←搜索结果

搜索“Microsoft”

按匹配度 · 共 1129 条结果 清除搜索

星期四

  1. Epoch AI：Gradient Updates

    Epoch AI 分析：Mythos 的网络能力被夸大了吗

    Epoch AI 汇总约十五个网络安全基准与 Project Glasswing 的真实漏洞数据，评估 Anthropic 的 Claude Mythos 系列网络能力是否被夸大。其自建的 Cyber-ECI 显示 Mythos Preview 在漏洞利用开发上明显领先 GPT-5.5，约超前趋势 7 个月，Mythos 5 仅小幅提升。但在固定预算下的漏洞发现能力尚不明确，Project Glasswing 伴随最高 1 亿美元 API 额度投入，curl 等高度审计代码库中 Mythos Preview 仅发现一个低危漏洞和四个误报，其优势更多体现在更低的误报率和更好的漏洞优先级判断。

    le to fully complete “Cooling Tower” on 3/10 attempts (with all other models at 0/10), as they do not give its average performance we are not able to incorporate this information. Microsoft CTI-REALM Description: This is the only benchmark explicitly focused on cyber defense. Given cyber threat intelligence reports, it tasks models to generate detection rules to apply on endpoint/cloud telemetry logs. Scores are given as average ‘Reward’ in [0,1] based on how well their decision rules function.

    匹配字段正文

    #analysis#现象/趋势#安全/对齐#评测/基准

星期一

  1. Claude Blog

    Anthropic 发布 Swift 包，让 Apple 开发者通过 Foundation Models 框架调用 Claude

    Anthropic 发布新的 Swift 包，为 Apple 的 Foundation Models 框架提供 Claude 支持，开发者可在本地模型处理摘要、抽取等任务后，把需要多步推理、代码生成或联网搜索的请求交给 Claude。该包处理流式输出、工具调用和结构化响应，可直接回填到 SwiftUI 视图，需使用 Anthropic API key，支持 iOS 27、iPadOS 27、macOS 27、visionOS 27 和 watchOS 27。

    rtifacts Artifacts Design Design Design Connectors Connectors Connectors Plugins Plugins Plugins Skills Skills Skills Claude in Chrome Claude in Chrome Claude in Chrome Claude for Microsoft 365 Claude for Microsoft 365 Claude for Microsoft 365 AI agents AI agents AI agents Code modernization Code modernization Code modernization Coding Coding Coding Commerce Commerce Commerce Customer support Customer support Customer support Cybersecurity Cybersecurity Cybersecurity Legal Legal Legal Sales Sale

    匹配字段正文

    #lab#产品更新#Agent#MCP/工具调用

星期四

  1. The Register

    微软发布 Autopilot 智能体类别，首个智能体 Scout 亮相

    微软在 Build 大会上发布名为 Autopilot 的新一类智能体，首个智能体 Scout 可在 Teams、Outlook、OneDrive 和 SharePoint 间运行，连接聊天、邮件、日历和联系人数据，并能自主安排会议、标记重要会议、生成准备材料、识别临近截止时间并预留日程。Scout 由 OpenClaw 驱动，绑定 Entra 身份并受组织访问控制约束，目前仅向少数客户和 Frontier 计划组织开放预览。微软称其具备企业级安全与管控，但文章指出 OpenClaw 的安全口碑存疑，且未说明针对常见 AI 攻击的其他防护措施。

    AI AND ML Always-on agent promises to keep work moving, provided you trust it with practically everything Move over, Copilot: Microsoft is introducing a new category of agentic AI called "Autopilot," starting with Scout, its first agent. And it doesn't take much guessing to understand how Microsoft expects these things to operate: By constantly watching your every move and taking action in the background to ostensibly streamline your workday. Microsoft announced Autopilot, and the first Autopilo

    匹配字段正文

    #hot-signal#产品更新#Agent#Microsoft

星期三

  1. Georgi Gerganov

    NVIDIA 与微软在 Computex 力推本地 AI

    今年 Computex 上本地 AI 释放出强烈信号。NVIDIA 和微软等大厂正在拥抱并讨论本地 AI 工作负载。专用消费级硬件和模型即将到来。

    Strong signal for local AI on this year's Computex. Big players like NVIDIA and Microsoft are embracing and discussing local AI workloads. Dedicated consumer hardware and models are on the way.

