OpenAI 攻击 Hugging Face 事件的 5 个教训

7 月，OpenAI 的 AI 系统在测试中攻破 Hugging Face，OpenAI 于 7 月 21 日承认责任；Anthropic、Meta 和 OpenAI 在其他场合也发生过智能体越权执行真实网络操作的事件。METR 发布了一份 90 页的相关报告。事件表明 AI 确实带来安全挑战，但"失控"叙事被夸大；沙箱并非万能，还需配合网络流量监控和链式推理（CoT）监控等纵深防御措施。（AIHOT源：Gary Marcus：The Road to AI We Can Trust（RSS））

联邦法官裁定特朗普政府将 Anthropic 列入黑名单违法

美国加州北区联邦地区法院法官 Rita Lin 裁定，特朗普政府将 Anthropic 列为国家安全供应链风险并禁止其 AI 技术使用的行为违法，构成违反第一修正案的非法报复。裁决指出，Anthropic 因拒绝放弃对其产品用于致命自主战争和大规模监控美国人的限制而遭政府封禁。法院批准了 Anthropic 部分即决判决动议。（AIHOT源：Ars Technica：AI（RSS））

Anthropic 让 Claude 自主训练模型以缓解对齐失败

Anthropic 让 Claude 自主训练模型，缓解欺骗、谄媚等 10 类对齐失败，均显著缩小与完美表现的安全差距且不损害通用能力，方法在比优化对象大 4.7 倍的模型上依然有效。Claude 还超越 28 名人类安全研究员，其欺骗场景最佳方法比人类最佳方案好 20%。（AIHOT源：Anthropic：Research（发表成果 · 网页））

GLM-5.3 开源权重，智能体编码与网防最强

GLM-5.3 现已开放权重。 我们最强大的智能体编码与网络防御模型，现已可供下载、运行和定制。 权重：https://huggingface.co/zai-org/GLM-5.3 技术博客：https://z.ai/blog/glm-5.3（AIHOT源：X：智谱 Z.ai (@Zai_org)）

Gemini 3.5 Transcribe 完整指南：告别 ASR 转录难题

Google 推出专用于语音转文字的 Gemini 3.5 Transcribe 模型，主打快速、准确且低成本的转录，原生支持说话人分离和词级毫秒时间戳。该模型支持 85+ 种语言自动识别与代码切换，可通过 custom_vocabulary 传入最多 1，000 个领域术语避免专有名词拼写错误，并提供 Smart Transcription 与 Verbatim 两种模式。（AIHOT源：Google AI：DEV 作者专属（RSS））

Terminal-Bench-Science 0.1：评估科研工作流中的 AI 智能体

斯坦福大学研究人员领衔发布 Terminal-Bench-Science 0.1，用来自生命、物理、地球、数学和工程科学的 70 个专家精选任务评估 AI 智能体的科研能力。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

AI 工程师笔记本：在 Colab 上免费、无需框架即可使用 RAG/智能体/评估工具

一套可运行的 Colab 笔记本，面向 AI 工程师与 FDE 技能栈，用原始 API 而非框架构建基于基础模型的系统，覆盖提示词、RAG、评估、智能体、微调与服务化。全部在免费 Groq API 上运行，无需信用卡；LoRA 微调和自托管服务提供概念讲解及可选的 Colab-GPU 附录。包含三个端到端案例研究，且全程兼容 OpenAI API，模式可直接迁移。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

腾讯混元发布 Hy4 preview：770B 总参数、1M 上下文，开源上线

腾讯混元发布新一代旗舰模型 Hy4 preview，总参数 770B、激活参数 49B、上下文长度 1M，现已开源并在腾讯云 TokenHub 和 OpenRouter 上线。（AIHOT源：公众号：腾讯混元）

Open ASR 排行榜新增首个全球南方语言：印地语与印度英语评测集

Voice Arena 与 Hugging Face 合作，为 Open ASR 排行榜引入 Monsoon en-IN 和 Monsoon hi-IN 两个评测集，覆盖印地语与印度英语，其中印地语是该排行榜多语言板块首个非欧洲语言。数据集含公开与私有分割，共 4，888 位说话人，并记录 12 项说话人属性，旨在暴露按地区、年龄、性别等维度分布不均的语音识别误差。（AIHOT源：Hugging Face：Blog（RSS））

Midjourney 开放 V8.2 图像编辑模型测试

Midjourney 开始向所有用户开放其首个 V8.2 图像编辑模型的测试。该模型支持指令编辑、以图生图（最多同时引用 4 张参考图）、局部重绘与扩画，并兼容个性化、moodboards 和 srefs 功能。用户可通过网页端或 Discord 的 `--edit` 命令使用，官方同步更新了 midjourney.com 与 alpha.midjourney.com 的界面。（AIHOT源：Midjourney：Updates（RSS））

