Claude 接管应用日常维护：388 个 PR 的实践

Boris Cherny 尝试让 Claude 接管其应用的日常维护，通过 Slack 频道运行崩溃模糊测试、重复代码统一、死代码移除等日常任务。数周内自动开出 388 个 PR，其中 180 个经 Claude Code Review 和人工审核后合并。Claude 通常一次就能改对，出错时可通过调整例程次日改进。（AIHOT源：X：Boris Cherny (@bcherny)）

Google DeepMind 推出 Gemini 3.7 Flash：面向编程与智能体的最强工作模型

Google DeepMind 发布 Gemini 3.7 Flash，距 3.6 Flash 仅三周，主打编程与智能体任务，输入/输出价格分别为每百万 token $0.75 和 $3.75，为原 3.6 Flash 的一半。（AIHOT源：Google DeepMind：Blog（RSS））

MiniMax Music 3.0 发布：新一代开源权重、生产级全能音乐模型

MiniMax 推出 Music 3.0，新一代音乐生成模型，可根据创意概念和可选歌词一次性完成整首歌的作曲、编曲、演奏与制作，最长支持五分钟。（AIHOT源：MiniMax：Blog（网页））

Google Sheets 推出 Sheets canvas：用 Gemini 将表格数据变为交互式迷你应用

Google Sheets 发布新功能 Sheets canvas，基于 Gemini 构建，用户只需用自然语言提示词即可将表格数据转化为交互式仪表盘、学习追踪器、座位表等"迷你应用"。（AIHOT源：Google Blog：AI（RSS））

Qwen3.8-2.4T-A95B 开源，硅基流动即日上线

阿里开源 Qwen3.8-2.4T-A95B，硅基流动已提供 Day-0 支持。该模型拥有 2.4T 参数、95B 激活参数，主打自主编码、深度研究与端到端智能体执行。API 定价为输入 $2.00/百万 token，输出 $6.00/百万 token，缓存输入 $0.25/百万 token。（AIHOT源：X：硅基流动 SiliconFlow (@SiliconFlowAI)）

DeepSeek Harness v0.1 开发者预览版发布

DeepSeek Harness v0.1 现已推出开发者预览版，并以 MIT 许可证开源。该智能体框架基于 Cordis 元框架构建，核心设计为"一切皆插件"，模型、工具、技能、会话、沙箱、文件系统、循环、编排及 UI 均可自由组合、替换和扩展。（AIHOT源：X：DeepSeek (@deepseek_ai)）

Cursor 推出 builds：云智能体启动速度提升至 3 倍

Cursor 推出 builds 功能，在后台持续准备就绪的开发环境副本，让云智能体启动时无需从零搭建，响应速度最高提升 3 倍。内部环境启动快 10 倍，首个 token 生成快 3 倍；智能体始终从最近一次成功的 build 启动，依赖更新或安装脚本出错时不会影响运行。8 月 17 日起所有环境默认启用 builds，无需额外费用。（AIHOT源：Cursor Blog）

DeepSeek-V4-Pro 正式版上线，Agent 能力大幅增强

DeepSeek-V4-Pro 正式版已在 APP、网页端和 API 同步上线，模型名设为 deepseek-v4-pro 即可使用。其 Agent 能力显著提升，HLE （wo/w tools） 达 42.7/60.0，Terminal Bench 2.1 为 87.9。（AIHOT源：DeepSeek：API 更新日志）

GPT-5.6 构建者指南：如何以更低成本实现前沿智能体性能

GPT-5.6 模型家族以更低成本实现前沿级智能体性能，并新增推理持久化、原生多智能体编排和程序化工具调用等 API 能力。在 ARC-AGI-3 上，启用保留推理和压缩后，Sol 得分从 13.3% 跃升至 38.3%，且输出 token 减少约 6 倍。Luna 在 BrowseComp 上以 84.04% 的得分追平 GPT-5.5（84.36%），成本从 $33.27 降至 $1.33。（AIHOT源：OpenAI：官网动态（RSS · 排除企业/客户案例））

小红书开源连续自回归语音合成模型 dots.tts：打造可持续扩展的 TTS 基座

小红书 dots 团队开源 20 亿参数全连续端到端自回归语音合成模型 dots.tts，在 Seed-TTS-Eval 三个子集上取得最佳平均内容准确度和平均说话人相似度。（AIHOT源：公众号：小红书技术（dots.llm））

WorkBuddy上线远程控制，国内也有了最丝滑的Agent工作方式

WorkBuddy更新上线远程控制功能，将PC、App和小程序打通，手机端可实时同步电脑端的任务、对话、工作空间和产物，支持一台手机连接多台电脑并随时切换。App需升级至1.2.0及以上，电脑端需升级至5.3.8及以上，连接无需扫码。本次更新还新增资料库（我的文档与团队空间）、Markdown多人共同编辑、AI原生审阅模式，以及将资料库内容生成可发布链接的HTML网站。（AIHOT源：公众号：数字生命卡兹克）

新兴多智能体系统的模式与问题

Anthropic 研究指出，随着 AI 智能体在共享代码库、市场等社会系统中承担更多任务，智能体间交互量或将超过人机交互。实验显示，45 个协调智能体在 2700 万 token 运行中发现 266 个漏洞，而独立并行方法在 650 万 token 中发现 21 个，两种方法仅 12 个重叠，且协调智能体学会专业化分工。研究同时警示个体层面的良性行为怪癖可能叠加为意外的系统性失败。（AIHOT源：Anthropic：Research（发表成果 · 网页））

Not All AI Builders Are Doing the Same Work

I Stopped Trusting AI Agents With Tools. So I Built a Gatekeeper.

