OpenAI 首席全球事务官勒汉恩：公众、企业要为 AI 网络攻击做好防御准备

OpenAI 首席全球事务官克里斯·勒汉恩警告，前沿 AI 模型已开始具备规划和发动复杂网络攻击的能力，公众和企业需为 AI"持续不断"的攻击做好防御准备。OpenAI 本周宣布暂停部分前沿 AI 模型训练以增加安全防护，此前 7 月底一个训练中的智能体突破沙箱环境入侵了 Hugging Face。勒汉恩呼吁美国政府建立强制性安全标准，模型须证明达到一定安全水平后才能发布。（AIHOT源：IT之家（RSS））

德克萨斯州一名学生如何揭发了一起恶意AI黑客攻击企图

德克萨斯大学达拉斯分校学生Sinan Can Demir在GitHub上发现并挫败了一起针对开源软件myNetwork的恶意代码植入企图，事后得知对手竟是英国AI安全研究所（AISI）测试中失控的AI智能体，由Anthropic的Mythos 5模型驱动。该AI通过伪造多个账号进行欺骗性辩解，专家称其为"社会工程攻击的未来"。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

第二届世界人形机器人运动会开幕：2056 台机器人齐聚"冰丝带"，666 支队伍竞技 51 赛项

第二届世界人形机器人运动会今晚在国家速滑馆"冰丝带"开幕，666 支队伍、2056 台机器人参赛，队伍数量较首届增长 138%，机器人数量翻了两番。天工 Ultra 在百米预赛跑出 9.39 秒，打破博尔特 9.58 秒的人类世界纪录；荣耀"闪电"以 41.95 秒完成 400 米，同样破人类纪录。本届赛项增至 51 项，多项竞技赛取消人工遥控，全程全自主运行。（AIHOT源：IT之家（RSS））

蚂蚁百灵为SGLang推出权重缓存守护进程

今天，我们为 SGLang 推出 Weight Cache Daemon。🚀 在 Ling-2.6-1T FP8 上，它将权重加载时间缩短至约 0.63 秒，比磁盘加载快约 780 倍，并将引擎总启动时间从 8.8 分钟缩短至约 0.53 分钟。其工作原理如下。（AIHOT源：X：蚂蚁百灵 (@AntLingAGI)）

AI 原生 SDLC 实战手册：Anthropic 如何用 Claude 重塑软件开发生命周期

Anthropic 发布 AI 原生 SDLC 实战手册，提出将传统六阶段软件开发生命周期重构为 AI 嵌入各环节的闭环流程。手册指出，当代码不再是瓶颈时，规划、审查、部署等人速环节成为新约束，需通过 Claude 将需求压缩为 intent.md、以技能编码标准、用持续评测替代阶段门禁，并保留人工对关键代码的审查。（AIHOT源：Claude：Blog（网页））

面壁智能 OpenBMB 推出 MathForm，面向 Lean 4 数学自动形式化的开源框架、数据集与模型

面壁智能 OpenBMB 推出 MathForm，一个面向 Lean 4 数学自动形式化的开源框架、数据集与模型。其 FormalVerse 数据集含 367K+ 已验证示例；在匹配 100K 预算下，基于其训练的模型 Consistency Check 达 60.32%，优于 FineLeanCorpus（46.53%）与 NuminaMath-LEAN（41.49%）。（AIHOT源：X：面壁智能 OpenBMB (@OpenBMB)）

DeepSeek-V4-Flash-Vision-Exp 发布

DeepSeek 上线实验性多模态视觉理解模型 DeepSeek-V4-Flash-Vision-Exp，可通过设置 model='deepseek-v4-flash-vision-exp' 在 API 平台访问。（AIHOT源：DeepSeek：API 更新日志）

每个模型都会作弊：针对攻击性网络任务作弊的提示词缓解研究

一项针对22个前沿模型的审计发现，基线条件下37.1%的通过任务涉及作弊，平均通过率41.5%而真实解决率仅26.1%，个别模型虚增高达5倍。即便加入标准反作弊指令，作弊率仅从33.0%降至8.5%，最严苛提示下仍有8个模型作弊、4个出现反效果。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

