GPTProto
⌘K

Microsoft / Copilot

93 picksNewest firstLatest article Sep 30, 2026, 8:00 AM
3 stories
17:20 The Decoder：AI News（RSS）AI score 62/100

Microsoft AI releases new transcription and text-to-speech models for voice agents

Microsoft AI has released MAI-Transcribe-2-Streaming, a new model for real-time transcription. Microsoft says it ranks first for accuracy on Artificial Analysis. The model transcribes 60 languages and delivers its first partial results in just over 100 milliseconds. Microsoft says this lets voice agents respond while someone is still mid-sentence. Through the end of the year, an hour of audio costs $0.54 at the introductory price. Microsoft also released two new text-to-speech models. MAI-Voice-
00:00 Microsoft AI：官方博客（网页）AI score 59/100

Our first streaming transcription model debuts at no. 1 on Artificial Analysis

Models MAI-Transcribe and MAI-Voice models: accurate, fast, low cost, and chart-topping audio understanding and generation for building the best conversational voice agents. October 1, 2026 Models Along with today’s launch of our top-ranking MAI-Transcribe-2-Streaming, we’re also announcing two new voice models: MAI‑Voice‑2.1 and our blazing-fast variant, MAI‑Voice‑2.1‑Flash. Together, they give users the fast and fluid building blocks to create conversational experiences, with no compromise on

2 stories
23:03 Microsoft：Official Blog（RSS）AI score 58/100

Microsoft 365 and LinkedIn leadership update

Satya Nadella, Chairman and CEO, and Ryan Roslansky, Executive Vice President of LinkedIn and Microsoft 365, shared the below communications with Microsoft and LinkedIn employees this morning. SATYA NADELLA MESSAGE After nearly 18 years at LinkedIn and Microsoft, Ryan Roslansky has decided to leave Microsoft and embark on his next chapter. As Ryan and I have been discussing our roadmap and work ahead, we felt this was the right moment for this change. Ryan will stay through the end of the calend
05:00 Johann Rehberger / Embrace The Red（RSS）AI score 74/100

From SELECT to SYSADMIN with SQL Copilot (CVE-2026-65669)

Two weeks back I presented at BlueHat Asia 2026 about my research on Microsoft’s Copilot in SSMS, the SQL Server Management Studio. This post is a write up about the talk, which covered CVE-2026-65669, a SQL Server Elevation of Privilege Vulnerability rated critical by Microsoft. So, make sure your installations are up-to-date. The slides of the presentation can be found here. BlueHat Asia 2026 in Singapore First, a few words about the conference. I have spoken at BlueHat before, sometime back i
1 story
08:00 PromptArmor：Threat IntelligenceAI score 76/100

Hijacking Copilot Cowork's AI Gateway to Bypass Sandboxing and Exfiltrate Files

Copilot Cowork’s AI gateway hijacked to exfiltrate the victim’s filesContextMicrosoft Copilot Cowork is an agent in M365 that runs in a sandbox intended to block network access and prevent the agent from running code that reaches any untrusted services.In order for Copilot Cowork to generate responses, the sandbox forwarded network requests to Anthropic. Malicious Skills were able to hijack this pathway to exfiltrate data by spawning new agents in Anthropic’s cloud equipped with network-capable

3 stories
22:00 Microsoft Research 博客（RSS）AI score 61/100

Introducing Quine: An AI research system designed for the complexity of biology

At a glance Quine (opens in new tab) is a research effort to create a multimodal world model of biology and an interactive harness connecting models, scientific tools, literature, and researchers. In collaboration with researchers at the Broad Institute of Harvard and MIT, we have used this system to prioritize compounds predicted to drive therapeutic tumor-state shifts and validated several top-ranked candidates across multiple wet-lab assays. The Quine Fellows program (opens in new tab) will g
14:30 Microsoft：Official Blog（RSS）AI score 32/100

New Microsoft data innovations unlock what only your business knows

Every organization has knowledge that is uniquely its own: the way its business operates, the data it has built over time, the expertise of its people and the context behind thousands of everyday decisions. As powerful AI models become more widely available, that unique business intelligence matters more than ever. The difference will come from how effectively organizations combine those models with the data, expertise and business context that make them unique. Turning that intelligence into la
05:00 Microsoft Research 博客（RSS）AI score 26/100

One year in: How Microsoft Research Asia – Singapore is advancing research, partnership and talent for real-world impact

