Claude Code 会话间可互发消息

Claude Code 新功能：你的会话现在可以互相发送消息了。 无需在另一个会话中重新解释自己，你现在可以让 Claude 代为传达。它会发送一份摘要（而非你的历史记录或文件），另一个会话会在任务进行中接收该摘要。（AIHOT源：X：Claude Devs (@ClaudeDevs)）

Kitesurf：一款在 V8 隔离环境中运行的"代理优先"浏览器

Cloudflare 推出 Kitesurf，一款专为 AI 智能体设计的浏览器，完全运行在 Workers 上，基于 V8 隔离环境，现已在 Browser Run 中免费开放测试。（AIHOT源：Hacker News 热门（buzzing.cc 中文翻译））

OpenAI 将 Astra 列为首个"关键"网络安全模型

OpenAI 在评估其即将推出的模型 Astra 后，依据"准备框架"将其列为首个"关键"网络安全模型。OpenAI 表示已为此情景做好规划，并将实施额外控制措施以确保 Astra 后续开发的安全。该公司正努力让 Astra 广泛可用，并将其先进网络能力交到防御者手中。（AIHOT源：X：OpenAI (@OpenAI)）

HPC-Ops × SGLang：腾讯混元开源高性能 Attention、Router GEMM 与 MoE 算子

腾讯混元开源算子库 HPC-Ops 已集成至 SGLang 主分支，其 Dynamic Attention 与 Fused MoE 在 Hy3 模型上最高降低 TPOT 48.8%。（AIHOT源：LMSYS：Blog（Chatbot Arena 团队））

LangChain 推出 Managed Deep Agents 公开测试版

LangChain 的 Managed Deep Agents 进入公开测试版，可将 Deep Agents 部署到托管的 LangSmith 运行时。该服务提供持久化执行、记忆、沙箱、通道、评估（evals）及生产级基础设施。（AIHOT源：LangChain：Blog（RSS））

Suno移动端上线Voices功能

Voices 功能已在 Suno 移动应用 iOS 和 Android 版正式上线！📱✨ 现在你可以直接在手机上录制人声，并将其用于你的歌曲中。只需在创作界面点击"+ Voice"按钮，录制至少一分钟，然后让音乐流淌起来。（Pro 和 Premier 套餐可无限使用，免费套餐可体验有限版本！） 打开应用，录制你的声音，并在评论区告诉我们你的使用体验！（AIHOT源：X：Suno (@suno)）

斯坦福与 Arc Institute 用 AI 设计全新病毒基因组，16 种在实验室成功杀死细菌

斯坦福大学与 Arc Institute 团队用 AI 模型 Evo 从零设计完整病毒基因组，并在实验室构建出 16 种自然界不存在的功能性病毒。Evo 提出 70 万个候选基因组，团队仅筛选最有希望的 285 个序列合成并植入细菌，其中 16 个成功复制并杀死宿主。该研究已通过同行评审发表于《Science》，但 Evo 未接受人类病原体数据训练，且能否推广至其他病毒类群仍是未知数。（AIHOT源：The Decoder：AI News（RSS））

蚂蚁百灵开源 Ling-3.0-flash：124B 总参数 MoE 模型，支持 API、单机与高性能三种部署

蚂蚁百灵正式开源新一代原生混合推理模型 Ling-3.0-flash，采用 124B 总参数、5.1B 激活参数的 MoE 架构，并提供 FP8、FP4、INT4 等多个版本。（AIHOT源：公众号：蚂蚁百灵（Ling））

Runway 上线 Seedance 2.5，支持 50 个角色参考

Seedance 2.5 现已登陆 Runway。每次生成最多可引用 50 个角色参考，构建充满角色的完整世界；可创作最长 30 秒、带完整音效与对白的片段，再按你的故事需求随意剪辑与延展。 点击下方链接立即开始。（AIHOT源：X：Runway (@runwayml)）