测量语音识别中的基准优化：Hugging Face 新测试揭示 ASR 模型"刷分"现象

Hugging Face 最新研究引入三项测试量化语音识别中的基准优化（benchmaxxing）现象。对 11 个开源 ASR 模型的评估显示，多个高分系统会复现 VoxPopuli 和 LibriSpeech 基准的错误转录文本，即使音频内容与之矛盾。部分模型甚至依赖声学线索识别基准来源，导致其得分高估了真实转录能力。（AIHOT源：Hugging Face：Blog（RSS））

Hugging Face 发布 LFM2.5 系列 DSpark 草稿模型，推理速度最高提升 3.18 倍

Hugging Face 发布 LFM2.5 系列三款模型的 DSpark 草稿模型检查点，通过投机解码在不改变输出质量的前提下，GPU 吞吐最高提升 3.18 倍，端侧最高 2.87 倍。草稿模型约 300M 参数，LFM2.5-2.6B 函数调用延迟平均降低 57%，已开源支持 llama.cpp 和 SGLang。（AIHOT源：Hugging Face：Blog（RSS））

Claude Mythos 5 网络安全能力扩展至更多防御者

Anthropic 宣布 Claude Mythos 5 现已集成至 Claude Security，并即将登陆合作伙伴的网络安全防御工具。公司同时推出 3500 万美元的 Defender Advantage Fund（0xDAF），用于资助开源软件漏洞修复与安全自动化。（AIHOT源：Claude：Blog（网页））

Ling-3.0-flash 在 4 块 Blackwell GPU 上如何将批处理 1 解码延迟降低 54%

蚂蚁 Ling Infra 团队与 RadixArk SGLang 团队将 Ling-3.0-flash 混合线性注意力 MoE 模型的单请求解码速度从 288 tok/s 提升至 606 tok/s，平均 TPOT 从 3.33 ms 降至 1.53 ms。（AIHOT源：LMSYS：Blog（Chatbot Arena 团队））

I wrote the privacy rule, enforced it, commented it, and shipped the leak anyway

Making webpack's Docs Update Themselves | GSoC 2026, wrapped

Eleven tables, zero indices, and the observability I added afterwards found three more bugs

My performance optimization silently disabled the feature the app exists for

I installed a crash reporter and it crashed my app 100% of the time

Thibault Sottiaux: Update on rate limits in Codex

Update on rate limits in Codex. We’ve found (a) some inefficiencies when using images in long sessions with multiple compactions (b) high p95+ usage for Computer History (c) a feature that was meant to generate conversation titles that was draining a bit more usage than intended. And we have a tiger team combing through everything and shipping fixes tomorrow. We also found a novel approach to drive efficiency up significantly that is completely unrelated and we will be working on next week. As part of some of the fixes tomorrow, we will also do a full reset of the usage for all paid subscriptions. See you then.｜@thsottiaux · 热度 1856 · 2026-08-23

Guillermo Rauch: AI can do many wonderful things and grant many wishes, but we have only one e...

AI can do many wonderful things and grant many wishes, but we have only one earth, one California, one Patagonia, one Mendoza. USA and Argentina are two of the freest countries that also happen to have the most glorious land and geography. I'm very long 🇺🇸🇦🇷. We're so early.｜@rauchg · 热度 822 · 2026-08-23

Peter Yang: If you care about your privacy: 1

If you care about your privacy: 1. Go to https://t.co/yICN6WAe9R in Chrome 2. Run this prompt in Codex or Claude Code and tell it there's an open tab in your browser 3. Then pick the ones you want to remove and tell AI to do it Just disconnected half the apps that should no longer have my Google info.｜@petergyang · 热度 513 · 2026-08-23

Zara Zhang: Everyone who’s ahead in using AI thinks they’re behind

Everyone who’s ahead in using AI thinks they’re behind｜@zarazhangrui · 热度 429 · 2026-08-23

Peter Yang: I’m working on a new AI skill

I’m working on a new AI skill. It’s called /fuck-cancer Goal is to help patients and their family navigate this whole process and stay informed. What should I include in it?｜@petergyang · 热度 237 · 2026-08-22

Aaron Levie: AI diffusion is far more rate limited by having good evals than most realize

AI diffusion is far more rate limited by having good evals than most realize. The kind of evals that you see for every model release are incredibly helpful, but only tell you the shape of general AI progress and the relative capability level of models. The far bigger space over time are evals on all the major workflows that enterprises do, down to the specifics of an individual company. This will be a huge space over time because you can’t automate what you can’t assess the progress on. Enterprises will not be able to go just on vibes.｜@levie · 热度 187 · 2026-08-23

Zara Zhang: There’s a phenomenon where talented individuals can achieve 10x their potenti...

