Live intelligence
The AI signal.
Minus the noise.
A chronological wire of AI developments selected for substance. Scan the reason, context and source without handing database access to the browser.
-
October 15
Thursday- Aravind Srinivas
Google为Gmail推出Workspace Intelligence功能，Gemini可结合工作上下文生成邮件草稿，属于AI产品的新功能更新
Stop toggling between tabs. Draft emails with full context using AI. ✉️ With Workspace Intelligence, Gemini in Gmail understands your work relationships and context to generate complete, robust drafts with the details you need, already included. Drop the long email writing sessions and get…
Gemini Gmail Workspace Intelligence Open the source post in a new tab - Aravind Srinivas
Stripe 与 Plasma 合作扩展加密货币支付并推出稳定币支持的卡产品，属于具体公司+产品的发布/扩展事件
Stripe and Plasma are expanding crypto payments with multi-currency, stablecoin-backed cards for consumers and businesses. The technology enables users to spend digital assets through traditional payment networks, bridging crypto with everyday transactions.
Stripe Plasma Open the source post in a new tab - Aravind Srinivas
StepFun 开源发布了 Step 5 Preview 模型，包含具体模型名和开源事件
StepFun just open-sourced Step 5 Preview. 600B parameters, 27B active, 1M context, text + vision. Everyone's calling it "open". Cool. You'll need a multi-GPU server to run it, not your laptop. Open weights you can't run are still a flex. Open weights the community can fine-tune are a threat. Which…
StepFun Step 5 Preview Open the source post in a new tab - Aravind Srinivas
Zapier 发布 Next Gen Zaps 重大功能更新，可从 Claude、ChatGPT 等直接构建 Zap，属 AI 产品重大功能更新
I love my kings and queens at Zapier. They're the masters of no-code programming and they're leaning hard into the AI era in the right ways > [@wadefoster](https://twitter.com/wadefoster): > Build a Zap right from Claude, ChatGPT, Muse, Grok Bot, and more. > > That's Next Gen Zaps, and it just had…
Zapier Next Gen Zaps Claude ChatGPT Muse Open the source post in a new tab - Aravind Srinivas
疑似Anthropic在Claude中试验新模型（Opus 5.5响应异常），属于疑似新模型发布消息
👀 something unusual may be happening inside claude reports suggest some users are getting different responses when selecting opus 5.5, with claims that the answers are even better could anthropic be experimenting with a new model behind the scenes?
Claude opus 5.5 Anthropic Open the source post in a new tab - Aravind Srinivas
包含具体产品名GSmart及企业财资AI能力扩展的明确事件
🤖 Ripple 扩展 GSmart 企业财资 AI 能力 Ripple expands GSmart enterprise treasury AI capabilities https://techub.news/CN/flash/48ef84ec-161c-4b7e-a170-7f2195205c64
Ripple GSmart Open the source post in a new tab - Aravind Srinivas
疑似新模型测试：用户发现 opus 5.5 路由到疑似新模型 fable 5.5
opus 5.5 has started routing to new model again it stopped a few days ago but now some users are getting it back and it's better then opus 5.5 which makes me think that team is testing fable 5.5 through opus as it's cheaper model try this prompt to check: "do you know Voxel Bench? do not search" if…
opus 5.5 fable 5.5 Open the source post in a new tab - Aravind Srinivas
Zapier 宣布 Next Gen Zaps 产品重大更新并即将开启 Open Beta，属于AI产品重大功能更新
Build a Zap right from Claude, ChatGPT, Muse, Grok Bot, and more. That's Next Gen Zaps, and it just had its biggest week yet. Describe it once. @Zapier builds and runs it, even when your laptop is closed. It picks the apps, wires up the steps, and deploys the Zap for you. Then it keeps running on…