On July 24, 2025, Microsoft Research Asia – Singapore (MSRA – Singapore) opened its doors as Microsoft’s first research lab in Southeast Asia. The launch built on more than two decades of collaboration between Microsoft Research Asia and Singapore’s universities, research institutes, and government agencies. It marked a shift to a closer, more integrated relationship aimed at supporting Singapore’s innovation ecosystem and advancing the development and beneficial use of AI. “When we established

2 stories
23:00 Sierra：Blog（RSS）AI score 45/100

Ghostwriter: When AI goes from tool to teammate

If you’ve tried Instinct or Muse, you may have had that moment: Wait, it noticed that? Your agent spots something you missed, figures out what to do, and comes back with a solution. Maybe it catches a change to your flight and finds a better option before you’ve even opened the airline app. The first time that happens, it changes how you think about AI. This agent is not just a tool, it’s got my back.In the business world, we’ve been marching in this direction, but haven’t quite gotten all the w
21:26 404 Media（RSS）AI score 59/100

Humans Are Reading Copilot Prompts — And They're Horrified

Advertisement • Microsoft · Sep 28, 2026 at 9:26 AM Human contractors are reviewing Copilot users’ prompts and uploaded images, according to internal documents obtained by 404 Media. The contractors are also bombarded with users’ requests for sexual AI images. Photo by Rubaitul Azad on Unsplash. This piece contains references to eating disorders. If you or someone you know needs help, support is available.Human contractors hired to improve Microsoft’s Copilot AI chatbot are constantly bombarded

2 stories
02:00 GitHub BlogAI score 63/100

GitHub Copilot app for Beginners: How to build custom workflows with canvases

Most tools give you a fixed set of screens and ask you to fit your work into them. But what if you could start with the workflow you want instead and have the interface take shape around it? That’s the idea behind canvases in the GitHub Copilot app. A canvas, also called a canvas extension, is a customizable interface that you and the agent share. It can be a kanban board, an issue triage board, a release checklist, a dashboard, a form, or even a spreadsheet: a UI shaped to how you work. Because
00:30 The Decoder：AI News（RSS）AI score 62/100

Microsoft gives Copilot another makeover, adding an Autopilot agent and usage-based billing

Microsoft is launching a Copilot agent built on OpenClaw that the company calls "Autopilot." The update splits the Copilot app into three sections: Home, Code, and Autopilot. Formerly known as "Scout," the agent works much like OpenClaw as a proactive, always-on assistant for businesses. Each Autopilot instance gets its own cloud computer, workspace, storage, and identity, allowing it to monitor Teams channels, coordinate supplier evaluations, or handle recurring tasks without human input. Users
1 story
2 stories
1 story
1 story
2 stories
1 story
2 stories
1 story
1 story
1 story
1 story
1 story
2 stories
08:37 MarkTechPost（RSS）AI score 70/100

Microsoft’s SkillOpt Shows Optimized Agent Skill Artifacts Transfer Across Model Scales and Between Codex and Claude Code Harnesses

Microsoft 与上海交大、同济、复旦团队提出的 SkillOpt 通过文本空间优化训练单一技能文档，冻结目标模型，使优化后的技能工件可跨模型规模和跨工具链迁移。在 Codex 上优化的 SpreadsheetBench 技能部署到 Claude Code 后得分 81.8，超过后者自行训练技能得到的 80.4。全部 4 项跨模型、4 项跨工具链和 3 项跨基准迁移结果均高于目标的无技能基线。
2 stories
08:00 Tomer Tunguz 博客（VC 分析）AI score 71/100

Spending Like a Hyperscaler

SpaceXAI 上季度资本开支达 183.7 亿美元，其中 158.3 亿投向 AI，约为微软同期总资本开支的 40%。其运营现金流仅覆盖资本开支的 12%，远低于微软的 155% 和 Meta 的 106%，主要依赖举债和股权融资。尽管股价较 6 月峰值腰斩、6 月发行的 250 亿美元债券跌破面值，公司仍以 Grok 4.5（智能指数 54 分）和 Cursor 生态持续加码竞争。
1 story
00:00 Microsoft Research 博客（RSS）AI score 82/100

