AuraTracer智迹闻
中文

Wednesday, September 9, 2026

One event, one storyline

Multi-source aggregation · faithful, traceable summaries · integrated timelines that restore the full picture

214events in 12h
140active sources
7categories
12hwindow
“Trace the signal. See the whole story.”— AuraTracer

Latest eventsLATEST EVENTS

Products & Apps 2026-09-09 02:29 Updated

Amazon SageMaker Feature Store introduces UpdateRecord for feature-level writes

Amazon Web Services (AWS) SageMaker Feature Store released new features on September 8, 2026, with the addition of the UpdateRecord API. This feature supports single calls to update one or more feature values without needing to read or rewrite the entire record, and is suitable for both standard (DynamoDB backend) and memory (ElastiCache backend) online storage layers. UpdateRecord can atomically merge changes and retain unmodified features, while automatically syncing the full snapshot to offline storage to ensure accurate training data. The API request requires specifying RecordIdentifierValueAsString and a list of Features (up to 100 features per call), with an optional TTL setting; if EventTime is updated, the new timestamp must be greater than the existing value otherwise the request will be rejected. This feature aims to address the read-modify-write issues caused by the traditional PutRecord mode…

Models 2026-09-09 03:00 Updated Launch

Introducing Muse: The World’s First Personal AI Agent Built for Everyone

On September 8, 2026, Meta officially launched Muse, the world’s first personalized AI agent designed for everyone. Built on the Meta Spark model and running on a dedicated Muse Secure VM, Muse can directly perform tasks such as sending emails, booking trips, and negotiating through dialogue. It can also autonomously plan resources to achieve complex goals. In terms of payment and security, Muse integrates Stripe Link and Shop Pay, using a single valid card to protect user information and providing purchase protection. It also supports login with OnePassword. To ensure security, Muse operates independently in the cloud, with network access controlled by the Sentinel proxy system-level isolation. Visibility of passwords and payment methods is strictly limited; all sensitive operations require user authorization, and data is not used for advertising or model training.

Models 2026-09-09 03:12 Updated

Pathway’s brain-inspired architecture development on Amazon SageMaker HyperPod

On September 8, 2026, the AI company Pathway launched its latest large-model architecture – BDH (Dragon Hatchling) – on the Amazon SageMaker HyperPod platform. This architecture abandoned the traditional chain-of-thinking reasoning model and adopted a brain-inspired sparse local interactive neuron graph structure. By using potential space for iterative calculations to update internal memories and decode candidate answers, BDH can handle new problems without generating intermediate text traces or fine-tuning. This design effectively overcomes the challenges associated with traditional Transformer models, such as low推理 efficiency and poor long-term consistency due to fixed context windows, catastrophic forgetting, and high computational costs, while maintaining state adaptation without relying on weight updates during testing.

Policy & Governance 2026-09-09 06:58 Updated

“America’s opposition is ineffective! The world map projection that has been used for over 400 years was suddenly changed by the United Nations.”

On September 4, 2026, the United Nations General Assembly passed a resolution by an overwhelming majority of 164 votes in favor, 6 abstentions, and 1 against. The resolution encouraged the gradual elimination of the Mercator projection, which had been used for over 400 years, in scenarios where accurate representation of area proportions was required, and the adoption of the “equitable Earth projection.” The purpose of the resolution was to address the issue of severe distortion in the areas such as Greenland caused by the stretching of high-latitude regions using the Mercator projection. It also believed that exaggeration of areas in Africa, South Asia, and other regions reinforced Western cultural hegemony. The United Nations emphasized that the resolution applied only to non-navigation scenarios such as education and public communication, and did not mandate the replacement of航海 or mobile navigation systems. The United States was the only country that voted against the resolution, considering it a waste of time by the United Nations. China currently uses the “equal-spacing parallel polyconic projection,” which balances visual and area representation, so the resolution has little impact on China.

Industry & Investment 2026-09-09 08:00 New

美国能源部向NextEra提供19亿美元贷款重启Duane Arnold核电站(615兆瓦)

On September 9, 2026, the U.S. Department of Energy announced a loan of $1.9 billion for NextEra to restart the Duane Arnold nuclear power plant, which had been shut down. The plant has a capacity of 615 megawatts. This loan is intended to support its reopening project, and specific technical details and subsequent plans will be determined based on relevant arrangements.

