Pai 的 AI 雷达

今日必看

含近几日仍热话题,按时间衰减排序。今天先看 langfuse/langfuse,其次 liketrek/TREK。并行还有 Claude Code 推出实验性 Code Mode 工具调用方式、Microsoft and Salesforce Announce Copilotforce, Bringing Trusted Customer Context and Action into Microsoft AI Experiences。

日期
2026/11/18
今日条目
491
#01

Claude Code 推出实验性 Code Mode 工具调用方式

Claude Code 推出实验性 Code Mode,一种新的工具使用方式:Claude 会编写脚本以复杂方式串联多个工具,并能在长命令仍在运行时回复用户。可通过设置环境变量 CLAUDE_CODE_EXPERIMENTAL_CODE_MODE=1 开启,文档见 http://code.claude.com/docs/en/code-mode。 🔗 阅读原文 via AIHOT · https://aihot.news/items/z3n160k1v5vsc90udwjd8x7pp

推荐理由它可以用来让系统通过自主编写脚本以复杂方式串联并调度多个工具,同时还能在耗时较长的命令仍在后台执行时即时向用户发送回复,实现高效的工具链协作与异步沟通。
编码代理AIHOT feed.xml
#02

Microsoft and Salesforce Announce Copilotforce, Bringing Trusted Customer Context and Action into Microsoft AI Experiences

November 17, 2026 Category: AI Microsoft and Salesforce Announce Copilotforce, Bringing Trusted Customer Context and Action into Microsoft AI Experiences Microsoft Source | 8 minutes New integrations connect Salesforce with Microsoft Copilot, Teams and Fabric, helping enterprises

推荐理由Microsoft and Salesforce Announce:November 17, 2026 Category: AI Microsoft and Salesforce Announce Copilotforce, Bringing。
开发与开源Microsoft AI Source
#03

GitHub Copilot Fusion updates: more availability, benchmarks, and models

Today, we’re expanding GitHub Copilot Fusion with new models, available on more Copilot surfaces, and results on two public Terminal-Bench leaderboards, giving you more ways to use Fusion and compare its performance. GitHub Copilot Fusion joins public Terminal-Bench 4.1 leaderboa

推荐理由GitHub Copilot Fusion updates: more:Today, we’re expanding GitHub Copilot Fusion with new models, available on more Copilot。
模型与评测GitHub Changelog
#04

GitHub Copilot app, Microsoft Ignite releases

Copilot is most useful when it can work where you do. The GitHub Copilot app now supports your Azure DevOps repositories, your GitHub Enterprise Server instance, and the browser you’re already signed in to. Work with Azure DevOps repositories in the GitHub Copilot app Clone an Az

推荐理由该应用能够直接接入并协同管理 Azure DevOps 代码仓库与 GitHub Enterprise Server 实例,同时支持在已登录的浏览器环境中无缝衔接日常工作,让用户无需切换工作环境即可快速克隆并直接操作多平台代码仓库以辅助日常开发。
开发与开源GitHub Changelog
#05

Co-Designing Sparse Attention for Efficient Long-Context Inference

Agentic workloads, long documents, and large codebases are pushing context lengths from tens of thousands of tokens to hundreds of thousands. At those lengths, attention becomes the most expensive part of a transformer layer. Dense attention scores every query against every key,

推荐理由Co-Designing Sparse Attention:Agentic workloads, long documents, and large codebases are pushing context lengths from。
Agent与工具NVIDIA Developer GenAI
#10

Integrating QPUs into Supercomputing Centers Using Slurm HRES Plugins

Quantum processors are being integrated into supercomputing centers worldwide. National programs in North America, Europe, and Asia are driving the installation of quantum processing units (QPUs) alongside high-performance computing (HPC) systems. Every supercomputer will eventua

推荐理由该技术能利用 Slurm HRES 插件将量子处理单元(QPU)深度集成到超算中心的高性能计算(HPC)系统中,实现经典超算与量子处理器的协同调度与算力融合,推动传统超算设施演进为具备量子加速能力的新型异构计算平台。
开发与开源NVIDIA Developer GenAI
#11

PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models

arXiv:2611.16748v2 Announce Type: replace-cross Abstract: Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality ju

推荐理由它能利用大语言模型对物理常识的认知,为视频生成模型提供细粒度的物理反馈与评判监督,从而驱动视频生成模型实现自我迭代与优化,有效解决生成视频中物理动力学失真与因果逻辑不连贯的缺陷,创造出符合现实物理规律与因果一致性的逼真动态视频。
模型与评测arXiv cs.AI 官方
#12

Agent-Native by Design: A Programming Language for Agentic Applications

arXiv:2611.21298v1 Announce Type: cross Abstract: Building agentic applications requires combining large language model (LLM) operations, workflow structure and application state, and model-guided routing. With SDKs and agent frameworks, developers connect these components throug

推荐理由Agent-Native by Design: A Programming:arXiv:2611.21298v1 Announce Type: cross Abstract: Building agentic applications。
模型与评测arXiv cs.AI 官方
#13

白嫖双强模型 API!长亭 MonkeyCode 不止免费,更是下一代云端 AI 开发工作台

📌 一句话摘要 MonkeyCode 是长亭科技推出的开源 AI 开发平台,提供永久免费的 MiniMax-m2.7 和 Qwen3.5-plus 模型 API,同时深度集成 GitHub,实现浏览器端的完整云端开发流程。 📝 详细摘要 文章介绍了长亭科技推出的 MonkeyCode 平台,这是一款开源的 AI 开发工具。平台的核心亮点是提供两大主流大模型(MiniMax-m2.7、Qwen3.5-plus)的永久免费调用权限,解决了开发者频繁充值的问题。此外,MonkeyCode 深度绑定 GitHub,用户无需本地 Git 配置,即可在浏览器中完成需