小红书联合浙大、复旦提出 CULTURE-MT：首个面向社媒翻译的「文化有效性」评测基准，入选 ICML 2026

小红书联合浙江大学、复旦大学提出 CULTURE-MT，这是首个面向中英社媒笔记翻译、兼顾文化符号传递与情感共鸣的评测基准，并首次提出「文化有效性」评估标准与自动评估模型 JUDGER（准确率 86.03%）。（AIHOT源：公众号：小红书技术（dots.llm））

Krea 推出 Seedance 2.5 视频模型

推出 Seedance 2.5。 30 秒连续视频、完整多镜头序列，以及最多 50 个参考。 立即试用 👇（AIHOT源：X：Krea AI (@krea_ai)）

独立开发者用 VoxCPM 克隆网红声音，让 AI 终于"会聊天"了

独立开发者叶小叔用面壁智能开源的 VoxCPM 克隆千万粉丝网红声音，搭建 STT → LLM → VoxCPM（TTS）三段式实时对话管道，TTS 首包延迟不到 1 秒，端到端体感 2 到 3 秒。（AIHOT源：公众号：面壁智能（MiniCPM））

What was your win this week?

The Year I Started Leaving Breadcrumbs Instead of Notes

ratatop: the process table, and the parentheses that ruin everything

Agent Sandboxes: Giving AI Agents Their Own Little Linux Box (And Why You Should Care)

linkedSignal() Writes Back: A Trace Through the Reactive Graph 🔍🚀

Claude Blog: Building intelligent apps for Apple platforms with Claude in the Foundation Models framework

Today we're releasing Foundation Models framework support for Claude through a new Swift package that lets Apple developers use Apple's Foundation Models framework to call Claude for more complex workflows. Apple’s Foundation Models framework gives developers access to tap into models natively from Swift. It is very easy to use and can return typed Swift values through guided generation in as few as three lines of code. Developers can use this to tap into Apple’s on-device models for fast, local tasks like summarization or extraction. Developers can now use Apple’s Foundation Models framework to hand off to Claude when a request calls for multi-step reasoning, code generation, and more. Claude can also search the web for current information and execute code for data analysis. Stream Claude's response back into the same view. Because Apple's framework returns typed Swift values from @Gene...

Claude: We’re updating Claude Fable 5’s biology safeguards to reduce false positives

We’re updating Claude Fable 5’s biology safeguards to reduce false positives. In our testing, this update reduced biology-related fallbacks by about 85% across our product surfaces. Fable can now assist on a wider range of everyday health and educational questions. We believe the biggest positive impacts of AI will be in biology and medicine, and we’re committed to putting frontier intelligence safely into the hands of as many researchers as possible. Fable will continue to fallback to Opus 5 for requests we consider dual-use—including virology, toxicology, and molecular design—so it isn't yet usable for professional biology research and drug development. We're committed to closing that gap through trusted access pathways for frontier biology capabilities. https://t.co/zbjxPhc909｜@claudeai · 热度 4793 · 2026-08-07

Thibault Sottiaux: You can just ask Codex with GPT-5.6 Sol the wildest things and it will just d...

You can just ask Codex with GPT-5.6 Sol the wildest things and it will just do it. I talk to it for 5 minutes straight with things that just seem to require weeks of work, get up to get something in the fridge, pet the dog, come back and ... it's done.｜@thsottiaux · 热度 3210 · 2026-08-07

Aaron Levie: Huge Atlassian quarterly beat

Huge Atlassian quarterly beat. There was a misplaced thesis over the past 6 months that somehow agents would be bad for certain software categories. There’s definitely truth in this in some areas, but many were parsing this poorly. In a world where agents are generating 100X more code, processing massive amounts of data, or making decisions across your systems, the role of the platforms that manage this data and these workflows becomes more important, not less. Enterprises care about governance, security, compliance, guardrails, safe access to data, and many more critical capabilities that go into these systems of record.｜@levie · 热度 1045 · 2026-08-06

Guillermo Rauch: Devtools must be 1️⃣ open source and 2️⃣ universally extensible

Devtools must be 1️⃣ open source and 2️⃣ universally extensible. AI coding agents are the most important devtools in the history of our industry. The Plugin standard lets anybody extend them uniformly. This is huge for the ecosystem. Build a Plugin → get exposure to the tidal wave of software creation from every agent: CLIs, IDEs, cloud agents, and even personal assistants.｜@rauchg · 热度 487 · 2026-08-06

Thibault Sottiaux: Free users of ChatGPT now have unlimited text chats, powered by GPT-5.6 Luna ...