Industry & Investment 2026-09-09 08:00 New

全文6193字 一、19亿美元复活一座废弃核电站,谷歌AI正在改写能源规则 1.背景+变化:2026年9月

In September 2026, Google plans to launch a $1.9 billion project aimed at reviving a abandoned nuclear power plant. This project will utilize artificial intelligence technology to reconstruct and optimize the energy system in order to address the challenges faced in the transition to traditional energy sources. This move is seen as a major breakthrough in the application of AI technology in the nuclear energy sector, with the potential to reshape global energy production rules. Currently, the project is still in the preliminary planning stage, and specific implementation details and timelines have not been fully announced.

Models 2026-09-09 08:00 New

张宏:在追问反常现象中寻找科学真问题

Zhang Hong, an academician of the Chinese Academy of Sciences and researcher at the Institute of Biophysics, won the Life Sciences Award of the 2026 Future Science Prize. Based on research on Caenorhabditis elegans, his team established a genetic screening system for autophagy in multicellular organisms and revealed the key mechanism by which local calcium signals on the endoplasmic reticulum regulate the initiation of autophagy. Regarding the current technical bottlenecks in the field of autophagy, such as the lack of biomarkers, inability to precisely regulate, and selective removal of substrates, Zhang Hong pointed out that artificial intelligence has limited role at this stage when the basic mechanisms are not fully understood. He emphasized the need to optimize the evaluation system constrained by journal impact factors, return to peer review, foster original innovation, be vigilant against the harms of “resource-based scientific research,” and call for the establishment of a scientific and technological evaluation system, adhering to an original research approach, and maintaining determination in facing the pressures faced by young researchers.

Products & Apps 2026-09-09 08:00 New

英伟达将推出新版RTX 5070显卡:改用GB203芯片 10752个CUDA核心

Nvidia plans to launch a new version of the RTX 5070 graphics card, which will use the GB203 chip. The GB203 chip has an area of 378 mm² and integrates 45.6 billion transistors. The complete configuration includes 10,752 CUDA cores, 84 fourth-generation RT Cores, 336 fifth-generation Tensor Cores, 336 texture units, and 112 raster units. The memory bus width is 256-bit. Currently, it is known that the RTX 5070 will only have 6,144 CUDA cores, approximately 57% of the total capacity, with the remaining space reserved for redundancy. Previously, Nvidia used the larger-size GB205 chip instead of the original GB206 for the RTX 5060. The complete configuration of the GB205 includes 6,400 CUDA cores and a 192-bit memory bus width. There are differences between the GB205 and GB203 in the GPC internal TPC and SM configuration ratios; whether these differences affect heat dissipation performance and performance output…

Industry & Investment 2026-09-09 08:00 New

Runway ARR 突破 2 亿美元:视频热潮转向世界模型押注

On September 8, New York-based artificial intelligence startup Runway announced that its annual recurring revenue (ARR) exceeded 200 million US dollars. This milestone marks a shift in the company’s strategy from being merely a provider of video generation tools to an infrastructure provider offering underlying “world model” technologies. Runway no longer sells only text-to-video clips but also offers simulation technologies that can simulate physical laws and assist in training robots and autonomous vehicles. Despite competition from large-scale cloud providers offering free services, Runway believes that long-term corporate contracts are more sustainable than炒作 cycles. Currently, the company’s valuation reaches 5.3 billion US dollars, with a price-to-book ratio of approximately 26.5 times. Its customer base has diversified to include film studios, advertising agencies, and large enterprise teams such as Chime, Robinhood, and Siemens. Its growth stems from structural factors rather than cyclical fluctuations. The tools have been deeply integrated into production pipelines like “The Whole Universe” and are used in partnerships with Lionsgate and AMC Networks, evolving into model routers…

Policy & Governance 2026-09-09 08:00 New

聚焦科技自立自强·看招丨将重资产转化为行业公共品

The State-owned Assets Supervision and Administration Commission of Shanxi Province issued the “Service Manual for Open Access to Pilot Testing Platforms in Shanxi Provincial Enterprises (2026 Edition)”, offering open access to 9 pilot testing platforms and 25 service items. This initiative aims to transform heavy assets of provincial enterprises into industry public goods, addressing the challenge of facilitating the pilot-to-commercial conversion of university research achievements due to lack of facilities, equipment, and operating conditions. The open platforms cover five areas: resource utilization of industrial solid waste and artificial intelligence in the coal industry. By making capacity parameters and contact information publicly available, it promotes precise matching of resources between small and medium-sized enterprises and research institutions for collaborative innovation. This measure helps to shorten the cycle of achievement conversion, reducing it from annual to quarterly. Relevant measures have been included in the “Action Plan for Systematic Promotion of Scientific and Technological Achievement Conversion (2026–2028)” and are supported by financial policies such as “Pilot Testing Loans” and “Pilot Testing Insurance”.