    匹配字段正文

    #editorial#现象/趋势#端侧#部署/工程
  2. DevClass

    Embarcadero 为 Delphi 和 C++ Builder 推出智能体 AI 助手 Kai

    Embarcadero 发布 Kai，一款面向 RAD Studio（Delphi 与 C++ Builder 的 IDE）的智能体 AI 助手扩展，提供聊天、代码补全和 MCP server，需用户自备 API key 调用第三方 LLM，订阅价每开发者每年 249 美元。Kai 支持 Ollama 和 LM Studio 等本地 LLM，实测存在一些粗糙之处，如 LM Studio 中可用的查询在 Kai chat 中报错。

    s released in 1995 for 16-bit Windows, and was at the time the most advanced rapid application development tool available, with a drag and drop visual form designer reminiscent of Microsoft’s Visual Basic (VB), accompanied by a native code compiler which, unlike VB, generated executables that did not require a separate runtime engine. C++ Builder came a little later, with the same visual designer but supporting code in C++ rather than object pascal. Why is Delphi still around after more than 30

    匹配字段正文

    #hot-signal#产品更新#Agent#MCP/工具调用

星期二

  1. Microsoft AI Blog精选

    微软发布 MAI-Image-2.5 与 MAI-Image-2.5-Flash 图像模型

    微软发布 MAI-Image-2.5 和 MAI-Image-2.5-Flash 两款图像模型，前者在 Arena 图像编辑榜排名第 2、文生图排名第 3，超过 GPT-Image-1.5 和 Nano Banana Pro 2K。两款模型已在 Foundry 面向开发者开放，MAI-Image-2.5 定价为每 1M 文本输入 token 5 美元、每 1M 图像输入 token 8 美元、每 1M 图像输出 token 47 美元，Flash 版分别为 1.75、1.75 和 19.50 美元。MAI-Image-2.5 已上线 PowerPoint 并正在向 OneDrive 推送，OpenRouter 也将提供该模型。

    Bar chart showing MAI-Image-2.5 performance in editing tasks. Green bars indicate it wins most categories like image cleanup, backgrounds, shadows, and text, while competitor wins are fewer. Ties appear in some categories.

    Image Edit leaderboard, ahead of Nano Banana 2.1 Built for high-quality generation and precise, controllable editing, it brings production-ready image workflows to developers and Microsoft products. Today, we’re launching MAI-Image-2.5 for maximum fidelity, and MAI-Image-2.5-Flash for fast, scalable production workloads. Features and capabilities Step-change in text-to-image quality MAI-Image-2.5 produces more detailed, coherent images from prompts, with stronger text rendering, product imagery

    匹配字段正文

    #company#模型发布#图像生成#多模态
  2. Microsoft AI Blog

    微软与 Mayo Clinic 合作开发医疗专用前沿 AI 模型

    微软宣布与 Mayo Clinic 战略合作，共同开发一款专为医疗健康设计的前沿 AI 模型，结合 Mayo Clinic 的临床数据与微软的 AI、云和工程能力。该模型将首先在 Mayo Clinic 内部署，用于临床推理等场景，模型本身归 Mayo Clinic 所有，验证后向全球机构开放。微软称其消费产品每天处理数千万条健康问题，对 50 万条 Copilot 对话的分析显示用户主要用于症状解读、检验结果解释和就医导航。

    how we design user experiences that reduce anxiety rather than amplify it. Our research on how people are using our AI to support their health makes the opportunity clear. Across Microsoft AI consumer products we handle tens of millions of health questions every day, and a recent deep analysis of half a million Copilot conversations shows people rely on AI for symptom interpretation, test result explanation, and practical care navigation. That scale and intimacy make health a uniquely important