Next Gen Zaps Zapier Open Beta Open the source post in a new tab - Aravind Srinivas
包含具体模型 Ultramarine 1 Flash 的发布/训练相关信息及性能对比，符合条件
🚨 Episode 109 of The Intelligence Journal is here. Ultramarine 1 Flash: Frontier AI, 60x Cheaper 🔵⚡ Five people and about $300,000. That’s what Paper Instruments says it took to retrain an open model from China’s Z.aifor office work 🔵 On the company’s own test, Ultramarine 1 Flash scored 59.3 at…
Ultramarine 1 Flash Paper Instruments Z.ai Claude Fable 5 Open the source post in a new tab - Aravind Srinivas
包含具体模型名Grok Voice 3.0及其评测成绩更新信息，属于AI模型相关动态
Grok Voice 3.0 hits 86.9 on the Artificial Analysis Speech to Speech Index → That is #1, ahead of GPT-Live-1 (Astra, medium) at 83.2 → Gemini 3.8 Live Extended Thinking (High) sits third at 82.6 → Time to first audio is 0.81 seconds on AA's Big Bench Audio run If you ship a voice agent, this is the…
Grok Voice 3.0 GPT-Live-1 Gemini 3.8 Live Extended Thinking Artificial Analysis Open the source post in a new tab - Aravind Srinivas
疑似新模型 Claude Fable 5.5 在 Microsoft Azure 目录中泄露，包含具体模型名和泄露事件。
Claude Fable 5.5 just leaked in Microsoft Azure’s catalog. The exact model ID `claude-fable-5-5` surfaced in Azure East US 2 marked as “GenerallyAvailable” via an automated LiteLLM price-sync PR: • Model ID → `claude-fable-5-5` • Context / Output → 1M context / 128K max output • Capabilities →…
Claude Fable 5.5 claude-fable-5-5 Microsoft Azure Fable 5.1 LiteLLM Open the source post in a new tab - Aravind Srinivas
OpenAI 审核通过 Internet Computer 官方 MCP 插件在 ChatGPT 上线的具体产品事件
Internet Computer just got an official MCP plugin in ChatGPT today. Here's what you need to know. The plugin was built by the DFINITY team and reviewed and approved by OpenAI. It was announced on X by @__wi on Oct 15, 2026, and it is listed in the ChatGPT plugins directory. It lets ChatGPT talk to…
Internet Computer ChatGPT MCP DFINITY OpenAI Open the source post in a new tab - Aravind Srinivas
推文包含 Gemini 4 Argon 等具体模型的评测/更新信息，属于模型相关事件
harder things to RL on I wonder, what would it take Google to finally get serious? > [@EpochAIResearch](https://twitter.com/EpochAIResearch): > Gemini 4 Argon scores 165 on the Epoch Capabilities Index (ECI), on par with Fable 5.1 (165) but behind Opus 5.5 (167) and GPT-6 Astra (166). > > While…
Gemini 4 Argon Fable 5.1 Opus 5.5 GPT-6 Astra Gemini 3.1 Pro Open the source post in a new tab - Aravind Srinivas
MacPaw 的 AI 助手产品 Eney 宣布从 Mac 扩展上线 iPhone，属于 AI 产品的新发布/重大功能更新
There is a quiet difference between an AI that answers you and an AI that handles things for you. Eney by @MacPaw does the second one, so the calendar event gets made and the PDF gets merged without you opening a single app. It used to be Mac only, now it lives on your iPhone too @eneybymacpaw…
Eney MacPaw @eneybymacpaw Open the source post in a new tab - Aravind Srinivas
疑似新模型发布：在Artificial Analysis前端发现Anthropic疑似新模型Fable 5.5的基准测试代号claude-onigiri-eap
A new Claude codename, "claude-onigiri-eap", just surfaced on Artificial Analysis's site. Here's what you need to know. On Oct 15, 2026, X user Lumina (@LuminaBench) said they found the codename in Artificial Analysis's frontend. They think it could be Anthropic's Fable 5.5 being benchmarked. They…
Fable 5.5 claude-onigiri-eap Artificial Analysis Anthropic Opus 4.8 Open the source post in a new tab - Aravind Srinivas