Products & Apps 2026-09-09 08:00 New

促进民营经济高质量发展丨中科绿碳用绿色材料助力“减碳降塑”

On September 6, Qingdao Zhongke Green Carbon Technology Co., Ltd. officially launched a new type of plant fiber cushioning material. This material is made from agricultural and forestry waste, with properties comparable to plastic foam and capable of complete biodegradation. Relying on the wooden fiber green carbon material team at the Qingdao Institute of Bioenergy and Process Technology of the Chinese Academy of Sciences, the company has mastered core technologies such as physical foaming at normal temperature and pressure and biomimetic cross-linking of plant fibers. Currently, the pilot production line has been fully put into operation. The product is targeted at four major sectors: packaging, building materials, automotive, and logistics. Its high-end wine linings and fruit and vegetable shock-absorbing pads have been verified through actual tests and launched on the market. Over the next two to three years, the company plans to expand its mass production line to support the industry’s deep carbon reduction transformation with independently developed green new materials.

Products & Apps 2026-09-09 08:00 New

郭守敬望远镜:给银河系“测基因”

郭守敬望远镜近期通过高精度观测取得多项突破性成果。该望远镜采用口径 4 米、视场 5 度的反射施密特设计,成功解决大视场与大口径兼得的难题,其球面主镜由 37 块六角形子镜拼接而成,改正镜集成薄平面与拼接镜面主动光学技术,焦面部署的 4000 根特种宽谱光纤已实现国产化。基于该设备对 4 万余颗贫金属红巨星进行的精确测距,科学家找到了 9 个超高速星候选体,并构建了包含近 10 万颗恒星的数据集。利用获取的光谱数据,“测基因”研究不仅发现了恒星级黑洞,还重新画像了银河系,发现其规模比原认知大一倍,同时构建了首幅覆盖全天的银河系三维尘埃消光规律图。截至目前,该望远镜近一年依托自身数据发表论文 417 篇,光谱发布量达 3082 万余条,稳居世界第一。

Models 2026-09-09 08:00 New

科学家揭示溶剂占据调控分子筛微孔催化扩散机制

中国科学院大连化学物理研究所邢嘉成副研究员、徐云鹏研究员及刘中民院士团队研究发现,溶剂分子在孔道内的竞争吸附与空间占据可显著调控大分子进入传统分子筛微孔的扩散过程。该机制被命名为“溶剂门控”。以 TS-1 催化环己烯环氧化为例,甲醇作为溶剂时转化率仅为 6.9%,而移除溶剂后转化率最高可达 88.8%。研究证实,小分子溶剂优先占据孔口及活性位附近空间,从而限制大分子接近内部活性位点。这一规律已在多种钛硅多孔催化剂及不同底物中得到验证。相关成果已发表于《美国化学会志》。

Policy & Governance 2026-09-09 08:00 New

单条政策支持1亿:上海徐汇力促量子计算产业发展

上海市徐汇区印发《徐汇区关于加快推进量子计算产业发展的若干措施》,通过十条支持政策推动产业发展。该政策旨在优化支出结构,重点发展量子计算整机、关键装备仪器、量智融合技术及软件算法等产业生态,并在漕河泾开发区等地推进产业布局。具体措施涵盖加速技术攻关、推广创新产品、鼓励商业化应用、构建开源生态、加大人才引育及强化金融服务等内容。在资金支持方面,对培育重点科研机构、创新企业及打造专业功能平台等提供最高 1 亿元资金;对首台套、首批次支持最高 2000 万元;对人才专项支持最高 100 万元;对开源生态及开放交流分别支持最高 500 万元和 200 万元。

Products & Apps 2026-09-09 08:00 New

引入中创新航 新款理想i6申报:彻底弃用宁德时 代电池

新款理想 i6 已正式向工信部提交申报。该车型彻底弃用宁德时代电池,转而采用欣旺达与中创新航提供的磷酸铁锂电池组,预计电池容量为 87 度,最远续航里程可达 720 公里。新车延续现款造型设计,车身尺寸调整为长宽高 4950/1935/1655mm,轴距为 3000mm,车高降至 1655mm。动力系统提供单电机(250kW)和双电机(150kW/250kW)两种配置。此外,新车配备激光雷达、电动伸缩牵引装置等选装件,并搭载自研的 5C 电池与马赫芯片技术。