Free users of ChatGPT now have unlimited text chats, powered by GPT-5.6 Luna https://t.co/eZq1e1hTQX｜@thsottiaux · 热度 257 · 2026-08-07

Peter Yang: Consumer is basically ChatGPT and Google's market to lose in my opinion

Consumer is basically ChatGPT and Google's market to lose in my opinion. ChatGPT already has 1B users and needs to get them to connect their favorite apps and use agents to do work. OpenAI is already pioneering new interaction patterns like voice. The biggest barriers to overcome are: 1) Normies not trusting AI and ChatGPT with full access to their apps and data. 2) Normies not realizing what ChatGPT is now capable of. In both cases, I think marketing and the right messaging are just as important as improving the product. Google Gemini also has almost 1B users and my gut is normies trust it more to connect to email, calendar, Workspace, Chrome, etc since it's all the same login. But Google is behind on 3rd plugins, browser use, and alot of other critical features. They need to make the Gemini app best in class and friendly for normies. They can just copy ChatGPT tbh except make improvements like not have 3 separate modes (ChatGPT/Work/Codex) that ChatGPT and Claude are stuck in. fwiw I don't think having acess to the best model is that important to normies vs. getting the user onboarding right. My non-AI pilled friends could care less about Sol vs Fable as long as the pricing is fair and the model can get their work done reliably.｜@petergyang · 热度 203 · 2026-08-06

Amjad Masad: It’s true, in 21/22 I went around the valley asking everyone to train coding ...

It’s true, in 21/22 I went around the valley asking everyone to train coding specific models with us: Google, Meta, everyone — no one thought it was as important as NLP use cases — eventually we trained our own: Replit-code-3b and then everyone got code pilled. https://t.co/BqVr3xFYML｜@amasad · 热度 136 · 2026-08-07

Madhu Guru: I’ve noticed people are much clearer when they speak through new ideas than w...

I’ve noticed people are much clearer when they speak through new ideas than writing a doc on it. Somewhere between their mouth and the doc, the core idea gets buried as they try to add context, polish it and make it sound smart. So here’s what I ask my teams to do: record yourself explaining the idea exactly as you would to a friend, use ai to do basic cleanup but keep the rough original structure and flow. Share that doc.｜@realmadhuguru · 热度 121 · 2026-08-06

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 · ⭐98225 · Search 增补 · 近快照 +825星 · 创建 2026-06-12

Panniantong/Agent-Reach

Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.｜Python · ⭐68330 · Search 增补 · 近快照 +767星 · 创建 2026-02-24

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 · ⭐39609 · Search 增补 · 近快照 +706星 · 创建 2026-03-17

Leonxlnx/taste-skill

Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop｜JavaScript · ⭐73831 · Search 增补 · 近快照 +670星 · 创建 2026-02-19

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 · ⭐104008 · Search 增补 · 近快照 +530星 · 创建 2026-04-03

AMD acquires Taalas to boost inference performance by etching models in silicon

https://ir.amd.com/news-events/press-releases/detail/1296/am... https://chatjimmy.ai/

New Mexico court orders Meta to pay $567m over harms to children’s mental health

https://www.reuters.com/world/new-mexico-court-orders-meta-p... https://www.theguardian.com/technology/2026/aug/06/new-mexic... https:/&#x2F

DeepSeek V4 Flash 0731

A year of fighting scrapers on my 1.5 million-page website

Oracle bans AI-generated code from OpenJDK

Anthropic 为自研 AI 芯片开出高薪，模型训练岗位薪资高于实际造芯工程师

IT之家 8 月 7 日消息，据 Wccftech 今日报道，Anthropic 目前正在扩大内部芯片设计团队，并计划开发自研 ASIC 芯片。 与此同时，公司不同芯片相关岗位之间出现了明显的薪资差异：负责训练 AI 模型进行芯片设计的研究工程师岗位，年薪最高可达 85 万美元 （IT之家注：现汇率约合 574.9 万元人民币） ，而实际负责设计 Anthropic 首款 ASIC 芯片的工程师岗位，薪资范围约为 32 万至 48.5 万美元 （现汇率约合 216.4 万至 328 万元人民币） 。Wccftech 称其为“薪酬倒挂”。 Anthropic 发布的招聘信息显示，负责“...