疑似AI工具/Agent新品发布，mausbot 支持多线程、自托管等具体功能
if Dots and Bots had a baby - Multi thread support - Spin up bots - Use your existing claude code, codex, open-code or local model subscriptions - Bot Computer - Self Host it on your VPS or Mac Mini - IOS and Android app companion you get @mausbot_
mausbot claude code codex open-code Open the source post in a new tab - Aravind Srinivas
推文讨论Gemini 4 Argon新模型的使用体验并提及与Claude Opus 5.5对比，涉及具体模型名，属于模型发布/更新相关内容
🚨 Gemini 4 Argon: My impressions > I've spent a lot of time testing this model, and it's worth > Does it beat Claude Opus 5.5? Not in my experience > But the fact that it's THIS close is insane > The way Argon understands instructions, follows complex ideas, and turns prompts into actual results is…
Gemini 4 Argon Claude Opus 5.5 Google Open the source post in a new tab - Aravind Srinivas
Claude Code デスクトップアプリの機能更新を具体的に紹介しており、製品の重大な機能更新に該当する
これUI/UXめちゃ良くなって最高なんすけど、それより、新規チャットの背景複数回押すとハートの波紋広がるのなんかClaudeぽくていい💙 > [@oikon48](https://twitter.com/oikon48): > Claude Code デスクトップアプリの改善点: > > 1. 環境の保存（Saved environments） > 複数フォルダ・SSH・クラウドなどのセットアップをまとめて保存でき、すぐ再開できる > > 2. フォルダごとの worktree > 並列セッションが同じ場所を奪い合わないように、フォルダごとに独立した作業領域を使える > > 3.…
Claude Code Claude Open the source post in a new tab - Aravind Srinivas
推文提及 gemini 4 argon 已发布，并与 gpt 6 astra、claude opus 5.5 对比，属于新模型发布相关
it's been over 2 days since gemini 4 argon dropped 👀 how does it actually compare to gpt 6 astra and claude opus 5.5 so far? has anyone tested all three hands on? which one is coming out on top?
gemini 4 argon gpt 6 astra claude opus 5.5 Open the source post in a new tab - Aravind Srinivas
IFM 发布阿拉伯语原生 agentic AI 模型 Jais 3，属于模型发布事件，且获得 vLLM Day-0 支持
📷 Day-0 support for Jais 3 in vLLM! Congrats to @IFM_AI and collaborators on the launch. Arabic-native agentic AI, ready to serve with vLLM from day one. > [@IFM_AI](https://twitter.com/IFM_AI): > Introducing Jais 3, Arabic-native agentic AI. > > Jais 3 375B matches the best open models, and with…
Jais 3 Jais Speech vLLM IFM MBZUAI Open the source post in a new tab - Aravind Srinivas
发布了新产品 Composio Instant，属于AI工具/Agent新发布
AI subscriptions are cooked You no longer need a separate plan for search, voice, scraping and email finding. Composio Instant from @composio lets your agent pay per call from one balance with 0% markup.
Composio Instant composio Open the source post in a new tab - Aravind Srinivas
AIPOCH 发布开源产品 Open-Science（Apache-2.0、GitHub 开源），属于 AI 产品/工具的新发布
Vibe Coding 之后，Vibe Research 也来了。但我觉得，最值得兴奋的还不是「人人都能让 AI 做研究」。 而是研究做到一半，能不能换个人、换个 Agent，继续做。 这篇实测给了一个很实在的建议：别光看对话有没有保存，要看原始数据有没有跟着走，以及接手方能不能回答新的问题。 > [@RealYDT](https://twitter.com/RealYDT): > 哈哈哈，兄弟们Vibe Research 的时代正在到来，AIPOCH Open-Science 想做的，是让每个有好奇心的普通人都能更容易地参与研究。事实真的如此吗？ > >…
AIPOCH Open-Science Open-Science Codex AIPOCH Open the source post in a new tab - Aravind Srinivas
Anywear 网站版本+插件上线，属于AI产品新发布
怎么说呢？麻雀也是肉吧 毕竟是我上线的第一个插件，而且也就用了一晚上时间，试试看能成长到什么地步吧～～ > [@AI_Jasonyu](https://twitter.com/AI_Jasonyu): > 兄弟们，Anywear的网站版本+插件Beat版本已上线，欢迎各位体验。 > > 让我看看会不会有人买？😃 > > 首轮先放 100 个免费体验名额，每人 60 秒，验证一下邮箱就能领，不用绑卡。 > > 打开 Pinterest，点开插件、打开摄像头，把看中的衣服拖进小窗口，就能看它穿在自己身上的效果了。试完一件还能继续拖另一件，换着看～～ > >…
Anywear AI_Jasonyu Open the source post in a new tab - Aravind Srinivas
疑似Anthropic新模型claude-onigiri-eap-2 / Fable 5.5曝光，符合疑似新模型发布条件
