Claude in Chrome 侧边栏升级为 Claude Cowork 会话

Claude in Chrome 浏览器扩展的侧边栏现已升级为 Claude Cowork 会话，对话会保存至历史记录，技能和连接器可在浏览器中工作，且任务可在桌面、网页和移动端应用间无缝切换。（AIHOT源：Claude：Blog（网页））

AutoGPT 如何用 AGENTS.md 和技能门控管理 AI 生成的拉取请求

AutoGPT 维护者发现，AI 智能体不会主动阅读文档，因此将指令放在 AGENTS.md 和技能文件中，并置于代码目录旁。他们通过强制 PR 模板、测试计划、CI 覆盖率门槛和 CLA 签名等门控机制，将智能体提交的 PR 从"不可用"转变为"可用但不符合路线图"。其中 CLA 签名因需浏览器和 OAuth 流程，被用作区分人类与智能体的"人类探测器"。（AIHOT源：GitHub Blog）

空货架还是丢钥匙？Google 研究：Recall 是参数化事实性的瓶颈

Google Research 提出知识画像框架，发现前沿 LLM（如 Gemini3、GPT-5）的事实编码接近饱和，但回忆（recall）能力不足，多数事实错误源于"丢钥匙"而非"空货架"。该框架将事实分为编码失败、回忆失败等五类画像，并配套推出 WikiProfile 基准，含 2，150 条维基百科事实，每条配 10 个问题，用于分别探测编码、回忆与识别能力。（AIHOT源：Google Research：Blog（网页））

阿里开放 Qwen3.8-2.4T-A95B 模型权重：2.4T MoE、激活 95B、原生 256K 上下文

阿里 Qwen 团队正式开放 Qwen3.8-2.4T-A95B 模型权重，这是 Qwen-Max 级别模型首次开源。模型总参数 2.4T，每个 Token 激活 95B，原生支持 262，144 Token 上下文并可扩展至 1，010，000 Token。（AIHOT源：IT之家（RSS））

微软首发自研推理模型MAI-Thinking-1

我们的首个推理模型 MAI-Thinking-1 从零开始构建，现已在 Microsoft Foundry 上线。为团队点赞！更多详情如下。（AIHOT源：X：Mustafa Suleyman（Microsoft AI CEO） (@mustafasuleyman)）

我写了一本 AI 教科书--AI 还要多久才能写得更好？

作者在完成一本 RLHF 教科书后反思，LLM 在长文非虚构写作上进展停滞，GPT 4.5 和 Kimi K2 等写作强模型已显老旧，而编码、数学等任务已接近超人水平。模型能改错字、做编辑，但组织整章内容时仍混乱且易出错，作者认为这阻碍了模型自主解决开放科学问题。（AIHOT源：Nathan Lambert：Interconnects（RSS））

Meta 开源 Muse Glimmer 登陆 OpenRouter

Meta AI 超级智能实验室的首个开放权重模型 Muse Glimmer 已在 OpenRouter 上线！ 这是一款 30B 密集文本+图像模型，采用 Apache 2.0 许可证，旨在成为可靠的本地智能体，MCP Atlas 得分 75.5，SWE-Bench Pro 得分 51.2。 https://openrouter.ai/meta/muse-glimmer-30b（AIHOT源：X：OpenRouter (@OpenRouter)）

LTX-2.5 模型登场：AI 生成 10 秒 720P 视频仅需 6.8 秒，原生集成 ComfyUI

LTX 推出 LTX-2.5 模型，原生集成 ComfyUI，在 2 张英伟达 GB200 配置下生成 10 秒 720P 视频仅需 6.8 秒。LTX-2.5 Fast 以每秒 0.09 美元生成带音频 720p 视频，10 秒片段成本 0.90 美元；年度经常性收入低于 1，000 万美元的组织可免费使用。（AIHOT源：IT之家（RSS））

Research Gold 号称"100%人类撰写、绝不使用AI"，实则全程由AI驱动

面向医学研究者的网站 Research Gold 宣称其服务"100%由人类撰写、绝不使用AI"，并列出多名博士审稿人。但调查发现，这些审稿人系AI生成、并不存在；部分真实方法学家的身份和照片未经许可被挪用。致电该公司时，自称"Sarah"的AI助手坚称自己是真人，邮件与聊天回复也均为AI生成。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

零基础用户半天上手AI的12步实操流程

文章给出一套零基础用户半天上手AI的12步实操流程：准备内存不低于16G的电脑，订阅ChatGPT并安装Codex或使用WorkBuddy，用语音输入法以【背景、痛点、需求】框架向AI交代任务，经苏格拉底提问澄清需求后投喂文件让AI直接完成，最后沉淀为可复用Skill。文中建议Codex选GPT-5.6 Sol最高模型，WorkBuddy选Kimi K3。（AIHOT源：公众号：数字生命卡兹克）

