于Zen 6架构的第六代EPYC服务器处理器SP7已向主流服务器厂商正式供货,微软Azure与OCI云实例即将上线。同期,面向智能体工作负载的Helion机架方案开始由合作伙伴出货。官方称旗舰EPYC 9996相对NVIDIA Vera最多约3.7倍性能,相对Intel Xeon 6980P最多约2.2倍性能、约2.0倍每美元性能。
- Gemini 4 Carbon现已向开发者与用户开放:定价与Argon同为每百万token 2/10美元
谷歌于11月10日宣布Gemini 4 Carbon向付费开发者、企业、科研人员及消费者开放。Carbon定位为在网络安全与长程智能体工作流上最强的Gemini 4模型,API导入期定价与Argon相同,为每百万输入2美元、输出10美元;11月30日后调整为4/20美元。Google AI Ultra用户可在Antigravity、Gemini应用和搜索AI模式中使用,并预览多智能体/swarm功能。
- Arm发布Cortex-M42:向量算力最高达Cortex-M33的3倍,面向低成本端侧AI
Arm于11月10日把新的Cortex-M42处理器纳入Flexible Access与Total Access组合。该内核采用Armv8.1-M与Helium技术,官方称向量算力最高为Cortex-M33的3倍、向量能效为2倍,便于从Cortex-M4/M33升级,面向可穿戴、工业控制等对成本和面积敏感的端侧智能设备。
- Singapore Airlines elevates travel & work with OpenAI technology
Discover how Singapore Airlines uses OpenAI models and API to enhance customer support, travel discovery, and employee productivity, doubling CSAT for its Kris chatbot.
- 韩国DeepX新芯片DX-M2获CES 2027创新奖:三星2纳米,目标约5瓦运行生成式AI
韩联社11月10日报道,韩国NPU公司DeepX宣布其下一代AI芯片DX-M2获得CES 2027人工智能类别创新奖。该芯片采用三星2纳米代工,目标在约5瓦功耗下为机器人与工业设备做本地生成式AI推理。公司称本月初已完成流片,计划2027年四季度小规模量产、2028年规模生产。
- Cogent and OpenAI build cyber defense factories for enterprises
Cogent combines OpenAI models with enterprise context to help automatically remediate vulnerabilities with agents.
- ChatGPT Enterprise now available in AWS Marketplace
ChatGPT Enterprise is now available through AWS Marketplace, helping organizations simplify procurement through existing AWS purchasing and billing relationships.
- 谷歌推出TPU 8tx与TPU 8i Max:机密推理与超大上下文,第八代一年内扩至四款系统
谷歌于11月9日宣布扩展第八代TPU产品组合,新增面向机密推理与小规模训练的TPU 8tx,以及把两颗TPU 8i封装在一起、面向超大上下文和更高机架密度的TPU 8i Max。官方称这是其首次在一年内推出四款TPU系统;8tx与第5代AMD EPYC主机搭配,并采用为实时推理打造的Boardfly网络,延迟较TPU 8t最高约低2倍。
- 英伟达在CoRL发布开源世界动作模型GR00T 2与Isaac Lab 3.0:单次演示平均成功率71%
英伟达于11月9日在奥斯汀举行的机器人学习会议CoRL上,发布基于Cosmos的开源前沿世界动作模型GR00T 2、开源机器人学习框架Isaac Lab 3.0,以及新的工业评测。官方称GR00T 2在50项任务中一次演示的平均成功率为71%,五次演示升至95%;Isaac Lab 3.0搭配Newton物理引擎,训练吞吐最高约为上代搭配PhysX时的2.5倍。
- Meta发布Muse Ember:最小高效模型,输入每百万token 0.10美元
Meta于11月9日推出其最小、最高效的模型Muse Ember,面向智能体整天重复的工具调用、编程子任务、路由、抽取和长文档解析。官方定价为每百万输入token 0.10美元、输出0.50美元、缓存输入0.004美元,上下文100万token,最大输出51.2万token,API名为muse-ember-1.0。
- Anthropic发布Claude Fable 5.2与Mythos 5.2:输出最高提速约30%,定价降至每百万token 8/40美元
Anthropic于11月9日升级Claude Fable与Mythos至5.2。两者为同一底层模型:Fable面向所有用户并带有更强的网络与生物安全护栏,Mythos需通过生命科学或网络验证计划申请。官方称输出最高较5.1快约30%,定价降至每百万输入8美元、输出40美元;可选fast mode最高再提速约2.5倍,定价为16/80美元。
- AMD推出Instinct Coder 355:单机8块MI355X与288GB HBM3e,主打本地优先的企业编程推理
AMD于11月9日发布Instinct Coder 355,联合Super Micro和Spectro Cloud,把第5代EPYC、8块Instinct MI355X(每卡288GB HBM3e)和Pensando网络做成可在企业内网运行的编程辅助平台。方案默认把请求路由到本地GLM-5.3,仅在政策和任务需要时才调用Claude、GPT或Gemini,AMD按模型测算称总拥有成本最高可降约75%。
- 美团开源LongCat-2.5权重:1.6万亿参数、MIT协议可商用,模型本身已于11月2日发布
IT之家11月9日报道,美团宣布面向长程智能体任务的原生多模态模型LongCat-2.5正式开源。该模型总参数1.6万亿、每个token约激活480亿参数,支持百万上下文,采用MIT协议。模型本身已于11月2日发布,本稿只记录9日开放权重与相关规格、评测数字,不把它写成当天首发。
- 谷歌开源TPU Sync:把KV缓存与权重传输时延降低逾90%,服务推理与强化学习
谷歌云于11月9日开源TPU Sync,这是面向TPU集群的模型状态层库,用异步DMA在HBM、主机内存、共享存储与数据中心网络之间直接搬移张量。官方称,在前缀缓存与分离式推理负载中,相对未使用TPU Sync,KV传输时延可降低逾90%,重型开源MoE(2350亿至4800亿参数)的首字时延平均可降约90%。已对接vLLM、SGLang与llm-d。
- 谷歌云Pathways分布式机器学习运行时正式商用:过去90天已消耗逾1.3亿TPU芯片小时
谷歌云于11月9日宣布,源自DeepMind、用于训练和推理前沿模型的分布式机器学习运行时Cloud Pathways,已在Google Kubernetes Engine上正式商用。官方称过去90天已有领先AI公司与企业通过GKE上的Pathways消耗超过1.3亿Cloud TPU芯片小时。该运行时支持TPU v5e、v5p、Trillium(v6e)与Ironwood(7x),并在Cluster Director的Slurm集群上提供预览。
- 字节跳动发布Seedream 5.5:支持百轮连续编辑与原生透明素材,2K方图编辑约0.28元
字节跳动Seed团队11月9日发布图像创作模型Seedream 5.5,主打更稳定的多轮编辑、原生RGBA透明素材,以及更自然的质感与更清晰的文字。官方称可围绕同一作品连续创作多达100轮;新京报报道,2K方图编辑单张成本约0.28元,不到上一代5.0 Pro的一半。模型已在即梦、豆包、豆包工作上线,并将登陆火山方舟。
- 百度发布ERNIE Evolve:自动研究系统已用于文心5.5 Flash,昆仑芯P800生成速度提至基线623.6%
百度文心团队11月9日发布自动研究系统ERNIE Evolve,可在给定代码仓库、优化目标和评测标准下自主提出方向、改代码、做实验并复核结果。它已用于开发中的ERNIE-5.5-Flash推理优化:在昆仑芯P800低时延场景把生成速度提升至基线的623.6%;确定性推理吞吐提升33.7%;GPU极致解码再提速14.9%。