推荐理由MonkeyCode 可用于永久免费调用 MiniMax-m2.7 和 Qwen3.5-plus 模型 API,解决频繁充值的问题,同时借助深度集成的 GitHub 生态,无需在本地配置 Git 即可在浏览器端直接跑通完整的云端开发全流程。
创作成片BestBlogs.dev
#14

Launch HN: StarSling (YC P25) – Self-improving CI for agent-volume repos

We’re Yonas and Daniel and we’re building StarSling ( https://starsling.dev ), fast GitHub Actions runners with AI agents that ship PRs to continuously improve your tests and workflows. Every CI job gets its own ephemeral sandbox on the fastest Zen 5 machines and NVMe Disk with u

推荐理由Launch HN: StarSling (YC P25) –:We’re Yonas and Daniel and we’re building StarSling ( ), fast GitHub Actions runners。
Agent与工具Hacker News
#15

Claude Code 团队验证方法解析:三大关键测试策略

📌 一句话摘要 作者转载 Addy Osmani 高质量文章,分享 Claude Code 团队的代码验证方法,特别推荐 E2E、Property 测试和 Differential 测试。 📝 详细摘要 作者认为 Addy Osmani 的文章质量极高,可作为爆款素材。重点分享《How the Claude Code team thinks about verification》中的内容:Claude Code 团队通过验证闭环实现 AI 写和审代码,作者认为最值得投入的三种测试是 E2E、Property 测试(首次听说)和 Differential

推荐理由📌 一句话摘要 作者转载 Addy Osmani 高质量文章,分享 Claude Code 团队的代码验证方法,特别推荐 E2E、Property 测试和 Differential 测试。
编码代理BestBlogs.dev
#19

Google’s Spark AI agent becomes ‘Tasks’ — and now it’s free

You can now switch from the “Chat” toggle to “Tasks.” | Image: Google Google is rebranding its Spark AI agent as "Tasks" and making it free to use. The feature, which is powered by Google's new Gemini 4 Hydrogen AI model, can do things like create meal plans, manage your calendar

推荐理由这款由谷歌全新 Gemini 4 Hydrogen 模型驱动且现已免费开放的智能代理工具,支持直接从聊天切换开启,能够协助用户制定日常膳食与饮食计划,并高效整理与管理个人的日程日历。
Agent与工具The Verge AI
#20

让 AI 写得像自己:Every 评测负责人试了六年,最后靠一个"盲审裁判"做到了

让 Claude 读完自己的文章,写一份风格指南,再拿这份指南去写稿。很多人都这么做过。 Mike Taylor 的实验结果很扎心: 这种 Claude 写的风格指南,对上三种优化方法时,在 80% 到 95% 的测试文章上都输了。 换成 Opus 5.5 时更离谱,不加这份指南反而写得更像他本人。 Mike Taylor 是 Every 的评测负责人,也是 O'Reilly《Prompt Engineering for Generative AI》的合著者。他从 2020 年就开始折腾这件事,这篇文章是他六年试错后总结的完整流程。 骗过了所有朋友,骗不

推荐理由通过这套历经六年试错总结的完整流程与盲审裁判方法,能帮你避开简单生成风格指南的低效陷阱,建立起可靠的客观评测标准,真正调教并验证 AI 稳定写出高度契合个人文风的高质量内容。
创作成片乔木博客
#21

New Ask Maps features to help with holiday travel and prep

Plan holiday road trips, get route tips, find local appointments, and search your saved lists with new Ask Maps features.

推荐理由借助这些全新的地图问答功能,你可以轻松规划节日假期的自驾游路线并实时获取沿途出行建议,还能快捷查找周边的各类本地预约服务,以及随时检索自己先前收藏保存的地点清单,全面搞定假期的出行筹备与行程安排。
产品接入Google AI
#24

Cloudflare Joins the Jevolution and Labels Itself a Neolab

Cloudflare has been one of the essential infrastructure companies of the web for the past decade and a half. But now, it’s turning into a neolab — at least according to a presentation at AI Engineer Code Summit by Michelle Chen , who runs Cloudflare’s AI Platform on the product s

推荐理由Cloudflare Joins the Jevolution:Cloudflare has been one of the essential infrastructure companies of the web for。
创作成片Latent Space
#27

[AINews] Perplexity hits a $1B revenue run rate, up 4x in 2026

Perplexity crossed a $1B annual revenue run rate today , CEO Aravind Srinivas told Semafor at a live event. The Information reported previously under $250M at the start of the year and over $750M by August, with Computer driving some of the growth. At usual agent lab multiples (e

推荐理由该资讯可用于评估人工智能搜索与智能体领域的商业化变现速度,并在分析AI初创企业估值倍数、研判行业增长趋势或撰写投融资报告时作为核心的实证数据与参考基准。
Agent与工具Latent Space
#28

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

推荐理由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。
开发与开源硬件边缘少数派
#30

From AI ambition to business impact

Most business leaders I meet no longer need to be convinced that AI works. They’ve seen the pilots, the demos and the copilots. The question I hear in boardrooms now is different: How do we move from AI that impresses to AI that changes how the business actually runs? That is the

推荐理由From AI ambition to business impact:Most business leaders I meet no longer need to be convinced that AI works。
开发与开源Microsoft AI Blog