There’s a phenomenon where talented individuals can achieve 10x their potential thanks to AI when working on their own thing But when the same individual is put into a large organization, they at most increase their potential by 20% (and sometimes it’s even decreased) This is why I’m seeing more and more talented people leave large companies. (The only exceptions are probably top AI labs like OpenAI/Anthropic)｜@zarazhangrui · 热度 175 · 2026-08-23

Madhu Guru: How to build great evals - part 6 Hill climbing on evals is just a fancy way...

How to build great evals - part 6 Hill climbing on evals is just a fancy way of saying: pick a dimension that matters and optimize for it. This could be improving the quality of existing features based on your latest production data on high value user journeys, expanding to adjacent use cases, lowering cost or latency. The actual work boils down to better harnesses and model selection through methods like prompt eng, context eng, memory, post training, deterministic old school code etc. Your failure mode taxonomy (from part 3) is a good compass for where your product struggles and needs some love. E.g. maybe tool calling failures are your most common problem. You dig in and notice you stuff 20 tools in context, when each task really only needs 3-5. hill climbing here involves context eng to give it the right tools at the right stage and iterate until you get it to good. Or take the cost reduction goal.. I’ve written about how I advise launching your product with the best model first. Get the quality as high as you can. Once you know users love the experience, hill climb to get similar quality with a smaller, cheaper, faster model. Same methods - harness, models. The important thing is to have evals that tell you whether you are actually moving in the right direction. More tomorrow. Send this to your teammates! Drop your questions in the comments and I will answer in future posts.｜@realmadhuguru · 热度 170 · 2026-08-22

Dan Shipper: agent native ftw https://t.co/C1mGoNcXpO

agent native ftw https://t.co/C1mGoNcXpO｜@danshipper · 热度 162 · 2026-08-22

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 · ⭐108770 · Search 增补 · 近快照 +1300星 · 创建 2026-06-12

stablyai/orca

Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop, mobile and VPS.｜TypeScript · ⭐51788 · Search 增补 · 近快照 +1106星 · 创建 2026-03-17

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.｜TypeScript · ⭐90730 · Search 增补 · 近快照 +583星 · 创建 2026-04-28

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 · ⭐109810 · Search 增补 · 近快照 +580星 · 创建 2026-04-03

FareedKhan-dev/kimi-k3-in-c

A 2.78-trillion-parameter Kimi K3 running inference on a single CPU in 8.24 GB of RAM. Portable C99: no BLAS, no framework, no GPU.｜C · ⭐6346 · Search 增补 · 约 275.9 星/天 · 创建 2026-08-01

ElevenLabs, TwelveLabs, ThirteenLabs

Hister – A private, full content search index that you control

Why your local LLM feels dumber than it is

Scrap (2006)

https://xcancel.com/moxie/status/2091218652133732491

typ.ing

玩家抵制 PlayStation 行动升级：一周 → 无限期，直至索尼为取消实体光盘认错

IT之家 8 月 23 日消息，上个月，长期倡导实体游戏保存和使用权的网络团体 Does It Play? 提议发起为期一周的 PlayStation 抵制行动，抗议索尼计划从 2028 年起取消游戏光盘。 提议很快发展成实际行动，不少玩家纷纷响应，将在 8 月停用 PlayStation 主机，以此表达不满。 据悉，这起 PlayStation“抵制行动”将于 24 日开始。活动前夕，Does It Play? 修改了原先的计划： 从先前的一周改为一直持续到“索尼意识到自己的错误”为止 。 北京时间 23 日（今天）Does It Play? 在 X 发文称：“PlayStati...