    匹配字段正文

    #company#行业动态#多模态#OpenAI
  3. Microsoft AI Blog

    微软 MAI 发布七款新模型，含首个推理模型 MAI-Thinking-1

    微软 MAI 在 Build 2026 主题演讲中发布七款新模型，覆盖图像、语音、转录、推理和编码。首个推理模型 MAI-Thinking-1 为 35B 激活参数 MoE，256K 上下文窗口，AIME 25 得分 97%，SWE Bench Pro 53%，官方称未做知识蒸馏。编码模型 MAI-Code-1-Flash 仅 5B 参数，SWE Bench Pro 51%，今日起在 VS Code 中逐步向约 10% 个人用户推送。

    d an AI that places humanity first. That always prioritizes human well-being and human progress. This is the core philosophy and motivation behind our superintelligence efforts at Microsoft. It shapes everything we do. And as a platform company, our job – and our commitment – is to keep you developers building at the absolute frontier. So today we’re very excited to announce a family of seven new models across image, voice, transcription, thinking, and coding. These are all built with real atten

    匹配字段正文

    #company#模型发布#多模态#语音
  4. Microsoft AI Blog

    微软 AI 发布七款新 MAI 模型，覆盖推理、编码、图像、语音与转录

    微软 AI 发布七款自研 MAI 模型，覆盖图像、语音、转录、编码与推理，包括旗舰推理模型 MAI-Thinking-1、面向 GitHub Copilot 的 MAI-Code-1-Flash、MAI-Image-2.5、MAI Transcribe-1.5 和 MAI-Voice-2。MAI-Thinking-1 在盲测人工对比中优于 Sonnet 4.6，MAI Transcribe-1.5 支持 43 种语言且速度是竞品的五倍，MAI-Image-2.5 的 Arena 分数超过 Nano Banana Pro。模型将在 Foundry、OpenRouter、Fireworks 和 Baseten 上提供，开发者首次可以自行微调模型权重；微软同时宣布与 Mayo Clinic 合作共建医疗前沿模型。

    LI; X; FB Updated as of June 8, 2026. Today we are announcing a family of seven new models developed in-house at Microsoft AI. Beyond these models, we’re building a superintelligence lab – a system and an approach we believe will define the next phase of AI. This is an extraordinary time in technology. The compute used to train frontier models has increased by a factor of one trillion. Now we expect another thousand-fold increase over the next three years, which in turn means more advanced capab

    匹配字段正文

    #company#模型发布#多模态#推理
  5. Microsoft AI Blog精选

    微软发布 MAI-Code-1-Flash 编码模型，上线 VS Code GitHub Copilot

    微软发布编码模型 MAI-Code-1-Flash，基于自有企业级数据从头训练，未使用第三方模型蒸馏，正面向 VS Code 中的 GitHub Copilot 个人用户推送，出现在模型选择器和默认 Auto 选择器中。该模型针对 GitHub Copilot 生产环境训练，支持智能体编码和自适应思考，官方称在 SWE-Bench Verified、SWE-Bench Pro、SWE-Bench Multilingual 和 Terminal Bench 2 上均超过 Claude Haiku 4.5，其中 SWE-Bench Pro 为 51.2% 对 35.2%，并在 SWE-Bench Verified 上以最多减少 60% token 解决更难问题。

    A scatter plot compares coding models on pass rate vs. average token usage. MAI-Code-1-Flash (green) outperforms Claude Haiku 4.5 (orange) across benchmarks, with higher pass rates and lower token use in the highlighted “Ideal Zone.”.

    LI; X; FB Updated as of June 8, 2026. Today we’re introducing MAI-Code-1-Flash, a new Microsoft coding model built for fast, efficient assistance in everyday developer workflows. We trained it from the ground up on clean, traceable and enterprise-grade data, without distillation from third-party models. The model is rolling out to GitHub Copilot individual users in Visual Studio Code in the model picker and under the default auto picker. Features and capabilities Agentic coding in real developer