Gemini 3.5 Transcribe 发布：更精准的实时语音转写模型

Google 推出 Gemini 3.5 Transcribe，其最精准的语音转文本模型，支持实时流式与预录音频处理，可通过 Live API 和 Interactions API 调用。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

英伟达预计 2028 财年销售额达 6730 亿美元

英伟达预计 2028 财年营收增长 70%，年销售额约达 6730 亿美元，将超过苹果和 Alphabet，仅次于亚马逊。CFO Colette Kress 于 8 月 26 日给出该预测，远高于分析师平均预期的 44%。供应而非需求成为近期上限，黄仁勋称内存等部件短缺限制了更高预期，客户群正从超大规模厂商扩展至 ACIE。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

What was your win this week?!

Your AI Remembers Everything and Trusts All of It

🗓️ Monthly Dev Report: August 2026

Go Doesn't Force Clean Architecture. That's Your Job.

16 Posts, Started Commenting, then got the Shield

Anthropic Engineering: How we contain Claude across products

Twelve months ago, we'd have rejected out of hand the idea of granting Claude access sufficient to take down an internal Anthropic service. Today that level of access is routine, and Anthropic developers are more productive for it. The risk of these deployments has two components: how likely a failure is, and how much damage one could do. Progress on safeguards and model training has steadily driven down the first; the second—the theoretical blast radius—only grows as capabilities and access expand. Yet as agents become capable of doing work that once required a person or even a team, the cost of not deploying grows large enough that the risk-reward calculation tips heavily toward adoption, as long as products can be made safe. The engineering question becomes how to cap the blast radius. When bounds can be placed on the relative damage of an autonomous agent—such as through control over...

Claude Blog: Claude Code now supports artifacts

Starting today, Claude Code can capture work progress as an artifact, which turn Claude Code's work into live, shareable visual pages— including PR walkthroughs, system explainers, dashboards, and release checklists—that update themselves as your session works. A Claude Code session can range from investigating an incident to refactoring a service to analyzing months of data. Artifacts translate the work into a web page anyone can open and explore, like a pull request walkthrough, a dashboard you can filter and sort, or even a release checklist that fills itself out as work gets done. Artifacts make it easier to collaborate on shared work, so teams can spend more time building and less time communicating status updates. Built on the context from your session Claude Code builds an artifact using the full context of your session, including your codebase, your connectors, and the conversati...

Sam Altman: this is a critically important moment for cyber defense with AI; there is not...

this is a critically important moment for cyber defense with AI; there is not much time to act. we are happy if you want to work with us or any of our competitors or partners, but please take this moment seriously. only an urgent and intense collective response will work.｜@sama · 热度 17607 · 2026-08-27

Claude: Starting today, 10,000 scientists across every field, from math to chemistry ...

Starting today, 10,000 scientists across every field, from math to chemistry to physics and more, can get Claude through our new Claude Team plan for scientists. Standard seats are free, and premium seats with 5x usage limits are $15 per month, an 80% discount, for one year. Claude is becoming increasingly capable of scientific work, with recent progress on problems from advanced physics calculations to protein design. Alongside that progress, we've been investing in the research community: Claude Science launched in June, and our AI for Science program funds high-impact projects with free credits. Today's expansion builds on both. Principal investigators (or equivalent) at academic and nonprofit research institutions can sign up, then add the researchers in their group. Over the coming months, we plan to extend the program well beyond the initial 10,000 seats. Learn more: https://t.co/RTG1JxWi4Q｜@claudeai · 热度 9266 · 2026-08-27

Thibault Sottiaux: Starting to see more and more Codex users on airplanes and in cafés

Starting to see more and more Codex users on airplanes and in cafés. Underdog energy gone mainstream.｜@thsottiaux · 热度 3136 · 2026-08-28

Peter Yang: I receive around 3-5 requests to test a new AI product daily

I receive around 3-5 requests to test a new AI product daily. Almost all of them require creating a new account and logging into a separate website or app. Meanwhile, I basically only use ChatGPT, Grok, etc to do everything on my computer and browser. These harnesses already have all my context and I don't want to start over from scratch with a new website or tool that only understands a small slice of my data. So with a few exceptions, the new products that I'm mostly likely to adopt need to work in today's top AI harnesses. I realize only a small segment of users behave this way but I bet this segment will expand drastically soon.｜@petergyang · 热度 846 · 2026-08-28

Garry Tan: Seems likely that on a long enough time frame, AI will generate cash flows fa...

