Chinese content10月22日AI日报：Claude Duet 开启等待名单，GPT-6 系列登陆 Google Cloud
这期日报信息量很密：Claude Duet 能帮你下单购物，GPT-6 全家桶上了 Google Cloud，还有 Storm-0940 用 AI 框架批量攻击的细节。
Microsoft微软
United States·Redmond
Copilot and the Azure AI cloud
What they do
Headquartered in Redmond, Washington, Microsoft is a major investor in and cloud partner of OpenAI. It sells models and compute through Azure and builds its Copilot assistant into Windows, Office and GitHub.
Latest coverage
Stories in the feed that mention this company
Chinese contentMicrosoft 披露黑客用开源智能体框架攻击超 2250 个 M365 租户
Microsoft 披露有黑客用开源智能体框架，每分钟试 2900 次登录，攻了 2250 多个 M365 租户，3 小时就能拿到特权权限，AI 驱动攻击的真实案例。
Chinese content微软研究员用 AI 辅助证明 31 年数学猜想 KLS
AI 智能体跑了约 100 条数学路径，帮研究员关掉 1995 年悬置的 KLS 猜想，证明过程怎么分工的这篇讲得细。
Chinese contentMicrosoft 与 CrowdStrike 报告 AI 代理驱动的攻击正在真实发生
AI 攻击从演示进入实战了：微软说有的组织 3 小时就被拿下云权限，阿根廷十几家医院中招，搞安全的都该看看。
Chinese contentMicrosoft 报告：中国背景组织 Storm-0940 用开源 AI 智能体 Artex 自动化攻击
微软点名 Storm-0940 用开源智能体 Artex 打自动化攻击，3 小时内拿下云权限，安全团队得看看这套打法。
GitHub scales Git infrastructure on Azure for agents
Migration to object storage means read throughput can grow separately from Git's traditional monolithic storage, a design choice driven by automated agents hammering repos at volumes human traffic never reached. Platform engineers building CI, mirror or agent systems on top of GitHub will want to study the per-reference quorum mechanism for consistency trade-offs. Azure-based replicas may also affect latency for tools integrating with GitHub APIs from Asian data centres.
Chinese contentSpaceX 盈利、Anthropic 招股书、英伟达 Vera Rubin 利好消息带动科技股反弹
SpaceX 要和特斯拉合并了，Anthropic 招股书曝光半年营收 163 亿美元，美股科技股一天全涨回来了，信息量很大。
IT之家Source
Chinese contentFPT 与 Microsoft 用智能体工程重构日本企业工作流
FPT 和 Microsoft 在日本推智能体工程，讲的是怎么围绕 Agent 重排业务流程并用 ROI 验证，做企业落地可以看看。
Chinese contentDeepSeek V4.1 Pro 预训练数据量或突破 50T tokens
博主在猜 DeepSeek V4.1 Pro 的预训练量会破 50T tokens，还拿 Microsoft 的 MAU 和 OpenAI 的规模做对比，算力党可以看看。
Chinese content微软员工总数十年来首次下降，降至约 22.3 万人
微软财报首次披露员工数十年来的下降，研发部门砍得最多，同时一个季度就砸 410 亿美元建 AI 数据中心，数据很能说明问题。
IT之家Source
Chinese contentMiniMax M3.1 Flash 搭配 SGLang Omni 打造可打断的语音智能体
MiniMax 把自家 M3.1 Flash 和 SGLang Omni 的流式、打断处理能力拼在一起，做能随时打断的语音智能体，做语音方向的朋友可以看看这套组合。
DeepSpeed adds AutoSP, Muon support and auto parallelism at PyTorchCon
AutoTP and AutoEP applied to Hugging Face models mean teams can distribute training without hand-tuning tensor or expert parallelism, lowering the barrier for mid-size labs. Practitioners running multi-vendor accelerator fleets gain a portable optimization stack outside vendor-locked toolchains. As DeepSpeed is open source, the features are immediately available to ASEAN startups building training infrastructure on limited budgets.