Chips & Hardware 2026-09-09 08:19 New

聚焦后端与 Agent 弹性计算 高通与亚马逊达成多代定制芯片合作

2026 年 9 月 9 日,高通与亚马逊宣布达成多代定制芯片合作。双方将共同开发针对特定需求的定制化芯片解决方案,重点聚焦于后端架构优化及引入 Agent 弹性计算技术。此次合作旨在提升系统性能与资源利用率。目前,相关细节尚未在公开信息中披露具体的技术参数或交付时间表。

Industry & Investment 2026-09-09 08:19 New

投后估值超 210 亿欧元 梁琛奇:AI 娱乐产品要让消费过程实时使用算力

2026 年 9 月 9 日,腾讯游戏完成对一家 AI 娱乐公司的收购,交易金额为 210.58 亿美元(约合人民币 1493.67 亿元)。此次交易完成后,该公司的投后估值超过 210 亿欧元。腾讯游戏首席执行官梁琛奇在相关报道中表示,AI 娱乐产品的核心方向是让消费过程能够实时使用算力。

Industry & Investment 2026-09-09 08:19 New

承销团,参与 Oura 上市 Mistral 完成 30 亿欧元 D 轮融资

2026 年 9 月 9 日,AI 公司 Mistral 宣布完成 30 亿欧元的 D 轮融资。本轮融资由 SoftBank Vision Fund 领投,Oura、NVIDIA、Microsoft、Meta、Palantir 等知名机构参与投资。该轮融资将用于加速 AI 大模型研发及全球市场拓展。

Products & Apps 2026-09-09 08:28 New

一字段,一份副本:Elasticsearch 列式存储如何丢弃倒排索引

Elasticsearch 采用列式存储技术重构其核心架构,实现每个字段仅存储一份副本,从而摒弃了传统的倒排索引机制。该架构优化使得 doc values 支持批量读取操作,同时利用跳数器(skippers)进一步提升数据访问效率。

Models 2026-09-09 08:37 New

折叠屏 iPhone 初期产量受限,每日仅数百部;环比增长 379%,腾讯 HY4 登顶全球大模型调用榜;特斯拉时隔 19 个月再降价|极客早知道

2026 年 9 月,苹果折叠屏 iPhone 初期因严格质控每日仅产数百部,正全力爬坡以达年内 800 万至 1000 万台目标。在人工智能领域,OpenAI 发布 ChatGPT Images 2.5,微软推出 Project Opal,Meta 发布个人智能体 Muse,腾讯混元 Hy4 preview 登顶全球大模型调用榜,周调用量环比增长 379%。此外,苹果收购脑成像公司 Sonera 以赋能 Apple Watch;特斯拉时隔 19 个月降价,Model 3/Y 现车起售价降至历史最低以冲刺交付。

Products & Apps 2026-09-09 08:58 New

今年最令人震惊的临床试验失败:降低Lp(a)没降心血管疾病风险

2026 年 9 月 9 日,一项旨在验证通过干预手段降低低密度脂蛋白(a)[Lp(a)]水平对心血管结局改善作用的临床试验结果公布。该研究未能证明降低 Lp(a) 水平能有效降低心血管疾病风险,目标未达成。这一结果挑战了此前认为降低 Lp(a) 可显著减少心血管事件的主流观点,表明单纯降低该指标未必能有效预防心脏病发作或中风等严重并发症。

Models 2026-09-09 00:58 New

Anthropic 开源 Agent Skills:定义可插拔能力模块以扩展 Claude 功能

2026 年 9 月 8 日,Anthropic 公司正式开源发布 Agent Skills 系统。该系统旨在为 Claude 技能提供标准化的定义与扩展指南。Agent Skills 被定义为可插拔的能力模块,本质上是包含特定功能的独立单元。此次发布标志着 Claude 技能系统的标准化进程正式启动,允许开发者通过模块化方式定制和扩展 AI 代理的功能边界,以应对多样化的应用场景需求。