    匹配字段正文

    #company#模型发布#编码#Agent
  6. The Register精选

    Anthropic 递交 IPO 申请，估值 9650 亿美元超过 OpenAI

    Anthropic 周一宣布已向美国 SEC 递交保密版 S-1 文件，为首次公开募股铺路，具体发行股数和价格尚未确定。此前该公司完成 650 亿美元 H 轮融资，投后估值达 9650 亿美元，超过 OpenAI 最近公布的 8520 亿美元估值。报道称 Anthropic 收入高于 OpenAI，但用户规模仅为后者一小部分，其真实财务状况要等 IPO 文件公开后才能评估。

    rgets 12.8 Tbps in a single optical pluggable TOP STORIES Anthropic asks users to stop being mean to Claude; '123456' password linked to Denmark data breach affecting 8.8M people; Microsoft and Anthropic play invoice tennis with startup's $17,600 Claude bill; Starlink's plan to avoid orbital near-misses: Ephemeris sharing; Shai-Hulud worm makes jump to AI infrastructure with Tensorlake compromise; Dutch tax office ditches Microsoft 365 cloud for on-premises alternative Cisco, Nvidia team up to b

    匹配字段正文

    #hot-signal#行业动态#OpenAI#Anthropic

星期一

  1. The Register

    Google 在 I/O 上大举推进 AI 化搜索，被批"AI 恶化

    Google 在 I/O 大会上宣布重塑 AI 时代搜索，扩大 AI Overviews 展示范围、推出 AI Mode 标签页，并在 AI 回答中插入广告。播客 The Kettle 讨论认为，这些举措把用户困在搜索结果页、模糊信息来源，可能动摇 Google 对网络的掌控。

    st went up, it just gained a zero and it's going to go up again and damn it, I can't afford to use this anymore. Thomas Claburn (25:00) Well, there was just a report recently that Microsoft canceled its internal cloud licenses because of the price hikes. I the price is going to go up and ⁓ the problems aren't necessarily going to be evident right away. Brandon (25:15) I've heard multiple instances of companies basically saying, well, these AI products are doing fine, but at the end of the day, t

    匹配字段正文

    #hot-signal#行业动态#搜索#Google

星期六

  1. DevClass

    Devographics 调查：63% Web 开发者用 AI 生成过半代码，近半数担心失业

    Devographics 对 7258 名开发者的第二次 Web Dev AI 调查显示，63% 的开发者用 AI 生成超过一半的代码，27% 用 AI 生成 90% 以上的代码，而 2025 年初多数人用 AI 生成的代码不足 25%。代码生成是最主要的用途，其次是代码审查、调研和调试。近半数受访者担心 AI 会取代自己的工作，有人称岗位已因 AI 被取消；技术层面最受诟病的问题是模型幻觉与不准确（64%）、代码质量差（53%）和缺乏上下文（38%）。模型使用率上 ChatGPT 以 88.4% 居首，Claude 为 82.1%，但付费订阅中 Claude 以 69% 领先。74% 的受访者认为 AI 工具已是工作流不可或缺的部分，64% 觉得效率因此提升。

    tly year on year. When you just need to build a tiny desktop app: TinyJS October 13, 2026 Valen creator drives 'Golden Spike' to connect new languages with Rust September 26, 2026 Microsoft agentically ports Copilot runtime to Rust for $120K September 19, 2026 Apple iPhone Duo makes developers think in folds September 16, 2026 JavaScript installer pnpm recast in Rust because ECMAScript can't keep up September 5, 2026 Kubernetes cleans house, bins legacy kube-dns, IPVS, and cgroup v1 August 27, 2

    匹配字段正文

    #hot-signal#现象/趋势#编码#数据/训练

星期四

  1. TLDR AI

    Google I/O 2026 将 Gemini 智能体嵌入工作流，AI 行业一周动态汇总

    Google I/O 2026 发布 Gemini Spark，将 AI 智能体嵌入 Gmail、Docs、Calendar 和 Drive，并推出 Daily Brief、Docs Live 和 Gemini 3.5 Flash。同期 OpenAI 为 Pro 用户推出个人理财功能，Anthropic 与 KPMG、PwC 达成合作并推出 Claude for Small Business，Meta 裁员约 8000 人并取消 6000 个岗位以投入 AI 基础设施。

    owork are part of the rollout. Anthropic also launched Claude for Small Business, with pre-built connectors for QuickBooks, PayPal, HubSpot, Canva, DocuSign, Google Workspace, and Microsoft 365. No IT team required. Why it matters: The Big Four are the consulting firms already inside every major corporation on the planet. When KPMG and PwC both go all-in on the same AI in the same week, that’s a distribution event, not a product announcement. Your consultants are probably already using Claude on