AI 需求挤爆云计算，消息称 AWS 要求工程师关闭闲置服务器减少资源浪费

IT之家 8 月 7 日消息，据《The Information》今日报道，亚马逊 AWS 最近对工程师提出一项颇为“朴素”的要求，让他们关闭不再使用的云计算机。 据报道，亚马逊此番要求主要针对闲置的 EC2 实例。随着 AI 正以越来越快的速度消耗计算资源，全球最大的云服务供应商也开始感受到算力紧缺的压力。 数据显示，AWS 的资源利用低下问题非常严重，大约 65% 的 EC2 实例在 30 天测量周期内，平均 CPU 利用率低于 20%。 亚马逊为此升级了计算优化器（IT之家注：Compute Optimizer），自动分析 14 天内的服务器 CPU 使用率和 IO 数...

Take-Two CEO 称《GTA6》定价“超划算”，因为原本可以定得更高

IT之家 8 月 7 日消息，在今日举行的 2027 财年第一财季财报电话会议上，Take-Two 首席执行官 Strauss Zelnick 谈到了《侠盗猎车手 VI》相关内容，他表示 79.99 美元 （IT之家注：现汇率约合 541 元人民币） 的售价对玩家而言“简直是捡了个大便宜”。 Zelnick 表示：“我们说过很多次了，我们的目标是向消费者提供远超收费的价值。说《GTA6》在游戏体验方面会物超所值已经算是保守的说法，因为 Rockstar Games 一向以超出预期而闻名。” 他进一步指出，以《GTA6》的热度而言，Take-Two 本可以将价格定得更高，但公司并未这么...

Gartner发布2026中国网络安全技术成熟度曲线，AI安全开始“挤泡沫”

点击查看原文>

蚂蚁开源Avernet，为多智能体协作搭建“操作系统”！内部跑通12大业务、任务完成率超90%

点击查看原文>

10万人排队、一键出片，AI创作越简单，爆款为什么反而更难做了｜请回答 WAIC 2026

点击查看原文>

Am I the problem? Interviewing another team to find out

A shell exclamation mark is not for yelling. Be lazy

What are you doing this weekend?

I'm leaving OpenAI to build Jurassic Park

tl;dv (Too Lazy; Didn't Validate): 181,874 Meetings Left Wide Open

The Download: a censorship conspiracy theory and the first virus created by AI

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 ideas of a vast censorship network moved from the online fringe to Trump policy For years, narratives about a “censorship-industrial complex” spread in rig...

How ideas of a vast censorship network moved from the online fringe to Trump policy

This article was produced in partnership with Type Investigations, with support from the Wayne Barrett Project. One morning in April 2025, employees of a small office in the US State Department got the email many of them had been dreading.  For months, Elon Musk’s Department of Government Efficie...

The Download: Google’s AI shake-up and Meta’s rogue model

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. Google’s AI empire is being reshaped. Here’s what’s changed. After a wave of painful losses in the tech talent wars, delays to its next flagship model, and mur...

After Rippling blew millions on AI in months, it built an employee ROI tool

After its own AI usage wake-up call, Rippling this week unveiled AI Spend Console, a product that tracks individual and team employee AI spending.

Wacom’s Movinkpad 11 is a fun and mid-priced entry point for digital artists

The Movinkpad 11 a versatile little graphics tablet that can help make your wildest digital art dreams come true.

Security researchers scanned the Polish web and found courts, hospitals, and airports at risk of hacks

Researchers found common points of failure, like software used to organize and display web content, could have allowed hackers to run riot through government websites.

The gaming site sponsored by Walmart lays off its editorial staff

Restart, a games media website launched in late 2024 that was sponsored by Walmart, has laid off its "entire editorial team," according to a Friday post from Brandy Berthelson, the site's former editor-in-chief. Other staffers have also posted about being laid off. Restart is operated by Moonrock...