Products & Apps 2026-09-09 07:09 New Launch

灵界 OS v5.0 发布:完成全英文适配并支持一键语言切换

2026 年 9 月 8 日,灵界操作系统正式推出 v5.0 版本。本次更新核心内容为全面完成系统国际化,所有界面元素包括菜单、按钮、对话框及提示信息均已适配英文环境。新增设置项提供语言切换入口,用户可在设置中一键完成中英文切换。该功能支持中英文环境独立运行,切换语言不会影响已安装的应用程序。

Products & Apps 2026-09-09 08:00 New

德国 8 月显卡销量 AMD 大获全胜:RX 9070 XT 单款销量超 NVIDIA 全系

2026 年 8 月,德国零售商 Mindfactory 的显卡销售数据显示 AMD 大幅领先。AMD 共售出 1570 块显卡,占比 55.87%,而 NVIDIA 为 1110 块,占比 39.50%。在架构表现上,RDNA 4 架构占据主导地位,其中单款 RX 9070 XT 销量达 870 块(占总销量 31%),两款 RX 9070 XT 与 RX 9060 XT 合计贡献了超过一半的 AMD 销量。NVIDIA 方面,Blackwell 架构占比 37.0%,Ada Lovelace 系列基本退出市场。在显存规格上,16GB 及以上容量显卡占比高达 75.8%。销量前三型号分别为撼讯 RX 9070 XT Reaper、蓝宝石 RX 9060 XT 脉动及微星 RTX 5060 Ti。

Models 2026-09-09 08:00 New

具身智能数据规模激增但需真实场景验证

多家具身智能企业近期发布阶段性数据成果,行业数据规模显著增长。其中,智元机器人旗下觅蜂科技累计生产超 100 万小时高质量无本体数据;北京人形机器人创新中心开源的 RoboMIND 数据集全球下载量超 2000 万次,年产能可达 18 万小时。尽管数据积累迅速,但行业普遍指出当前数据质量参差不齐,且缺乏真实产能驱动与随机性,尚未完全实现模型验证。为打通从采集到部署的闭环,北京人形数采基地通过复刻家居、商超等场景及专业设备,采用“真机 + 无本体”矩阵式采集模式,上半年交付项目验证通过率超 95%。然而,多家企业认为必须将数据采集延伸至快递物流等真实作业环境,通过 POC 验证及长期迭代积累数据,以形成正向飞轮并提升模型泛化能力与落地效果。

Models 2026-09-09 08:00 New

生数科技发布 MoT 统一架构:视频生成与机器人动作闭环融合

2026 年 9 月 9 日,生数科技推出 MoT 统一架构,旨在构建具身智能的通用大脑。该架构将视频生成与机器人行动耦合为闭环系统,通过联合训练使同一模型既能输出像素级视频,也能执行机器人动作。在 RoboTwin 2.0 评测中,其 Motubrain 模型于双场景测试中排名第一,随机环境下平均得分达 95%。真机测试表明,仅需 50 至 100 条同本体样本即可适配新人形机器人,支持插花、倒水等任务的断点续做与多任务并行处理。尽管在线延迟估计尚未解决且端侧部署存在限制,生数科技仍致力于构建理解、预测与行动协同的通用基座模型。

Industry & Investment 2026-09-09 08:00 New

中国供应链 AI 应用加速落地:采购周期缩短超五成,迈向全球枢纽

根据中国物流与采购联合会发布的《中国供应链发展报告(2025-2026)》,2025 年我国供应链人工智能应用显著加速。采用智能采购的企业相比传统模式,采购效率提升超过 30%,采购周期缩短 50% 以上。在物流领域,人工智能渗透率已超 37%,其中运输环节达 78%,安全合规与智能仓储分别达到 52.73% 和 47.27%。报告指出,我国供应链出海正从“产品输出”阶段迈入“产业链供应链系统性出海”新阶段,加速向“全球供应链枢纽”演进。

Products & Apps 2026-09-09 08:00 New

曹操出行联合豆包推出 AI 打车服务,北京杭州苏州首批上线

2026 年 9 月 9 日,曹操出行与豆包联合推出的 AI 打车服务在北京、杭州、苏州三城正式上线。该服务将 AI Agent 能力延伸至物理世界履约环节,支持用户在豆包端咨询本地生活信息时主动生成出行方案。系统基于场景服务体系,可智能匹配专车或智能大白车,识别轮椅等需求并推荐无障碍专车,同时允许用户直接设定行车路线、空调温度、通风模式及音乐风格等个性化偏好。