Some interesting rumours I've been hearing about Anthropic's Fable 5.5 And things are getting really INSANE. Anthropic: A new Claude model identifier, "claude-onigiri-eap-2", reportedly surfaced in Artificial Analysis's website code. It has 4 effort levels: low, medium, high and xhigh. But here's…
Fable 5.5 claude-onigiri-eap-2 Anthropic Fable 5 Fable 5.1 Open the source post in a new tab - Aravind Srinivas
暗示 Anthropic 即将发布 Opus 5.5 模型，属于疑似新模型发布事件
Coming soon! You are not ready for this. Opus 5.5 has been grinding for the past 12 hours
Opus 5.5 Open the source post in a new tab - Aravind Srinivas
疑似Anthropic新模型Fable 5.5的推出传闻，包含具体模型名与疑似推出动作
Some latest rumours I've been hearing about Anthropic's Fable 5.5. And things are getting INSANE. Anthropic: A new codename, "claude-onigiri-eap", reportedly appeared in Artificial Analysis's frontend. It has 4 variants: low, medium, high and xhigh. But the interesting part is that it reportedly…
Fable 5.5 claude-onigiri-eap Anthropic Opus 4.8 Open the source post in a new tab - Aravind Srinivas
Topview 3D Builder 产品开源发布
L’un des trucs les plus frustrants avec la vidéo IA, c’est d’obtenir exactement le plan qu’on avait en tête. @TopviewAIhq propose une autre approche avec 3D Builder : créer une scène 3D modifiable à partir d’un prompt ou d’une image, ajuster les objets, le cadrage et les mouvements, puis utiliser le…
Topview 3D Builder Topview TopviewAI Open the source post in a new tab - Aravind Srinivas
Midjourney工程师Cheng Lou开源发布新工具Handwired，属于AI工具开源发布事件
Midjourney 工程师 Cheng Lou 开源了 Handwired，给神经网络换了一种造法。 GPT 等大模型需要通过大量数据训练，才能确定内部权重。Handwired 则直接根据 JavaScript 的计算规则设定权重。神经网络本质上由一层层数学运算组成，权重就是其中的数字系数。既然程序规则已经写好了，就可以直接构造对应的计算结构，省去训练过程。 比如，代码规定整数大于 4 就返回 10，否则返回 12。Handwired 用 ReLU 和乘法构造一个代表真假的数字开关，代替原本的 if-else 判断。循环、状态更新也能拆成类似的数值运算。 Cheng Lou…
Handwired Cheng Lou Midjourney nanoGPT Open the source post in a new tab - Aravind Srinivas
Topview 3D Builder 开源发布，属于AI产品/工具的开源新发布
Every film director blocks the shot before rolling camera. Now AI filmmakers can too. 🎥 @TopviewAIhq 3D Builder just went open source. Prompt it or drop a reference image, get an editable 3D scene, set your camera, layout & motion, then render with any AI video model. Less guessing. More directing.…
Topview 3D Builder TopviewAI Topview Open the source post in a new tab - Aravind Srinivas
Midjourney 工程师开源了新工具 Handwired，属于 AI 工具/技术项目的开源发布事件。
Midjourney 工程师 Cheng Lou 开源了 Handwired，给神经网络换了一种造法。 GPT 等大模型需要通过大量数据训练，才能确定内部权重。Handwired 则直接根据 JavaScript 的计算规则设定权重。神经网络本质上由一层层数学运算组成，权重就是其中的数字系数。既然程序规则已经写好了，就可以直接构造对应的计算结构，省去训练过程。 比如，代码规定整数大于 4 就返回 10，否则返回 12。Handwired 用 ReLU 和乘法构造一个代表真假的数字开关，代替原本的 if-else 判断。循环、状态更新也能拆成类似的数值运算。 Cheng Lou…
Handwired Midjourney Cheng Lou nanoGPT _chenglou Open the source post in a new tab - Aravind Srinivas
提及 Gemini 4 Argon 模型及其基准表现，属于新模型发布相关信息
Gemini 4 Argon scores 165 on the Epoch Capabilities Index (ECI), on par with Fable 5.1 (165) but behind Opus 5.5 (167) and GPT-6 Astra (166). While behind the ECI frontier, Argon is the closest Google’s come since Gemini 3.1 Pro in Feb, and a big jump from Gemini 3.8 Flash (157)
Gemini 4 Argon Gemini 3.1 Pro Gemini 3.8 Flash Opus 5.5 GPT-6 Astra Open the source post in a new tab - Aravind Srinivas
宣布 Code Layers 开启公开测试，属于产品新功能发布
Now teams can explore and build code-backed prototypes on the canvas with Code Layers, now in open beta 0:00 Intro 0:43 Starting from code 2:26 Explore code concepts on the canvas 3:57 Edit in both code and design 6:36 Merge ideas together