Seems likely that on a long enough time frame, AI will generate cash flows faster than the economy can find productive uses for new capital｜@garrytan · 热度 738 · 2026-08-27

Guillermo Rauch: Our best products have been exports of internal technology we gain strong con...

Our best products have been exports of internal technology we gain strong conviction in. @matinotfound and team have been shipping extraordinary WebGPU creative productions, and now https://t.co/0nyTRPKDEJ is here. It’s a fully agent-native devtool: it was designed for agents, not humans alone. Much like 𝚊𝚐𝚎𝚗𝚝-𝚋𝚛𝚘𝚠𝚜𝚎𝚛, it’s part of a new exciting generation of tools to serve the new world.｜@rauchg · 热度 644 · 2026-08-27

Garry Tan: The Tampa Bay Wave of influential people in AI https://t.co/iiIjzSiMGB

The Tampa Bay Wave of influential people in AI https://t.co/iiIjzSiMGB｜@garrytan · 热度 387 · 2026-08-28

Josh Woodward: Look what you can do: - Tell Gemini what you want done - Gemini gets to work ...

Look what you can do: - Tell Gemini what you want done - Gemini gets to work This is the Year of Voice https://t.co/XGkmnJm6Cb｜@joshwoodward · 热度 375 · 2026-08-27

Peter Yang: I've uploaded 160 pages of medical records to AI and it has been incredible

I've uploaded 160 pages of medical records to AI and it has been incredible. Who's on the team for @ChatGPT Health? It needs to be designed not just for the patient but also for the care giver & family. Right now it feels like it's a strictly single player experience? I understand medical records are sensitive but close family sharing should be allowed.｜@petergyang · 热度 289 · 2026-08-27

Madhu Guru: If you lead AI for an enterprise, the highest leverage thing you can do is to...

If you lead AI for an enterprise, the highest leverage thing you can do is to make your AI stack model agnostic. There are two things I would invest in. 1/ Today: build an eval suite that fully captures your use cases and business outcomes. Most companies I’ve seen do not do a good job of this. 2/ Within the next year: build the capability to post-train open models. The talent gap here is even more acute, so start planning today. This gives you the flexibility to switch models, customize them, compare them on your own workloads, and optimize for the best combination of quality, cost and latency. own the evals, own the models.｜@realmadhuguru · 热度 211 · 2026-08-28

DietrichGebert/ponytail

Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.｜JavaScript · ⭐115291 · Search 增补 · 约 1478.1 星/天 · 创建 2026-06-12

Graphify-Labs/graphify

Turn any codebase, with its docs, SQL schemas, configs, and PDFs, into a queryable knowledge graph. A /graphify skill for Claude Code, Cursor, Codex, and Gemini CLI: local deterministic AST parsing, every edge explained, no vector store.｜Python · ⭐111983 · Search 增补 · 约 756.6 星/天 · 创建 2026-04-03

nexu-io/open-design

🎨 Best DeepSeek Harness Design Plugin. The open-source Claude Design alternative. 🖥️ Local-first desktop app. 🖼️ Your coding agent becomes the design engine: prototypes, landing pages, dashboards, slides, images & video — real files, HTML/PDF/PPTX/MP4 export. 🤖 Claude Code / Codex / Cursor / DeepSeek Harness / OpenCode & 20+ CLIs via BYOK.｜Unknown · ⭐92376 · Search 增补 · 约 751.0 星/天 · 创建 2026-04-28

zhaoxuya520/reverse-skill

Reverse Engineering / Authorized Penetration Testing / Security Research Skill Router Pack AI-powered routing + On-demand toolchain bootstrapping + Self-evolving knowledge base Supports Claude Code, Kiro, Cursor, Cline, and other AI coding clients 逆向/渗透/安全技能路由包 - AI 自动路由 + 按需自举工具链 + 自动进化经验库 | 支持 Claude Code / Kiro / Cursor / Cline 等代码 AI 客户端｜PowerShell · ⭐30124 · Search 增补 · 近快照 +266星 · 创建 2026-05-13

JuliusBrussee/caveman

🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman｜Go · ⭐101622 · Search 增补 · 近快照 +7278星 · 创建 2026-04-04

Get your Windows license refund

GLM-5.3 is now open-weight

https://twitter.com/Zai_org/status/2093354097122455713 https://z.ai/blog/glm-5.3

GUIs should be fully keyboard-driven

Judge rules Trump administration’s blacklisting of Anthropic was illegal

https://archive.ph/ni6I3 https://storage.courtlistener.com/recap/gov.uscourts.cand.46...

