Pai 的 AI 雷达
含近几日仍热话题，按时间衰减排序。今天先看 kharmanskyi/open-steps，其次 Upcoming deprecation of selected GitHub Copilot models in December。并行还有 Advancing science & engineering with Antigravity’s /teamwork、Run Native DLA Inference in Python on NVIDIA Jetson AGX Orin Using cuDLA。
We will deprecate the following models across all GitHub Copilot experiences (including Copilot Chat, inline edits, ask and agent modes, and code completions) on December 14, 2026: Model Deprecation date Suggested alternative Claude Fable 5 2026-12-14 Claude Fable 5.2 Claude Haik
Education Innovation
AI applications at the edge often need to run multiple neural networks within limited compute and power budgets. NVIDIA Deep Learning Accelerator (DLA) platforms provide dedicated hardware for deep learning inference. On NVIDIA Jetson AGX Orin, DLA executes supported neural netwo
You can now bring your work on GitHub Enterprise Server (GHES) into the GitHub Copilot app. Add your GHES account to work with remote repositories, issues, and pull requests. This feature was previously an opt-in preview, and is now generally available. You’ll need to sign in wit
Introducing ChatGPT Wallet, a secure way for agents in ChatGPT Work, Codex, and dots to complete purchases with user approval and existing card protections.
Anthropic 的 Claude Duet 更新：截至昨日在等待名单上的全部美国 Pro、Max 和 Ultra 订阅者均已收到邀请，等待名单入口为 https://claude.com/product/claude-duet。研究预览附带限额重置，Pro 为 3 次，Max 和 Ultra 为 1 次；官方计划很快向美国以外更多国家开放，并继续征集反馈。 🔗 阅读原文 via AIHOT · https://aihot.news/items/ajfypca37l5knyrnij49v5ofz
Running a frontier-scale model locally means keeping all its weights in GPU memory; the largest models no longer fit on one machine. Two NVIDIA DGX Stations connected over ConnectX-8 RDMA serve one model through a single endpoint, with tensor parallelism splitting the model weigh
Human motion offers roboticists a great starting point for teaching humanoid robots how to move. A human demonstration captures the coordination and timing behind natural movement, from everyday actions to complex tasks. But a robot cannot simply copy the same joint motion from a
Recommendations for designing, running, and reporting safety evaluations, based on our own experiences.
Built primarily on OpenAI models, Amplitude’s proactive product agent, Wave, identifies what to build, helps teams ship changes with coding agents, and measures the impact.
Always-on agents run indefinitely, deciding how and when to act across many sessions. This guide walks through their core primitives and how they fit together.
We are bringing Gemini 4 Hydrogen to Google Search globally to provide you with more helpful, visual, and intuitive responses.
📌 一句话摘要 前 OpenAI 研究员声称通过模型权重指纹技术检验到中国 AI 实验室使用了禁运 NVIDIA 芯片。 📝 详细摘要 该推文是 Depue 关于中国模型是否使用禁运 Blackwell 芯片的首发声明。他声称通过分析模型权重初始化时的随机数布局模式（即「指纹」），检验出超过 6 万亿种组合，匹配到 DeepSeek、月之暗面、MiniMax 和蚂蚁的模型在 Blackwell GB200/GB300 上初始化。他进一步估算涉及数千块芯片、数亿美元成本。声明称结论经 OpenAI 和 Anthropic 的「独立验证」，并附带报告链接供查
📌 一句话摘要 前 OpenAI 研究员声称中国 AI 实验室使用禁运 Blackwell 芯片，但报告承认的限制在推文中被简化，引发技术社区质疑。 📝 详细摘要 前 OpenAI 研究员 Will Depue 宣称发现中国模型（DeepSeek、月之暗面、MiniMax、蚂蚁）使用被禁运出口的 NVIDIA Blackwell 芯片，涉价 5 亿美元。然推文中声称的结论多数不在原报告中，报告本身承认关键限制：指纹仅证明初始化芯片，无法说明实际训练情况、芯片所在地及是否违规。推文还简化了「5 亿美元」估算依据，省略了假设性质。批评者指出，Depue 在推
Claude Code, Codex & Cursor limits, paced in your menu bar Discussion | Link
Former OpenAI researcher Will DePue says he has produced the first technical evidence that DeepSeek, Moonshot AI, MiniMax, and Ant Group are using Nvidia's latest Blackwell GPUs despite US export controls. His forensic method reads characteristic chip signatures directly from pub
Create product videos and courses from any AI chat Discussion | Link
Congrats to Gemini 4 Hydrogen on their launch and it seems Anthropic’s IPO is well under way, but today’s headline was only ever going to be about one thing, the most earthshaking leadership change in AI (after Demis/Jeff/et al at Google DeepMind). Yesterday we covered reports th
The theme of day two of the AI Engineer Code Summit in San Francisco was increasing ambition using AI . And there have been few projects as ambitious as ArtCraft in the AI era. So far the fledgling company has released 17 open source versions of popular software tools , starting
PLUS: SoftBank backs $8.5B German humanoid
苹果支持社区将于 11 月 20 日永久关闭 再曝婚外情，曝英伟达董事会考虑罢免黄仁勋 余承东官宣华为子品牌 OSTA 星芒，11 月 20 日首秀 曝字节计划推出「中国版 Muse」 小米澎程首个完整交付月，两款车零售合计超 2.7 万辆 张雪机车第 10 万台整车在重庆下线 腾讯入局搜索，网页端元宝搜索上线 曝 Anthropic 承销商考虑上调 IPO 发行价 我国算力总规模达 310 万 PFLOPS，全国数标委成立具身智能数据标准起草组 字节 AI 芯片负责人被曝提出离职，团队正研发下一代芯片 北京印发具身智能机器人三年行动方案，提出上市一对一
Pinecone integrated embedding now supports Google's Gemini Embedding 2
Research: parse_number() as a SQLite extension: C vs Rust I was out walking the dog when I got curious about how easy/difficult it would be to write a SQLite C extension that adds a parse_number() function for turning strings like -$1,234.56 or '12.5%' into numeric values that co