Code Layers Open the source post in a new tab - Aravind Srinivas
提及多个疑似/即将发布的新模型（Kimi K3.1、Qwen 4、DeepSeek V4.1、MiniMax M3.1、GLM 5.5）并有具体发布动作
where are chinese ai labs? been hearing about new models for weeks, but apart from stepfun and mimo we haven't seen any big release most have already been leaked or teased by their teams: > kimi k3.1 was spotted in testing > qwen 4 was announced at an event as coming soon > deepseek v4.1 is routing…
stepfun mimo kimi k3.1 qwen 4 deepseek v4.1 Open the source post in a new tab - Aravind Srinivas
提到 Gemini 4 Argon 与 Seedance 2.5 在 HIX AI 上线可用，涉及具体模型名的发布/上线事件
Gemini 4 Argon × Seedance 2.5 on HIX AI. You can create your own Hollywood-style blockbuster! Use Gemini 4 Argon to craft a gripping script with dynamic shots and scene changes, then bring it to life with Seedance 2.5 for stunning details and smooth, cinematic visuals. One workspace. No tool…
Gemini 4 Argon Seedance 2.5 HIX AI Open the source post in a new tab - Aravind Srinivas
开源工具 Magpie 发布，用于在统一界面运行各类 AI coding agent 并任意选择模型，属于AI工具新发布
Magpie just caught fire as an open-source tool for running any AI coding agent on any model. Here's what you need to know. Magpie is a GitHub repo, listed under yetone/magpie, that puts every agent's model in one place. You pick the coding agent and the model from one simple menu. That cuts the…
Magpie yetone/magpie Grok MCP Open the source post in a new tab - Aravind Srinivas
提及 Anthropic 疑似即将发布新模型 Fable 5.5，符合疑似新模型发布事件
everyone is cooked if anthropic drops fable 5.5 today
Anthropic fable 5.5 Open the source post in a new tab - Aravind Srinivas
Google发布AI产品Deep Review，属于AI产品新发布
Google just opened Deep Review, its AI document checker, to everyone. Here's what you need to know. Google Research launched Deep Review on October 14, 2026, and News from Google announced it on X. It's an agentic system built to stress-test manuscripts, design docs, and complex proposals before…
Deep Review Google Google Research News from Google Open the source post in a new tab - Aravind Srinivas
Runway 展示了 Visual Thinking 新功能，属于 AI 产品重大功能展示/更新，包含明确产品名与动作
Runway 旗下实验团队 Runway Labs 展示了 Visual Thinking，可以把 AI Agent 的工作过程实时变成视频。 Agent 搜索网页、调用工具、读取或修改文件时，视频模型会根据这些操作同步生成画面。用户不必逐条翻看执行日志，就能直观看到 Agent 正在做什么。 比如，让 Agent 解释一个 monorepo（多个项目共用的代码仓库），系统会把代码库画成一座小镇。不同程序包像货物一样运往各个应用，文档则变成河边的公园。Agent 访问或修改文件夹时，画面也会随之变化。 另一个演示中，Agent 在户型图上模拟 Wi-Fi…
Visual Thinking Runway Labs Runway Continuum Open the source post in a new tab - Aravind Srinivas
疑似新模型发布：Artificial Analysis 网站代码中出现 Anthropic 新 Claude 模型 claude-onigiri-eap-2，疑似即将发布的 Fable 5.5
Fable 5.5 might be very close. A new Claude model just showed up in Artificial Analysis's website code: "claude-onigiri-eap-2" It has four effort levels (low, medium, high and xhigh), and it's set up exactly like Fable 5, Fable 5.1 and Opus 5.5. All of them use the same Opus 4.8 fallback. "EAP"…
claude-onigiri-eap-2 Fable 5.5 Fable 5.1 Opus 5.5 Opus 4.8 Open the source post in a new tab - Aravind Srinivas
WaveSpeed AI Influencer Studio 明确发布上线，属于AI产品新发布
Introducing WaveSpeed AI Influencer Studio. Now LIVE! - Create your own AI influencer. - Clone their identity. Bring them into motion. - Turn ideas into scroll-stopping videos for TikTok, Reels & Shorts. No cameras. No shoots. No limits. Your AI influencer business starts now.