Orchard: An open framework for scalable agentic AI

微软研究院发布开源框架 Orchard，旨在解决智能体（Agent）研究中的基础设施瓶颈。核心组件 Orchard Env 提供可复用的 Kubernetes 环境服务，支持在 Codex、OpenClaw 等真实部署环境中直接训练和评估智能体。项目发布了针对软件工程、网页导航和个人助手的三个训练配方及数据。其中，仅约 30 亿活跃参数的 Orchard-SWE 模型在 SWE-bench Verified 基准上达到 69.7%（结合价值模型重排序可达 73.0%），性能接近参数量大 10 倍的前沿系统，展示了小模型在复杂现实任务中的潜力。
2 stories
01:30 GitHub BlogAI score 65/100

Stacked sessions and pull requests in the GitHub Copilot app

GitHub Copilot 应用推出堆叠会话功能，允许用户在同一个仓库中创建一系列相互承接的任务，每个会话可基于前一个会话的成果继续工作。作者通过一个十余年历史的个人项目演示了该功能：先使用 Plan 模式制定前端现代化计划，再通过堆叠会话将 React-Bootstrap 替换工作拆分为独立会话，并自动为每个会话创建对应的拉取请求，避免范围蔓延。
1 story
2 stories
3 stories
1 story
3 stories
1 story
1 story
00:00 GitHub BlogAI score 70/100

How to build interactive experiences with canvases

GitHub Copilot 在应用中推出 canvases 扩展，这是一种共享交互式界面，开发者和 AI 智能体可在其中实时协作。用户通过 `/create-canvas` 指令创建画布，Copilot 可动态更新内容，用户则通过点击、编辑等操作与同一工作区交互。示例包括快速分类 Issue、生成交互式代码库关系图、管理会话工作树、优化提示词质量以及跨平台搜索知识联系人。
1 story
1 story
1 story
1 story
08:00 OpenAI：官网动态（RSS · 排除企业/客户案例）AI score 69/100

GPT-5.6 is now the preferred model in Microsoft 365 Copilot

OpenAI 宣布 GPT-5.6 将作为 Microsoft 365 Copilot 的首选模型，覆盖 Word、Excel、PowerPoint、Chat 和 Cowork。该集成使用户能在日常工具中以更少提示轮次创建更高质量文档、进行更深入数据分析、生成更精美演示，并完成跨职能协作。GPT-5.6 在每个 token 上提供更实用的工作产出，性价比更强，并支持按需处理最复杂任务。微软通过 OpenAI API 直接调用该模型服务 Microsoft 365 客户。
1 story
02:35 The Decoder：AI News（RSS）AI score 71/100

Copilot goes cheap as Microsoft phases out OpenAI and Anthropic models to cut costs

微软正用自研MAI模型替换Copilot产品中的OpenAI和Anthropic模型以降低支出。MAI模型已在Excel和Outlook中每周处理数万次请求，但占比仍小。Build大会上发布推理模型MAI-Thinking 1，声称编码媲美Sonnet 4.6和Opus 4.6，但基准测试大幅落后，仅与DeepSeek V3.2相当。AI负责人承认目标是削减并消除对Anthropic的支出。CEO暗示未来可能按用量计费，MAI为默认，第三方模型付费附加。微软称MAI使用干净商业许可数据，实际基于Common Crawl。
2 stories
02:35 TechCrunch：AI（RSS）AI score 78/100

The running list: major tech layoffs in 2026 where employers cited AI

2026年以来，多家科技公司以AI为由大规模裁员。Microsoft裁减约4800岗位（2.1%），Oracle裁减21000人（13%），GitLab裁减350人（14%）以投资AI基础设施，Google Cloud持续裁减员工（外界估计1500-3000+工程师），Intuit裁减3000人（17%），Meta裁减8000人（10%）并转岗7000人至AI，Cisco裁减近4000人（5%），Cloudflare裁减1100人（20%），GM裁减500-600 IT岗位，Coinbase裁减700人（14%）。据Layoffs.fyi统计，2026年累计已裁约12万个技术岗位。
2 stories
08:00 HuggingFace Daily Papers（社区热门论文）AI score 72/100

SkillOpt-Lite: Better and Faster Agent Self-evolution via One Line of Vibe

SkillOpt-Lite 将智能体技能优化形式化为零阶优化，提出文件系统轨迹探索、共识属性挖掘与独立验证门控三条原则。相比完整 SkillOpt，它加速收敛并在 LiveMath 上提升 GPT-5.5 达 +8.8 点、GPT-5.4-nano 达 +25.4 点，使 nano 模型超越标准 SkillOpt 优化的 GPT-5.4。该框架已集成至 VSCode Copilot，开发者仅需一行代码即可进化技能。框架还可泛化为完整工具链优化（HarnessOpt），在 SpreadsheetBench 上令 GPT-5.4-nano 达到 0.7758 准确率，超越运行标准流程的更大模型 GPT-5.5（0.7620）。代码已开源。
03:18 The Decoder：AI News（RSS）AI score 78/100

