AI News

TestingCatalog 11月9日 AI 日报:OpenAI、Anthropic、Google 动态

Testing Catalog更新于 11月9日 17:46

DAILY AI BRIEF 🗞 — Nov 9

AI 导读

TestingCatalog 11月9日 AI 日报汇总多家公司动态:OpenAI 方面,Dots 可从关联邮箱读取登录验证码并自动填入,1M 上下文的隐身模型 Wyvern 在 Devin 中被发现;Anthropic 据 FT 报道拿下约 34 亿美元燃气发电机订单,对应 1.7GW 数据中心电力;Google 正为 Gemini Enterprise 测试下一代 Gemini 智能体,早期 checkpoint Krypton-a 在编码演示中据称胜过 Carbon 和 Argon。

引用

🚨 AI News | TestingCatalog

@testingcatalog

每日 AI 简报 🗞 — 11月7日 OPENAI 🔥: - 彭博社报道,OpenAI 在一轮融资中迄今已筹集 110 亿美元,公司估值达 1.25 万亿美元。 - ChatGPT Remix 正向 Business 和 Enterprise 推出,让团队可以复刻同事的站点并在聊天中继续编辑。 ANTHROPIC 🔥: - Pro、Max 和 Ultra 上的 Claude Code 云会话现在能自行处理依赖项、网络访问和 API 密钥。 - Claude Duet 可以查看正在运行的 Claude Code 会话、批准更改,并从手机上发送更新。 - Claude for iOS 正在测试侧边导航改版,将 Chats、Projects 和 Code 分为独立标签页。 META 🔥: - Muse Spaces 让用户从聊天中生成可交互、可更新的页面,并将其固定在应用中。 - Muse 和 Meta AI 现在运行带有 Muse Wonder 的 Active Memory 和 Active Focus 的专用模型。 SPACEXSI 🔥: - Grok API 文本水印功能今日起对列出的 Grok 4.x 模型默认开启,团队可在 Console 中选择退出。 ALIBABA 🔥: - Qwen 发布了“The becoming begins”预告,Qwen4-Max、Flash、Plus 和 27B 预计很快推出。 NOUS RESEARCH 🔥: - Nous Research 发布了 Arete-27B,这是一个面向西方价值观的 Qwen3.8-27B 后训练模型。 * 使用 Grok 撰写本简报,挑选新闻并进行了一些后期编辑。

查看原文

DAILY AI BRIEF 🗞 — Nov 7 OPENAI 🔥: - Bloomberg reports OpenAI has raised $11B so far in a round valuing the company at $1.25T. - ChatGPT Remix is rolling out to Business and Enterprise so teams can fork a colleague’s site and keep editing it in chat. ANTHROPIC 🔥: - Claude Code cloud sessions on Pro, Max, and Ultra now figure out dependencies, network access, and API keys on their own. - Claude Duet can check running Claude Code sessions, approve changes, and send updates from your phone. - Claude for iOS is testing a side-nav redesign with separate Chats, Projects, and Code tabs. META 🔥: - Muse Spaces let users generate interactive, updatable pages from chat and pin them in the app. - Muse and Meta AI now run specialized models with Active Memory and Active Focus from Muse Wonder. SPACEXSI 🔥: - Grok API text watermarking turns on by default today for listed Grok 4.x models, with team opt-out in Console. ALIBABA 🔥: - Qwen posted a “The becoming begins” teaser, with Qwen4-Max, Flash, Plus, and 27B expected soon. NOUS RESEARCH 🔥: - Nous Research released Arete-27B, a Qwen3.8-27B post-train aimed at Western values. * Used Grok to compose this brief, cherry-picking the news and doing some post-editing.

查看原帖

正文

DAILY AI BRIEF 🗞 — Nov 9

OPENAI 🔥: - Dots can now pull sign-in codes from a connected inbox and fill them in to log in for you. - Faro and Tonada, two Argentinian Spanish ChatGPT Voice voices, are live globally. - A Codex Edition of the ModRetro Chromatic handheld is coming to OpenAI Supply Co. - Wyvern, a stealth OpenAI-routed model with 1M context, was spotted inside Devin (via abhinavflac).

ANTHROPIC 🔥: - FT reports Anthropic took over about $3.4B of gas-generator orders for 1.7GW of data-center power. - Claude Code Projects are getting a collaborative mode with recent activity and scheduled-task overviews. - Create files and artifacts is rolling out to free Claude users (via BartokGabi17). - Anthropic is also working on Duet invites and merging Claude Enterprise organizations. - Claude "Reflection" feature is rolling out to more users. - Anthropic is working on the invite logic for Claude Duet.

GOOGLE 🔥: - Google is testing a next-generation Gemini agent for Gemini Enterprise, already in internal use. - An early Gemini checkpoint called Krypton-a reportedly beats Carbon and Argon in coding demos (via chetaslua).

SPACEXSI 🔥: - Grok Bot Slack integration is rolling out so you can DM or @ mention your bot from Slack. - Elon says Grok Bot token-efficiency and smarter-behavior fixes are landing overnight. - Grok Bot Meetings briefly appeared for some accounts before being pulled (via alexmichael).

APPLE 🔥: - Bloomberg reports CEO Ternus pressed Federighi to accelerate AI, with Apple eyeing agentic Siri as early as next year.

PERPLEXITY 🔥: - Perplexity is working on a Multiplayer mode for Projects.

* Used Grok to compose this brief, cherry-picking the news and doing some post-editing.

来源

Testing Catalog