Cursor 与 SpaceXAI 联合发布 Grok 4.6

Cursor 与 SpaceXAI 今日发布 Grok 4.6，重点强化长时运行智能体与交互式视觉任务，在多项智能体编程与知识工作基准上达到前沿水平，并在 Artificial Analysis Intelligence Index 上追平 GPT-5.6 Sol。（AIHOT源：Cursor Blog）

xAI 发布 Grok 4.6，强化长时运行智能体能力

xAI 今日发布 Grok 4.6，在 Grok 4.5 基础上重点强化长时运行智能体及更复杂的交互式与视觉工作能力。该模型在多项智能体编码与知识工作基准上达到前沿水平，在 Artificial Analysis Intelligence Index（九项基准综合分）上追平 GPT-5.6 Sol。（AIHOT源：xAI：News（网页））

Bug Smash: restoring dropped Gemini chat config in Sentry's JavaScript SDK

I Built a Notebook for Sharing Notes That Doesn't Ask You to Sign Up First

The Next Evolution of Software Developers

Latency vs. Tokens: What I Learned Optimizing an Agent with Gemma (and What Didn't Work)

Managed Inference on Google Cloud: Pairing the Gemini Enterprise Agent Platform with Cloud Run

Thibault Sottiaux: We did it, finally..

We did it, finally... Codex & ChatGPT desktop, now on Linux. Thanks for waiting and you can cancel that MacBook order if you got impatient. It’s that good. 👀 https://t.co/QiiDPFct2v｜@thsottiaux · 热度 11254 · 2026-08-11

Thibault Sottiaux: I previously promised a reset for every 1M in additional active users for Cod...

I previously promised a reset for every 1M in additional active users for Codex, until 10M. We blew past that and have been silent since 10M. Little surprise for you tomorrow.｜@thsottiaux · 热度 3605 · 2026-08-12

Boris Cherny: LLMs still produce bugs, but those bugs are different than what they used to be

LLMs still produce bugs, but those bugs are different than what they used to be. It’s less off-by-ones and more about system design, ui usability, missing broader context. Some kinds of coding has been solved, but not all. While models continue to improve, adversarial code review has been an incredibly powerful tool to catch many of these kinds of bugs. It can be as simple as a one line prompt - “use a dynamic workflow to adversarial test every edge case in an iOS simulator”, or use Claude’s built in /code-review (or /code-review low, /code-review medium, etc.)｜@bcherny · 热度 2862 · 2026-08-11

Thibault Sottiaux: Import your world

Import your world. Codex. Run. https://t.co/Epj4ULlAC6｜@thsottiaux · 热度 2402 · 2026-08-11

Thariq: This is part of working with the EU AI Act, other labs are adding similar wat...

This is part of working with the EU AI Act, other labs are adding similar watermarking. It’s hard to identify AI-generated text, and this gives people better tools to do that. We’ll also be a shipping text detection API that you can use yourself. https://t.co/OZtG0fNsqc｜@trq212 · 热度 1504 · 2026-08-11

Garry Tan: Deep alignment of your AI with you and your context is massively important I...

Deep alignment of your AI with you and your context is massively important Important new work by @ibab and team https://t.co/TijcK96ckO｜@garrytan · 热度 386 · 2026-08-11

Guillermo Rauch: The growth of @aisdk has been astonishing

The growth of @aisdk has been astonishing. ~80.5M downloads every 30 days, and growing faster than all AI labs SDKs. Most importantly: open and provider-agnostic. https://t.co/EZxyu94gk0 https://t.co/sdVZTZhsCP｜@rauchg · 热度 344 · 2026-08-12

Peter Yang: Trying to onboard my parents to ChatGPT desktop app and the separation betwee...

Trying to onboard my parents to ChatGPT desktop app and the separation between Chat, Work, and Codex and the lack of consistency across ChatGPT web, desktop, and mobile apps is all incredibly messy. I think the team should do a clean up / quality pass or maybe ask Codex to do it. (I say all this as a Codex fan)｜@petergyang · 热度 326 · 2026-08-12

Madhu Guru: There’s going to be a ton of money in making open-weight models exceptional a...