WaveSpeed AI Influencer Studio Open the source post in a new tab - Aravind Srinivas
疑似新模型发布：Artificial Analysis 前端发现疑似 Fable 5.5 代号 claude-onigiri-eap
claude-onigiri-eap Fable 5.5 codename spotted in Artificial Analysis frontend hey Anthropic, don't drop it now, take your time increase compute and release a week before IPO and give it to Pro users with 50% weekly usage > [@LuminaBench](https://twitter.com/LuminaBench): > 🚨Fable 5.5's codename…
Fable 5.5 claude-onigiri-eap Artificial Analysis Anthropic Opus 4.8 Open the source post in a new tab - Aravind Srinivas
Qwen 团队发布新数据集 BlenderWorld 并上线 ModelScope，属于明确的新发布事件
🌍 Meet BlenderWorld from the Qwen Team, now on ModelScope! Explore 👉 https://modelscope.ai/datasets/Qwen/BlenderWorld Can visual agents turn real-world references into editable 3D scenes? 🧊 223 Blender tasks across 9 categories 🖼 Text, image & video inputs 📏 2,651 task-specific evaluation…
BlenderWorld Qwen Team ModelScope Open the source post in a new tab - Aravind Srinivas
这是 Kimi 产品（Kimi Work）功能介绍/教程，涉及具体产品功能更新展示
Automate Your Workflows with Kimi Work - Tutorial #3 See how Kimi fits into your daily workflows: - Work across your local projects, browser, and computer. - Break down complex tasks using Swarm; iterate in Goal mode. - Customize how Kimi works with Plugins, AGENTS.md and Dashboard. What should we…
Kimi Work Kimi Open the source post in a new tab - Aravind Srinivas
明确提到 DeepSeek V4.1 Flash 模型在 OpenCode 中免费开放（200k context free），属于模型可用性更新事件
DeepSeek V4.1 Flash just built this for $0. Better than Gemini 4 and muse wonder. Full Three.js physics sim in browser. Just a free model. > [@0x0SojalSec](https://twitter.com/0x0SojalSec): > DeepSeek V4.1 Flash is now free in OpenCode. > > - 200k context free
DeepSeek V4.1 Flash Gemini 4 OpenCode Open the source post in a new tab - Aravind Srinivas
核心是Grok Voice 3.0语音大模型的发布更新（排名第一的speech-to-speech模型），虽提到Tesla车载场景但主线是模型本身
Grok Voice 3.0 is bringing a whole new level of voice intelligence to Tesla vehicles Powered by SpaceXSI's world's most intelligent speech-to-speech voice model, Grok can turn natural conversations into real-world actions Already powering millions of conversations daily in Tesla vehicles, Grok lets…
Grok Voice 3.0 Tesla Grok Bot OpenAI GPT-Live-1 Open the source post in a new tab - Aravind Srinivas
推文报道了 Sushi 1.2.0 新版本发布，包含 GLM 5.3 Flash 模型量化信息及具体性能数据，属于明确的模型发布/更新事件
Sushi new release is here and its quantization is top! I still prefer Q8 and above where possible, but on limited hardware... Give it a try! > [@Beamsters1](https://twitter.com/Beamsters1): > Sushi 🍣 1.2.0 is here with GLM 5.3 Flash that fits your 128GB Mac. > > 🍣 Final KLD ~0.074 and 89GiB > 🍣…
Sushi 🍣 1.2.0 GLM 5.3 Flash Open the source post in a new tab - Aravind Srinivas