Fenix Flexin isn’t even denying using AI to make ‘Rubberz’ anymore

It took long enough, but now LA rapper Fenix Flexin appears to have admitted using AI for the 80s synth pop-themed song "Rubberz." His comments follow the producer Medasin's videos claiming that an AI tool called Treblo (formerly Sonauto) was used to make the song, and the company releasing an AI...

Watching Roku’s AI channel is like eating from a trough

The appeal of free ad-supported streaming television (FAST) channels has always been the way they make it easier to (re)discover classic films and series. But Roku's latest experiment in the FAST space has less to do with traditionally produced entertainment and is entirely focused on giving view...

本周看什么 | 最近值得一看的 10 部作品

📅本周新预告《生化危机：爆发夜》中字预告8月5日，电影《生化危机：爆发夜》发布了全新中字预告，将于9月18日在北美上映。扎克・克雷格（《凶器》《野蛮人》）执导，奥斯汀·艾布拉姆斯主演，一份双倍酬劳的郊 ... 查看全文

角落新声｜从合租到整租，我在客厅听见了自己的声音

当我坐下来，这个小角落就慢慢有了属于我的声音。 查看全文

游戏辅助后续之yolo

之前的游戏辅助跑了一个月了, 因为刷了很多资源, 账号能力提升很多, 于是就有了刷下一级副本的能力. 下一级副本效率高了50%, 单位时间产出也高了50%, 但多了难度是 boss 会释放随机技能,

漫话火山 Milvus｜为 Agent、RAG 、语义搜索而生，AI 时代的极致性价比之选

火山Milvus是基于开源Milvus的全托管向量数据库，支持Serverless，联动火山大模型生态，适配RAG、Agent，免底层运维，解决锁定、扩容、成本难题。

AI 培训现场翻车后的应急方案设计：从工具选型到教学流程的系统化思考

导语 8 月 5 日，我在丽水为近 20 位企业主开展了一场 AI 培训。现场遭遇学员全员未携带电脑的情况，凭借"三端齐全"的工具选型策略和结构化教学流程，成功将课程从电脑实操切换为手机教学，原定 1

特朗普重启解雇美联储理事丽莎·库克的行动

白宫本周表示，特朗普总统正在“考虑”解雇美联储理事会理事丽莎·库克，这重新启动了一项曾在6月被最高法院阻止的行动。 白宫总统助理丹·斯卡维诺在周三致库克的一封信中表示...

摩根士丹利对日元转向偏空 干预影响逐渐消退

摩根士丹利外汇策略师对日元转向偏空，他们认为，随着事实证明美日当局支撑日元汇率的措施效果短暂，日元将逐步走弱。 当局联手支撑日元仅短暂提振了日元需求...

大宗商品综述：油价持稳 铜价微跌 金价大涨

原油持稳，中东紧张局势升温之际，交易员密切关注伊朗与阿曼就恢复霍尔木兹海峡通航达成协议的进展。铜价周五微跌，但仍然录得单周上涨，市场出现铜供应趋紧的迹象。

8月8日外盘头条：伊朗内部辩论霍尔木兹协议措辞 非农爆冷削弱加息预期 SpaceX将自建发电厂保障电力供应

全球财经媒体昨夜今晨共同关注的头条新闻主要有： 1、伊朗内部辩论霍尔木兹协议措辞 特朗普坚称谈判“正在推进” 2、特朗普力推30亿美元关键矿产投资项目 3...

三菱日联：建议做多澳元/日元 美联储加息押注降温提振套息交易

三菱日联分析师建议做多澳元兑日元，因干预带来的影响正在消退，而交易员削减对美联储加息的押注改善了套息交易的环境。

阿里推出国内首个AI语音平台CosyVoice Studio，将语义理解融入语音能力

AI语音“听、说、创”一站式满足

AI批量轰炸苹果bug赏金计划，审核团队已下线

openJiuwen发布业界首个企业级分布式蜂群架构，联合邮储成功落地金融生产环境

从「能用」走向「规模化落地」