There’s going to be a ton of money in making open-weight models exceptional at boring, specific business domains. Pick a size x domain combination and go deep. Eg mid-market legal, smb retail, enterprise logistics. The hyperscalers have the primitives but will struggle with domain depth, scrappiness and sheer will to make a model exceptional for one particular kind of business. And that is your opportunity.｜@realmadhuguru · 热度 201 · 2026-08-11

Aaron Levie: Good post on how to think about FDEs

Good post on how to think about FDEs. The key is that FDEs are real and not going away for AI any time soon. The reason this is happening now at a scale that has never been seen before is because AI is fundamentally about adding a non-deterministic, rapidly changing system to workflows that largely have never been automated before. This looks very little like implementing traditional software. “If you're building an AI agent for accounting in 2026, there is no established workflow, because literally nobody has ever used one of these. Nobody knows what the user journey looks like — not you, and importantly, not your customer either. They can't tell you what they want, because the thing they'd want doesn't have a shape yet.” Software has largely always been deterministic and once implemented effectively worked the same for customers. This meant the upfront implementation work was *relatively* uniform across similar customers, and the system wasn’t regularly being upgraded in fundamental ways. AI agents are entirely different on nearly every dimension. The customer’s business process has to change to work with agents, there is heavy -necessary- customization to get agents to work in the customers end-state process, evals need to be run constantly, the AI models are constantly changing and updates need to keep getting incorporated, the underlying harness and broader system are often changing due to customer feedback, and much more. This is real work for the customer, systems integrators, and the applied AI vendors. Even as AI capabilities improve dramatically, this work remains (or even gets more complicated) given enterprises will just throw increasingly more complex processes at agents. Great time to be an FDE.｜@levie · 热度 126 · 2026-08-12

Thariq: All Claude generated text will have this embedded watermarking

All Claude generated text will have this embedded watermarking. For example, you could check if a PR was generated by Claude Code. That said, it does have limitations. https://t.co/umfn6nU5up｜@trq212 · 热度 124 · 2026-08-11

Josh Woodward: 🌏On behalf of the whole Gemini team, thank you Whether you use Gemini on the ...

🌏On behalf of the whole Gemini team, thank you Whether you use Gemini on the Web, Android, iOS, or inside Chrome, we’ll keep listening to your feedback as we keep building for the next billion!｜@joshwoodward · 热度 101 · 2026-08-11

Tailscale Traces Database Corruption to 16y/o SQLite WAL-Reset Bug

AI is removing the middle class of software engineering?

DeepSeek V4 Pro 0813

License plate reader searches should require a warrant

2026 Eclipse Webcams

深夜放大招！DeepSeek V4 Pro 正式版 API 更新上线，多项测试性能接近 Fable 5

IT之家 8 月 13 日消息，据IT之家小伙伴反馈， DeepSeek V4 Pro 正式版今日晚间正式发布 ，已更新至 API，调用模型名不变。 新版本增强了 Agent 能力，支持 Responses API 和 Codex 接入。 从官方群放出的评测对比表可以看到，DeepSeek V4 Pro 正式版（DeepSeek-V4-Pro-0813）在多项测试中 接近 Fable 5 水平 ，相比之前的预览版能力大幅提升。 IT之家附定价如下： 百万 tokens 输入（缓存命中）：0.025 元 百万 tokens 输入（缓存未命中）：3 元 百万 t...

剑指 GPT-5.6 Sol，SpaceXAI 发布 Grok 4.6 模型

IT之家 8 月 12 日消息，北京时间今天（12 日）晚间，Grok 4.6 正式发布。新模型在 Grok 4.5 基础上进一步强化长时间运行的智能体任务，以及复杂的交互和视觉工作。 按照发布信息，Grok 4.6 能够持续处理包含大量步骤的复杂任务，包括 资料研究、信息分析、大型代码库处理 ，以及将产品构想转化为完整应用或工作成果。 基准测试方面，Grok 4.6 在多项智能体编程和知识工作测试中达到前沿水平。在综合 9 项基准测试的 Artificial Analysis Intelligence Index 上， Grok 4.6 与 GPT-5.6 Sol 取得相同成绩 。...

从亚百毫秒级启动到生产级部署，腾讯云为何重构 Agent 沙箱？

点击查看原文>

Vercel 发布新语言 Zero：代码不是写给人看的，而是写给 AI 的

点击查看原文>

别再所有人平分AI算力：顶级模型给资深工程师才省钱，新人刷题式成长已失效

点击查看原文>

I hate packaging my software for Linux

How Tailscale helped find the SQLite WAL-Reset bug

Guide to (not) fucking up QR codes

I'm done using AI

My Homelab Got Hacked - A Postmortem

Scientists just created female clones of male mice

Scientists have deliberately turned male mouse embryos into females for the first time. A team based in Japan used a CRISPR-based approach to remove the Y chromosome from male cells and create female clones of male mice.  “No one has done this before,” says Monika Ward, a reproductive biologist a...

Scaling AI agents with trustworthy data

Business and technology leaders need no convincing that the time of agentic AI is here. Organizations are rapidly adopting agents, and few executives doubt the technology’s potential to transform work. But many organizations find that realizing the desired return on investment (ROI) from AI hinge...