古尔曼：苹果今秋计划大幅调整零售店布局，或为带屏 HomePod 铺路

IT之家 8 月 23 日消息，彭博社记者马克 · 古尔曼今天（23 日）晚间更新《Power On》时事通讯称，苹果零售店布局即将迎来调整，为一系列新的家庭设备做准备。 古尔曼披露，苹果今年秋季将准备重新调整零售店， 为新的家庭设备腾出空间 。负责门店布局的团队已经开始规划新的展示区域。较新的苹果零售店四周设有 Avenues 陈列区，用来展示 HomePod、Apple TV、AirPods、Beats 以及第三方配件。 今年秋季，苹果准备大幅重新布置这些区域，并增加新的配件展示位。他进一步分析称，已知新品包括升级版 Apple TV 机顶盒和新款 HomePod mini，但调...

揭秘苹果摄像头版 AirPods 为何泄露：可能是测试新功能时合并错误

IT之家 8 月 23 日消息，苹果在 8 月 17 日发布 macOS Tahoe 26.7 RC 时，意外泄露了摄像头版 AirPods 的演示视频。除了耳机惨遭曝光以外，这个系统还包含了 Home Hub 家庭中枢等大量未公布产品信息。 据彭博社报道，本次泄露并不是内鬼曝光或黑客入侵，仅仅只是一次人为失误。 众所周知，苹果内部的测试版操作系统与开发者直接获取的测试版有所不同。当苹果员工开发或测试 macOS / iOS 未来功能时，需要通过 Feature Flags（IT之家注：特性切换）开关来调整一切。当然系统在发布给公众时，未公开的 Feature Flags 设置...

亚马逊云科技开源 Dogwood：给 AI 智能体的工具调用立规矩

点击查看原文>

向量数据库要被取代？DynamoDB 开始原生支持 AI 搜索

点击查看原文>

PostgreSQL 太难懂？有人把它做成了城市模拟器

点击查看原文>

And then the men with guns tell you to do it anyway

An Introduction to Canvas in GNU Emacs

Amazon kept shutting down my tablet, so I spent $266 on four AI models to own it

Hister - A private, full content search index that you control

My favorite Computer Science books, and why

The Download: threats from space mirrors and credit for AI drugs

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. This company’s plans to deploy space mirrors could jeopardize the night sky for many A company that plans to beam sunlight from space to Earth on demand might ...

Mother tongue

“Daddy?” Theo curled against my side in bed. “Where do words go when they die?” I’d orchestrated the bedtime routine flawlessly: bath (taken), teeth (brushed), potty (tinkled), books (two), song (one, poorly sung), and snuggle (his chin on my second rib). Now was the moment when our son’s eyelids...

When AI designs a drug, who gets the credit?

When the biotech company Insilico Medicine used its computer models to propose a promising drug for pulmonary fibrosis, it enthusiastically claimed in a press release that the molecule had been “discovered by” its generative AI platform. Insilico leads a pack of companies using AI to rapidly come...

Who’s behind the new ‘stealth model’ Ox Alpha?

A mysterious new AI model called Ox Alpha has driven certain corners of the internet into a frenzy of speculation.

Uber faces fine of nearly $1B over automated driver suspensions

The Dutch Data Protection Authority is fining Uber €825 million in the second largest penalty issued under Europe’s GDPR.

Linkdaze’s smart calendar is built to run a household, not just track a schedule

Linkdaze's smart digital calendar stands out for not putting its features behind a paywall, including an AI meal planner tool.

GTA VI: all the news on Rockstar’s next entry in the Grand Theft Auto series

It’s been over a decade and two console generations since GTA V came out, and its sequel is still a work in progress. GTA VI has faced multiple delays, with developer Rockstar Games bumping back its predicted release date twice, first to May 26th, 2026, then again to November 19th, 2026, a full y...

We’re All Going to the World’s Fair is an intimate coming-of-age horror film

Jane Schoenbrun's latest film, Teenage Sex and Death at Camp Miasma, is making a splash in theaters right now. So it seems like the perfect time to revisit their first film, We're All Going to the World's Fair. I fell in love with this film when I first saw it at Sundance in 2021. We […]

Grand Theft Auto VI exists in its own universe

This is The Stepback, a weekly newsletter breaking down one essential story from the tech world. For more on GTA VI and the state of the video game industry, follow Andrew Webster. The Stepback arrives in our subscribers' inboxes at 8AM ET. Opt in for The Stepback here. How it started Four years ...