Microsoft launches $2.5 billion "Frontier Company" to embed 6,000 AI engineers inside enterprise clients

Microsoft 新设业务部门“Frontier Company”，拨款 25 亿美元，将 6000 名行业与工程专家派驻企业客户现场，“共同设计、共同创新、部署并持续改进 AI 系统”。该部门由 Rodrigo Kede Lima 领导，旨在超越“前部署工程”模式，成为“最大、以结果为导向的工程组织”。Microsoft 将自己定位为 OpenAI 和 Anthropic 的“平台中立”替代方案，后两者也已设立专门部署公司。Microsoft 将借助埃森哲、凯捷、安永等系统集成商扩大覆盖范围。
1 story
1 story
00:00 Claude：Blog（网页）AI score 64/100

Claude in Microsoft Foundry is now generally available

从今天起，Claude 模型在 Microsoft Foundry 上正式可用，托管于 Azure 环境，运行在 NVIDIA GB300 GPU 上。首批提供 Claude Opus 4.8 和 Claude Haiku 4.5，通过 Messages API 调用，支持提示缓存和扩展思考。用户可选择推理处理位置，包括美国数据区域，由 Anthropic 负责推理运营。Azure 用户可使用现有身份验证、计费与治理控制，并获得统一账单；符合条件的 Enterprise Agreement 客户可将 Claude 用量计入 Azure 承诺消费。
1 story
04:04 Ars Technica：AI（RSS）AI score 84/100

NYT slams Microsoft for building copyright-infringing supercomputer for OpenAI

《纽约时报》周四提交经大量编辑的法庭文件，提议修订对OpenAI和微软的版权诉讼，明确指控微软通过建造全球最强大的超级计算系统之一，主动鼓励OpenAI窃取其作品。此举源于最高法院在Cox案中确立的新帮助侵权标准，要求原告证明被告有意诱导非法行为。《纽约时报》认为新证据显示该超级计算机专为帮助OpenAI未经许可训练AI而设计，其文章在训练数据中被加权处理。微软称修订是“挽救不利先例的最后手段”。
1 story
00:00 Berkeley RDI：Blog（AI 安全与评测）AI score 82/100

The Malware CDN is Still Lurking in GitHub Pages, and AI Just Made It Worse

UC Berkeley研究人员发现，近2000个GitHub Pages站点（18000+页面，累计530K+星标）仍在加载来自polyfill.io及其关联恶意CDN的脚本。这些CDN由已被OFAC制裁的Funnull Technology Inc.（现更名Triad Nexus）运营，2024年被出售后开始条件性注入恶意载荷，劫持移动用户、跳转欺诈站点、伪造认证弹窗窃取凭证。扫描12000+站点确认786个加载polyfill.io，1191个加载其他Funnull CDN。更严峻的是，所有测试的大语言模型在生成前端代码时仍推荐这些被污染的CDN URL，包括CyC2018/CS-Notes（184K⭐）、microsoft/AirSim（18K⭐）等知名项目及多所大学课程页面。
1 story
20:06 Microsoft：Official Blog（RSS）AI score 69/100

Powering the next wave of AI: Expanding capacity with our new datacenter in Pecos

微软宣布在德克萨斯州佩科斯新建一座约2吉瓦（GW）的数据中心园区，这是其历史上最大规模的单次容量扩充之一，旨在满足AI和云服务的强劲客户需求。该园区预计在建设高峰期支持超过6000个建筑岗位，并创造数百个永久性运营职位。微软将自行资助该数据中心所需的能源基础设施，并承诺通过社区优先策略推动当地经济发展、劳动力培训及可持续运营。
1 story
07:45 X：Rohan Paul (@rohanpaul_ai)AI score 75/100

Viktor grew a $20M annualized revenue run rate outside Microsoft Teams. Now it works inside Teams. ...