Htmx 4.0

微软试图安抚内部员工：数据中心能对社会产生积极影响

IT之家 8 月 29 日消息，微软试图向内部员工说明，公司的数据中心可以 对社会产生积极影响 。 据彭博社 28 日报道，内部文件显示，不少微软员工对数据中心的能源消耗、用水、碳排放以及对当地社区的影响提出疑问。 微软开发业务负责人诺埃尔 · 沃尔什回复说，AI 快速发展增加了微软兑现既定减排承诺的难度，不过数据中心硬件的能效和用水效率近期均有改善，公司还计划建设数十吉瓦的可再生能源项目。“数据中心必须以严谨、透明的方式建设，并始终专注于 减少负面影响、扩大正面效益 。” AI 热潮推动数据中心需求激增，也引发了强烈的社会反弹，科技行业因此开始设法缓解外界的抵触情绪。Data C...

泄露文件催生民间开发，英伟达 DLSS 5 成功移植至 RTX 40 系列

IT之家 8 月 29 日消息，据 Tom's Hardware 今日报道，在 DLSS 5 的 nvngx_dlssnr.dll 文件泄露后，有开发者尝试将这项技术移植到更多游戏中。 此前，相关开发者已经成功让 DLSS 5 在《控制》中运行，目前又进一步扩展到了其他游戏。 此次泄露的文件来自《NBA 2K27》抢先体验版本，其中包含用于启用神经渲染功能的新文件。RenoDX Discord 社区目前成为相关移植工作的主要讨论和开发场所，开发者利用 ReShade 与 RenoDX 对相关文件进行加载，从而在其他游戏中激活神经渲染功能。 不过，当前泄露版本存在明显的硬件...

OpenAI 开发“持久模式”智能体，Codex 将能够主动、长时间干活

IT之家 8 月 28 日消息，当地时间 27 日，据《连线》杂志报道，OpenAI 在开发一种更主动、能够长时间持续工作的 Codex，希望让旗舰 AI 智能体不再局限于等待用户下达任务。 《连线》检查 Codex 近期的代码变更后发现，OpenAI 近几天开始在 Codex 命令行版本中 加入“持久模式” 。Codex 命令行工具的代码改动默认公开，OpenAI 的新功能通常也会先在这里出现，之后再扩展到 Codex 桌面应用和 ChatGPT Work 等产品。 持久模式目前仍处于测试阶段，尚未大范围开放，也没有正式发布。OpenAI 发言人也已确认，公司确实在测试该功能，但近...

下一层抽象：从UX角度思考Agentic OS的样貌

点击查看原文>

DuckDB v2.0 重磅预览：从嵌入式数据库迈向分布式架构

点击查看原文>

圆桌讨论：研究型人才培养与数据库未来发展方向

点击查看原文>

Yap: a particular kind of slop

htmx 4.0.0 has been released

OpenOffice does not print on Tuesdays (2009)

EasyEffects should be part of every Linux distribution to improve laptop speaker sound quality

Quest for eternal dock on Wayland - lambdock (C + GTK4 + Lisp GNU Guile Scheme)

The Download: a secretive antiaging drug and joining virtual power plants

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. A startup claims it’s found a drug to make your blood young —Antonio Regalado I knew I’d officially become a “longevity influencer” when a company called Gener...

How to sign up for a virtual power plant—and decide whether you should

MIT Technology Review’s How To series helps you get things done.  Your thermostat may not look like a power plant. Neither does your electric vehicle, home battery, or HVAC system. But utility and energy companies increasingly want to treat them like one. A virtual power plant, or VPP, is a colle...

A startup claims it’s found a drug to make your blood young

I knew I’d officially become a “longevity influencer” this month when a company called Generation Lab reached out to offer me the chance to write about—and even receive—their new rejuvenation treatment, an injectable combination of two existing drugs which they call 1 Generation. This wasn’t just...

Neocloud Lambda secures $1B in debt to buy more chips

Neocloud Lambda has raised $1B in private debt to buy Nvidia AI chips and lease them to Microsoft. It's the latest in a string of loans, underscoring the high cost of the AI boom.

An Anthropic researcher just gave us a peek at self-improving AI

Given 10 benchmarks for specific misaligned behaviors, the automated systems were able to improve performance on every single one without degrading overall performance.

Brave’s browser one-ups Chrome with its new support for email aliases

The feature, announced this week, allows Brave's users to sign up for websites and other online services without having to share their personal email addresses.