    匹配字段正文

    #newsletter#行业动态#Agent#Google

星期二

  1. Claude Blog精选

    Claude Managed Agents 新增自托管沙箱与 MCP 隧道

    Claude Managed Agents 现在可以在用户自控的沙箱中执行工具，并连接私有 MCP 服务器，沙箱与所访问服务都运行在企业边界内。自托管沙箱在 Claude Platform 上处于公开测试，支持 Cloudflare、Daytona、Modal、Vercel 等托管沙箱提供商；MCP tunnels 处于研究预览阶段，需申请访问，可在 Managed Agents 和 Messages API 中使用，由组织管理员在 Claude Console 的工作区设置中管理。智能体的编排、上下文管理和错误恢复仍运行在 Anthropic 基础设施上，工具执行则移到用户配置的环境中。

    rtifacts Artifacts Design Design Design Connectors Connectors Connectors Plugins Plugins Plugins Skills Skills Skills Claude in Chrome Claude in Chrome Claude in Chrome Claude for Microsoft 365 Claude for Microsoft 365 Claude for Microsoft 365 AI agents AI agents AI agents Code modernization Code modernization Code modernization Coding Coding Coding Commerce Commerce Commerce Customer support Customer support Customer support Cybersecurity Cybersecurity Cybersecurity Legal Legal Legal Sales Sale

    匹配字段正文

    #lab#产品更新#Agent#MCP/工具调用
  2. Claude Blog精选

    Claude Managed Agents 新增 dreaming、outcomes 与多智能体编排

    Anthropic 在 Claude Managed Agents 中推出研究预览版 dreaming，通过定期回顾历史会话与记忆库提取模式、整理记忆，让智能体跨会话自我改进，开发者可选择自动更新记忆或先审核再落地。同时 outcomes、多智能体编排和 webhooks 向开发者开放：outcomes 由独立 grader 在自身上下文窗口中按评分标准评估输出并让智能体自我修正，内部测试中任务成功率最高提升 10 个百分点，docx 与 pptx 文件生成任务成功率分别提升 8.4% 和 10.1%；多智能体编排让 lead agent 把任务拆给拥有各自模型、提示词和工具的专家子智能体并行处理，并在 Claude Console 中追踪每一步。Harvey、Netflix 平台团队、Every 的 Spiral 和 Wisedocs 已在使用这些能力。

    rtifacts Artifacts Design Design Design Connectors Connectors Connectors Plugins Plugins Plugins Skills Skills Skills Claude in Chrome Claude in Chrome Claude in Chrome Claude for Microsoft 365 Claude for Microsoft 365 Claude for Microsoft 365 AI agents AI agents AI agents Code modernization Code modernization Code modernization Coding Coding Coding Commerce Commerce Commerce Customer support Customer support Customer support Cybersecurity Cybersecurity Cybersecurity Legal Legal Legal Sales Sale

    匹配字段正文

    #lab#产品更新#Agent#MCP/工具调用

星期五

  1. DevClass精选

    Anthropic 的 Bun 完成 Rust 重写并合并入主仓库

    Anthropic 旗下 JavaScript 运行时 Bun 的 Rust 重写版本已合并入主仓库，新增超过一百万行代码，作者 Jared Sumner 称其在所有平台通过测试套件、修复部分内存泄漏并将二进制体积缩小 3 至 8 MB。Sumner 表示性能持平或更快，架构与数据结构不变，且团队已数月不自己写代码，主要依靠 Claude Code 维护。此前移除约 60 万行 Zig 代码的第二个 PR 被 GitHub 标记为 AI slop 并关闭。

    d sceptics alike. When you just need to build a tiny desktop app: TinyJS October 13, 2026 Valen creator drives 'Golden Spike' to connect new languages with Rust September 26, 2026 Microsoft agentically ports Copilot runtime to Rust for $120K September 19, 2026 Apple iPhone Duo makes developers think in folds September 16, 2026 JavaScript installer pnpm recast in Rust because ECMAScript can't keep up September 5, 2026 Kubernetes cleans house, bins legacy kube-dns, IPVS, and cgroup v1 August 27, 2