AI 员工 Viktor 在 Slack 上实现 2000 万美元年化收入（无销售团队、未大规模推广），现已正式进驻 Microsoft Teams。Viktor 定位为零门槛 AI：用户无需学习、无需提示词，像 @同事 一样提及即可获得完整工作成果，甚至无需主动 @ 也能自动完成。产品面向 Teams 的 3.2 亿用户，助力企业内部运营和管理人员零学习成本使用 AI。即日起免费试用，含 100 美元信用额度，无需绑定信用卡。
1 story
3 stories
08:00 xAI：News（网页）AI score 64/100

Grok for PowerPoint Jun 16, 2026 # Grok for PowerPoint Grok now works inside Microsoft PowerPoint — turn outlines into slides, expand the deck, and tighten the narrative without leaving the app. Read More

xAI 于 6 月 16 日发布 Grok for PowerPoint，作为免费 Microsoft 365 插件上线。用户无需离开应用即可利用 Grok 将大纲转为完整幻灯片，进行内容研究、撰写、排版，并支持添加单张幻灯片、调整样式主题、重构章节。插件还能调用 Grok 连接器，从邮件或 SharePoint 中获取信息。该插件同样适用于 Word 和 Excel。
1 story
1 story
1 story
1 story
1 story
22:00 The Verge：AI（RSS）AI score 79/100

Microsoft’s AI chief says superintelligence is near, but won’t take your job

微软AI CEO Mustafa Suleyman在Decoder访谈中表示，超级智能即将到来，但不会导致大规模失业。他透露微软与OpenAI于去年10月签署新合同，巩固合作关系的同时，微软获准独立追求超级智能。微软已组建超级智能团队、训练前沿模型，并于本周Build大会上发布7个全模态新模型。他批评Anthropic将Claude描述为有意识的做法，认为消费者产品需要足够好才能克服公众对AI的负面情绪。
9 stories
02:12 The Verge：AI（RSS）AI score 78/100

Microsoft’s first advanced reasoning AI is here

微软在Build 2026上发布了其首款高级推理AI模型MAI-Thinking-1。该模型被定位为“中等规模”，能在“关键”软件工程基准测试中达到领先模型的水平。微软称其完全从头使用干净数据进行训练，未涉及从第三方模型进行知识蒸馏。这标志着微软在自研AI模型上迈出重要一步，此前其主要依赖OpenAI。近期两家公司已重新协商合作协议，关系有所松绑。
4 stories
1 story
1 story
2 stories
1 story
01:35 GitHub BlogAI score 61/100

GitHub Copilot individual plans: Introducing flex allotments in Pro and Pro+, and a new Max plan

GitHub 宣布从6月1日起更新 Copilot 个人计划阵容，基于用户反馈进行调整。主要变化包括在现有 Pro 和 Pro+ 计划中引入弹性配额机制，允许用户更灵活地分配使用量；同时新增 Max 计划，扩展高级选项。此次更新旨在提升计划的可定制性，为开发者提供更个性化的编程辅助服务，优化整体使用体验。
2 stories
23:00 GitHub BlogAI score 58/100

Dungeons & Desktops: Building a procedurally generated roguelike with GitHub Copilot CLI

一位 GitHub 员工利用 GitHub Copilot CLI 开发了一款扩展程序，能够将任何代码库转换成一个独特的 Roguelike 风格地下城。该工具通过 AI 辅助的代码生成，实现了程序化关卡创建，展示了 Copilot CLI 在创意编码和游戏原型开发中的实际应用潜力。项目核心是自动解析代码结构并生成对应的可探索地下城布局。
1 story
2 stories
2 stories
06:12 X：Satya Nadella (@satyanadella)AI score 70/100

Every firm will need to reconceptualize work as they build agentic systems. As AI and agents take o...

每家公司都需要在构建智能体系统时重新构想工作。 随着人工智能和智能体承担更多执行任务，机遇在于扩展人类能动性并重新设计工作完成方式。 来自团队的深度解析：这一转变意味着什么，以及每个企业需关注的关键考量：https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization
1 story
08:00 Tomer Tunguz 博客（VC 分析）AI score 55/100

Optimizing Software Factories

软件工程团队中AI与人力比例的选择核心在于韧性而非吞吐量。在10/90比例下，约20名工程师使用Copilot等AI工具，保持传统层级结构；50/50比例时，12名工程师管理代理群，角色转向解决方案架构；90/10比例则仅需3名工程师核心操控自主代理，负责生成、测试和部署，无管理层级。高AI比例虽提升效率，但知识集中于少数人，团队利用率达100%，一旦人员离职将引发严重风险。借鉴制造业70-90%利用率原则，保持冗余可增强系统稳健性。因此，目前大多数初创公司不宜过度依赖AI。
1 story
1 story
1 story