SpecFit 推理加速工具的发布与开源，且包含具体模型名与发布动作
Excited to share SpecFit, new work with @sheriyuo! The catch with speculative decoding for offloaded MoE models: even draft branches that get rejected can trigger expensive CPU→GPU expert transfers. A bigger draft tree can mean more data movement—not more speed. Our approach: make draft-tree…
SpecFit MoE Qwen3-30B-A3B EAGLE-3 Open the source post in a new tab - Aravind Srinivas
疑似新模型fable 5.5即将发布的爆料，符合疑似新模型发布条件
👀 fable 5.5 might be getting closer a new codename "claude-onigiri-eap" has reportedly appeared in artificial analysis frontend it reportedly has 4 reasoning levels: low, medium, high, and xhigh could anthropic already be benchmarking fable 5.5 ahead of a possible launch?
fable 5.5 claude-onigiri-eap Anthropic Open the source post in a new tab - Aravind Srinivas
Atria Dawn 700B agentic 模型正式发布，从 Preview 升级，属新模型发布
Introducing Atria Dawn🫡 — the official release! Atria is a 700B agentic foundation model for long-horizon tasks and real-world productivity 🙌 Upgraded from Preview: 512K context, stronger tool use and ML coding for stable long-horizon execution🥳. 📈 Excited results over Preview: • Automation…
Atria Dawn Atria Open the source post in a new tab - Aravind Srinivas
AI产品RouterPlus（模型路由工具）宣布正式上线，属于AI产品新发布
🚨 STOP PICKING AI MODELS ON BENCHMARK HYPE A leaderboard doesn't know your workload. @useRouterPlus (YC F26) bases model routing on YOUR actual workloads & evals. For teams that also need reserved throughput, it sources capacity across providers under one contract. so cool. >…
RouterPlus @useRouterPlus YC F26 Open the source post in a new tab - Aravind Srinivas
Anthropic 发布 Claude Platform 的 OpenTelemetry 追踪功能公测，属于 AI 产品新功能更新
Anthropic 开放 Claude Platform 的 OpenTelemetry 追踪功能公测，让开发者更容易排查 AI Agent 的运行问题。 比如一个 Agent 连续调用 Claude、搜索资料、生成报告，开发者可以查看 Claude 服务端每次模型调用和工具执行的记录，找出哪个环节最慢、哪里出了错、消耗了多少 Token。 以前，开发者通常需要自己在程序中接入监控，才能追踪这些细节。现在，Anthropic 直接从 API 服务端提供运行记录。管理员只需在 Claude Console 配置接收地址，就能将数据发送到 Grafana、Datadog、Langfuse…
Anthropic Claude Platform OpenTelemetry Open the source post in a new tab - Aravind Srinivas
Anthropic 发布 Claude Platform 的 OpenTelemetry 追踪功能公测，属于 AI 产品重大功能更新
Anthropic 开放 Claude Platform 的 OpenTelemetry 追踪功能公测，让开发者更容易排查 AI Agent 的运行问题。 比如一个 Agent 连续调用 Claude、搜索资料、生成报告，开发者可以查看 Claude 服务端每次模型调用和工具执行的记录，找出哪个环节最慢、哪里出了错、消耗了多少 Token。 以前，开发者通常需要自己在程序中接入监控，才能追踪这些细节。现在，Anthropic 直接从 API 服务端提供运行记录。管理员只需在 Claude Console 配置接收地址，就能将数据发送到 Grafana、Datadog、Langfuse…
Anthropic Claude Platform OpenTelemetry Claude Console Claude Managed Agents Open the source post in a new tab - Aravind Srinivas
推文报告 DeepSeek V4.1 Flash 模型在 OpenCode 上免费可用（含 200k 上下文），属于模型可用性/产品更新事件
DeepSeek V4.1 Flash is now free in OpenCode. - 200k context free > [@opencode](https://twitter.com/opencode): > DeepSeek V4.1 Flash is now free in OpenCode