The Download: our 35 young innovators and the “censorship-industrial complex”

This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. How we picked 35 of the world’s top young scientists and engineers On September 8, MIT Technology Review will reveal its 2026 list of Innovators Under 35, reco...

Some Claude users are mad that Anthropic’s new watermarks will catch them cheating at their jobs, classes

Is Anthropic's new watermarking system a travesty? Some have taken to social media to complain that it is.

AI nuclear power firm Fermi finally has a new CEO

Lee McIntire, an independent member of Fermi's board, has been hired as CEO, more than three months since the company fired co-founder Toby Neugebauer from the top post.

Northrop’s robot space mechanic is a new way to keep satellites at work longer

The Mission Robotic Vehicle is making the first attempt to attach a new thruster to an aging satellite.

Pixel 11 event live blog: Let’s watch Trevor Noah introduce Google’s new phones

It's almost time for the Made by Google keynote, where the company will show off the brand-new Pixel hardware it announced today. Like last year, it'll be a celebrity-packed live show, though Trevor Noah is hosting instead of Jimmy Fallon. If the 2025 show was any indication, today's broadcast mi...

Cats and dogs are missing meals after a popular smart feeder went down

A Petlibro outage is preventing its smart pet feeders and other devices from performing scheduled tasks, like dispensing food. The outage began on Tuesday, with users across Reddit reporting that their smart feeders, litter boxes, and water fountains have gone offline. Though Petlibro maintains t...

把「闪念胶囊」（Almost）放到 iPhone 锁屏

想法被及时、完好的记下来，然后呢？ 查看全文

对话少数派老麦：从 Palm、WebOS 到 AI 硬件，依然不想成为多数派

《三五环》主播刘飞邀请老麦聊了聊他从 Palm 时代开始的二十年数码时代的经营历程。复盘了少数派如何从一家媒体，一步步跑通「内容+产品」的商业闭环；也聊到了他与罗永浩那段「不打不相识」的往事，以及那些改变了他对科技媒体商业模式认知的故事。 查看全文

我，与「唯一」的 iPhone Air

在 iPhone Air 之后才发现，有些产品存在的意义，是帮我确认我真正需要什么。 查看全文

uni-app 跨端布局与适配技术指南

> **核心目标**：解决“碎片化”屏幕尺寸问题。 > **适用人群**：iOS/Android/鸿蒙原生开发者、前端工程师。 > **核心价值**：理解 uni-app 如何通过单位转换和

OctaFuse 2.4.0：让 DashScope 的 ASR、TTS 语音能力进入统一路由

当语音识别和语音合成进入真实业务，网关面对的不再只是一次普通的 HTTP 请求：ASR 可能需要同步转换或异步任务，TTS 可能返回完整音频或持续输出音频流，实时语音还需要在 WebSocket 上双

浅析大模型推理十二篇之关键指标

前两篇分别讲了 Prefill / Decode 两阶段，以及 KV Cache 如何用「空间换时间」避免重复计算。理解机制之后，下一步是：怎么衡量一个推理系统好不好？

芯片制造商Cerebras硬件销售下滑 盘后股价大跌

Cerebras Systems Inc．盘后下跌，此前该公司公布硬件收入下降，显示这家上市不久的公司在销售采用新型芯片设计的计算机方面进展缓慢。

美联储料维持准备金管理购买节奏

美联储预计将在周四公布新一期准备金管理购买计划时，继续安排下一阶段的国库券购买。随着未来数月流动性预计减少，决策者希望借此充实准备金。

8月13日外盘头条：德黑兰正为地区冲突长期化做准备 美国7月核心通胀表现温和 思科公布远超预期财测

全球财经媒体昨夜今晨共同关注的头条新闻主要有： 1、美伊谈判陷入僵局 德黑兰正为地区冲突长期化做准备 2、美国7月核心通胀表现温和 美联储加息概率下降 3...

多空在160大关附近鏖战 交易员密切关注日元会否再遭干预

日元兑美元周三进一步逼近160大关，令投资者密切关注当局是否会再次进场干预。 当日纽约尾盘日元兑美元下跌0.1%至159.43。本月以来日元已贬值逾1%...

大宗商品综述：油价走势震荡 铜价续跌 金价上涨

原油盘中反复震荡，伊朗与美国围绕霍尔木兹海峡的谈判几乎未见突破迹象，与此同时，交易员研判多家主要能源预测机构发布的一系列数据。

《置身谷内》！Jeff Dean上顶会自曝离职现场：被1500人围堵

回消息到凌晨2点半

Anthropic CEO整天神神叨叨，投资人受不了了

Dario能不能少吓唬点人…

国产具身智能创全球新纪录！以30%成本跑赢 Figure AI 45%效率，聪明的具身大脑成关键

具身模型一小时狂拣1816件异形包裹