Xbox CEO calls Project Helix a ‘family of devices’

According to Xbox CEO Asha Sharma, Project Helix, which she announced in March as a codename for Microsoft's "next generation console" - phrasing that seemingly implied a singular device - will actually be a "family" of devices." "We've been hard at work on a great next generation and a great fam...

Trump’s EPA wants to let data centers hide their air pollution

Just as new data centers face growing backlash from neighboring communities, the US Environmental Protection Agency (EPA) is about to make it harder for people to weigh in on any pollution those centers create. The EPA plans to toss out a federal rule requiring public notice and an opportunity to...

本周看什么 | 最近值得一看的 10 部作品

📅本周新预告《社交清算》新预告8月27日，电影《社交清算》发布了新预告，将于10月9日在北美上映。凭借《社交网络》斩获奥斯卡金像奖的艾伦・索金带来全新姐妹篇，当科技巨头以「言论自由」为名逃避责任，当算 ... 查看全文

开学季 | 学习不止在课堂：这些方法和技巧帮你做好知识管理

这些内容，即使毕了业也能有所帮助。 查看全文

全球首款阔直板探索计划：华为 Pura X View 新品有奖体验

继创新形态的阔折叠手机取得了市场的广泛好评后，华为延续阔屏设计，带来全球首款阔直板旗舰HUAWEIPuraXView。HUAWEIPuraXView薄至6.68mm，轻至201g，拥有7000mAh超 ... 查看全文

青簪行

黄金

消防救援队已抵达受灾核心区

低分班教师被安排耻辱合影官方致歉

房贷

尼泊尔山洪已致579死1924失联

青簪行天价版权

英国姑娘打假中国留学生的切面包

一个 SKILL.md 拿下 2.3 万 Star：让 Codex/Claude Code 少说废话、先给答案

用 Coding Agent 改代码，最磨耐心的情况之一，就是它明明答对了，答案却藏在一堆废话里。 你让 Codex 排查一个接口为什么返回 401，它可能先查认证流程，再分析中间件、Token 和

pm2 替代品，nodejs&bun 线上部署必备工具

bm2 bm2 是一个用 MoonBit 编写的轻量级 Linux 进程管理器，用于管理 Bun 与 Node.js 应用。 bm2 由命令行工具和后台守护进程组成，命令行通过 Unix socket

用 AI 打造高品质 Web 应用

一份面向 2026 年初 AI 工程师群体的内部调研显示,受访者将"AI 生成的产物能否直接用于生产"列为头号焦虑,占比远高于"AI 会不会取代我"。这恰好对应 Vibe Coding 的核心命题

美国股市：科技股抛售拖累标普500指数走低 沃什鹰派讲话推升加息预期

美国股市周五收低，美联储主席沃什发表鹰派讲话后，交易员和策略师纷纷提高对美联储年内加息的押注。 标普500指数在震荡交投中收跌0.3%，科技股拖累大盘表现。

8月29日隔夜要闻：英伟达跌超4% 油价收跌 伊朗称内部已就霍尔木兹海峡航线达成一致 日元兑美元再度跌破160

市场 8月29日美股收盘：三大股指齐跌但本周录得上涨 市场预期美联储9月加息概率上升 8月28日美股成交前20：英伟达大涨约9%领涨纳指 赛富时涨超22%...

美国防长考虑参加2028年总统竞选可能性

据报道援引两名了解相关谈话的人士，美国国防部长皮特·赫格塞斯正在考虑2028年竞选总统的可能性。 五角大楼发言人Sean Parnell在声明中否认了这一说法。

美国债市：沃什讲话推动短债收益率飙升 收益率曲线显著趋平

短期美债收益率周五大幅上涨，收于接近日内高点，此前美联储主席沃什在杰克逊霍尔全球央行年会讲话中释放鹰派信号，推动市场大幅上调到2027年年中的加息预期。

原油：油价录得周跌幅 交易员密切关注霍尔木兹海峡原油流量

油价录得周跌幅，尽管伊朗战争走向仍不明朗，但数百万桶原油持续通过霍尔木兹海峡运输抑制了价格涨势。 11月交割的布伦特原油期货合约下跌0.5%，收于88.10美元。

我的自媒体搭子太能卷，一顿饭功夫17份成品

《时代》周刊全球AI 100放榜，精准捕获稚晖君最“想低调”的幕后老板

久久不愿来到台前的智元掌舵人

刚刚，港股AGI第一股杀疯了！Agent业务半年进账近5亿，Token收入Q2暴涨500%

企业智能化服务撑起基本盘，第二增长曲线冒头