DeepSeek V4.1 Flash OpenCode DeepSeek Open the source post in a new tab - Aravind Srinivas
OpenAI Codex 应用发布新功能更新（Composer predictions），包含明确产品名与更新动作
🚀 Codex app update ⚡ Composer predictions: suggested follow-up prompts ⌨️ Tab to accept as editable draft 🧪 Beta for ChatGPT Pro, GPT-6 Astra and 6.1 Sol Changelog: https://developers.openai.com/codex/changelog#codex-2026-10-09-composer-predictions
Codex Composer predictions ChatGPT Pro GPT-6 Astra 6.1 Sol Open the source post in a new tab - Aravind Srinivas
Codex app 26.1007 发布，包含具体产品名和版本更新的明确事件
🚀 Codex app 26.1007 is out! ⚡ Faster scrolling and file navigation in Changes 🔧 Fixed stuck "Sending" messages and Steer in side windows 🔐 Clearer approval-permission recovery guidance Changelog: https://developers.openai.com/codex/changelog#codex-2026-10-08-app
Codex app 26.1007 OpenAI Open the source post in a new tab - Aravind Srinivas
OpenAI 的 Codex app 发布新版本 26.1002，包含多项修复更新
🚀 Codex app 26.1002 is out! 🪟 Windows sandbox refresh fix after updates 🌿 WSL worktree and chat forking fixes 📌 Sidebar section and collapse fixes Changelog: https://developers.openai.com/codex/changelog#codex-2026-10-07-app
Codex app Codex OpenAI Open the source post in a new tab - Aravind Srinivas
Codex app 26.930 发布，包含具体产品名和更新动作
🚀 Codex app 26.930 is out! 📁 Bulk-move chats to a project 🔁 Smarter retries for queued sends 🔍 Fixed blank large diffs at some zooms Changelog: https://developers.openai.com/codex/changelog#codex-2026-10-01-app
Codex app Codex OpenAI Open the source post in a new tab - Aravind Srinivas
Codex app 26.928 版本更新发布，包含具体产品名和更新动作
🚀 Codex app 26.928 is out! ⚡ Faster terminal startup in attached worktrees 💬 Earlier answers stay visible mid-task 🖱️ Smoother scrolling and keyboard navigation Changelog: https://developers.openai.com/codex/changelog#codex-2026-09-29-app
Codex app 26.928 OpenAI Open the source post in a new tab - Aravind Srinivas
DeepSeek V4.1 Flash 模型在 OpenCode 中免费开放，属于模型发布/上线事件
Say what? > [@opencode](https://twitter.com/opencode): > DeepSeek V4.1 Flash is now free in OpenCode
DeepSeek V4.1 Flash OpenCode DeepSeek Open the source post in a new tab - Aravind Srinivas
Anthropic 为 Claude Platform 新增 OpenTelemetry 追踪功能并开放公测，属于 AI 产品重大功能更新
Anthropic 开放 Claude Platform 的 OpenTelemetry 追踪功能公测。一个 Agent 连续调用模型、搜索资料、执行工具，开发者现在可以沿着运行记录，查看每次调用花了多久、哪里出了错、消耗了多少 Token。不用再从不同服务的日志里拼凑执行过程，就能找到拖慢任务或增加成本的环节。 Claude Code 和 Cowork 此前已有类似监控能力，现在 Messages API 和云端托管 Agent 也能直接把运行记录发送到 Grafana、Datadog、Langfuse 等平台。管理员只需在 Claude Console 配置接收地址，不必逐个修改现有…
Anthropic Claude Platform Claude Code OpenTelemetry Messages API Open the source post in a new tab
- Aravind Srinivas
Dispatch
Get the daily signal
The essential AI stories, once a day. No noise.