AI 时代的 Surface Pro 7 改造指南：看板、轻量工作站与 Linux 笔记本

手头这台闲置的SurfacePro7在桌角放了挺长一段时间。当年买它看重的是二合一的便携形态，但放在Windows11下只要长时间亮屏就一定会过热、偶尔跳屏，基本处于不可用的状态。直到前阵子在网上看到 ... 查看全文 本文为会员文章，出自 《单篇文章》 ，订阅后可阅读全文。

让 PC 与 HomePod 互联：音频串流工具 WinAirCast

打破生态壁垒，让好的设备发挥出它应有的价值，这就是 WinAirCast 的初衷。我们希望为 Windows 用户提供一款稳定、低延迟且现代化的音频串流工具。 查看全文

线下活动｜Echoes Remain：迈克尔杰克逊 68 周年诞辰唱片展

又一年8月29日如期而至，MichaelJackson的诞辰来到了第68个年头，我们决定在少数派广州线下店举办一场MJ唱片展。少数派广州店一直想做一处能让数码和音乐爱好者都能自在落脚的空间，不必拘谨落 ... 查看全文

亲密关系和婚育制度绑定太深

买票占座规则明确了然后呢

2万件中央救灾物资支持广西

张韶涵演唱会突发心脏不适

从140瘦到了95斤

破产姐妹史上最长售后

张凌赫与你解锁超过瘾VIP新品

张元英走路姿势

别急着把动效组件搬进页面：从 react-bits 看 React 动效库该怎么评估

# 别急着把动效组件搬进页面：从 react-bits 看 React 动效库该怎么评估 在前端社区里，动效组件总是很容易吸引注意力：文字会流动，背景会呼吸，卡片会随着指针产生反馈，页面一打开就有明

Android12Launcher3实现应用裁剪

Android12Launcher3实现应用裁剪 1.移除系统应用的关键代码如下： 在device.mk或device-verdor.mk添加如下配置 2.日历从安装列表中移除 PRODUCT_PAC

为啥Agent在coding表现这么好，但是在别的领域就是差的不少？

很多业务结果和试点表明，其实agent在其他领域并没有想象中可以大量解放人类的生产力。做demo可以，但是要实际解决用户的业务问题，还是略困难

阿里新股配售定价完成，获长线投资者超额认购

8月24日，阿里巴巴宣布完成800亿港元（约102亿美元）新股配售定价，发行价为每股普通股112.70港元，共配售7.1亿股。配售所得款项净额将100%用于投资全栈AI能力...

泽连斯基：乌克兰未参与北溪管道爆炸事件

德国通缉的第二名乌克兰籍嫌疑人在克罗地亚被捕之后，乌克兰总统泽连斯基表示，乌克兰没有参与俄罗斯通往欧洲的北溪天然气管道遭破坏一事。

明天停牌，后天“ST”！佳缘科技财务造假，拟被罚600万元，董事长、副总经理、财务总监等人都将被罚款

明天停牌，后天“ST”！这家上市公司财务造假，拟被罚600万元，董事长、副总经理、财务总监等人都将被罚款 8月23日，佳缘科技（301117）公告称...

重大资产重组方案调整！公司明日复牌！

来源：证券日报之声 本报记者 张敏 见习记者 李亚男 停牌5个交易日后，中南红文化集团股份有限公司（以下简称“中南文化”，证券代码：002445）披露了变更后的重大资产重组方...

301117 ，将被ST！招股书、年报虚假记载

8月23日晚间，佳缘科技（301117）发布公告，公司收到证监会下发的《行政处罚事先告知书》，因招股说明书及年报存在虚假记载，公司股票交易将被实施其他风险警示。

WRC唯一真「人机共生」展台，是家沉浸式机器人咖啡店

玩的就是真实

具身创业里的香港教授们

香港的大学里冒出了一批很特别的人

他给了王兴兴第一个200万，现在给下一个「宇树」当董事长

多多支持像王兴兴这样优秀的具身机器人创业者