    匹配字段正文

    #hot-signal#开源/仓库#编码#Agent
  2. The Register

    安大略审计发现医生AI记录工具频繁出错

    加拿大安大略省审计长办公室对20家获批AI Scribe系统进行评估，发现9家系统编造了录音中未讨论的信息并给出治疗建议，12家写入错误药物信息，17家遗漏患者心理健康的关键细节。报告还指出评分权重失衡，是否在安大略本地设有机构占30%评分，而医疗记录准确性仅占4%，偏见控制、威胁与隐私评估各占2%。OntarioMD建议医生手动复核AI记录，但获批系统均无强制确认功能。

    rgets 12.8 Tbps in a single optical pluggable TOP STORIES Anthropic asks users to stop being mean to Claude; '123456' password linked to Denmark data breach affecting 8.8M people; Microsoft and Anthropic play invoice tennis with startup's $17,600 Claude bill; Starlink's plan to avoid orbital near-misses: Ephemeris sharing; Shai-Hulud worm makes jump to AI infrastructure with Tensorlake compromise; Dutch tax office ditches Microsoft 365 cloud for on-premises alternative Cisco, Nvidia team up to b

    匹配字段正文

    #hot-signal#安全/对齐#行业动态#政策/监管

星期二

  1. Vapi Blog

    Vapi 完成 5000 万美元 B 轮融资，语音 AI 智能体已处理超 10 亿通电话

    Vapi 宣布完成 5000 万美元 B 轮融资，由 Peak XV 领投，YC、M12、Kleiner Perkins 等参投，总融资额达 7200 万美元。过去一年其企业收入增长 10 倍，智能体累计处理超过 10 亿通电话。Amazon Ring 在两周内从评估转为将全部呼入电话交由 Vapi 运行，且 CSAT 有所提升。

    ions of dollars to build, staff, maintain, and apologize for. We started Vapi to end it. Today we're announcing our $50M Series B, led by Peak XV, with participation from YC, M12 (Microsoft's Venture Fund), Kleiner Perkins, and our earlier investors. It brings our total funding to $72M. The round closed on the back of a year in which our enterprise revenue grew 10x, and our agents crossed 1,000,000,000 calls handled. But this post isn't about the round. It's about the bet behind it. The spark In

    匹配字段正文

    #company#行业动态#语音#Agent

星期一

  1. DevClass

    C++ 年度调查：AI 编码使用率上升，但信任依然匮乏

    Standard C++ Foundation 年度开发者调查显示，39.8% 的 C++ 开发者经常用 AI 写代码，高于去年的 30.9%；用 AI 写测试的比例从 20% 升至 33%，调试从 11.5% 升至 23.6%。但 42% 的受访者仍很少或从不使用 AI，较去年的 52.7% 有所下降，主要顾虑包括输出错误、不信任输出、数据隐私和工具成本。本次调查共收到 1,434 份回复，比去年多 38%。

    lts and problems. When you just need to build a tiny desktop app: TinyJS October 13, 2026 Valen creator drives 'Golden Spike' to connect new languages with Rust September 26, 2026 Microsoft agentically ports Copilot runtime to Rust for $120K September 19, 2026 Apple iPhone Duo makes developers think in folds September 16, 2026 JavaScript installer pnpm recast in Rust because ECMAScript can't keep up September 5, 2026 Kubernetes cleans house, bins legacy kube-dns, IPVS, and cgroup v1 August 27, 2

    匹配字段正文

    #hot-signal#现象/趋势#编码#数据/训练