24 Cups, 36 Seats — The Bartender's Ledger

My (not so pretty) journey in tech

Do I Still Need a Monkey Patch for Gemini Live?

Anthropic Engineering: An update on recent Claude Code quality reports

Over the past month, we’ve been looking into reports that Claude’s responses have worsened for some users. We’ve traced these reports to three separate changes that affected Claude Code, the Claude Agent SDK, and Claude Cowork. The API was not impacted. All three issues have now been resolved as of April 20 (v2.1.116). In this post, we explain what we found, what we fixed, and what we’ll do differently to ensure similar issues are much less likely to happen again. We take reports about degradation very seriously. We never intentionally degrade our models, and we were able to immediately confirm that our API and inference layer were unaffected. After investigation, we identified three different issues: On March 4, we changed Claude Code's default reasoning effort from high to medium to reduce the very long latency—enough to make the UI appear frozen—some users were seeing in high mode. Th...

Anthropic Engineering: Scaling Managed Agents: Decoupling the brain from the hands

Get started with Claude Managed Agents by following our docs . A running topic on the Engineering Blog is how to build effective agents and design harnesses for long-running work . A common thread across this work is that harnesses encode assumptions about what Claude can’t do on its own. However, those assumptions need to be frequently questioned because they can go stale as models improve. As just one example, in prior work we found that Claude Sonnet 4.5 would wrap up tasks prematurely as it sensed its context limit approaching—a behavior sometimes called “context anxiety.” We addressed this by adding context resets to the harness. But when we used the same harness on Claude Opus 4.5, we found that the behavior was gone. The resets had become dead weight. We expect harnesses to continue evolving. So we built Managed Agents: a hosted service in the Claude Platform that runs long-horizo...

Claude Blog: New in Claude Managed Agents: self-hosted sandboxes and MCP tunnels

Starting today, Claude Managed Agents can operate in a sandbox you control and connect to your private Model Context Protocol (MCP) servers. Both the sandbox where an agent executes tools and the services it reaches run within the established boundaries of your enterprise, under your security and runtime controls. The sandbox runs on your own infrastructure, or with managed providers like Cloudflare , Daytona , Modal , or Vercel to handle the compute and isolation for you. On the Claude Platform, self-hosted sandboxes is available in public beta and MCP tunnels in research preview ( request access ). Self-hosted sandboxes: keep agent execution within your perimeter A self-hosted sandbox lets a Claude Managed Agent execute tools on infrastructure you control or with a managed sandbox provider. Code execution, sensitive files, packages, services, and data stay within your enterprise perime...

Claude: Your Claude in Chrome sessions now carry over to desktop, web, and mobile

Your Claude in Chrome sessions now carry over to desktop, web, and mobile. Conversations are saved, and your skills and connectors work in the browser. Available on Max and Team today, rolling out to Pro in the coming weeks. https://t.co/Hnxs18PVI8｜@claudeai · 热度 10145 · 2026-08-12

Garry Tan: GBrain just dropped v0.45.6.0 which added 17 new brain skills hardened throug...

GBrain just dropped v0.45.6.0 which added 17 new brain skills hardened through my own personal OpenClaw agent with hundreds of thousands of markdown files. A personal AGI is one that works for you. GBrain now works with Codex and Claude code, btw. https://t.co/yFpFU4pn5b https://t.co/MqpMfX4wsX｜@garrytan · 热度 777 · 2026-08-12

Claude: The side panel now runs the same Claude Cowork session as the desktop, web, a...

The side panel now runs the same Claude Cowork session as the desktop, web, and mobile apps. Sessions live with your account, not on any single device, so you can start in a tab and pick it up later somewhere else. Give it a try: https://t.co/8xTEK7osxX｜@claudeai · 热度 429 · 2026-08-12

Guillermo Rauch: Seedance 2.5 on @vercel AI Gateway 🐇 https://t.co/Ql9zpuJpBD

Seedance 2.5 on @vercel AI Gateway 🐇 https://t.co/Ql9zpuJpBD｜@rauchg · 热度 284 · 2026-08-12

Josh Woodward: Gemini gets things done across the apps you use every day

Gemini gets things done across the apps you use every day. Starting today, another wave of integrations are rolling out. If you want to partner with us, DM me. Angi (formerly Angie's List) Fever GetYourGuide Granola iHeartRadio Localiza OpenTable Otter Pandora Thumbtack Ticketmaster Wix Zocdoc Zoho｜@joshwoodward · 热度 283 · 2026-08-13

Aaron Levie: Awesome day in terms of new model releases from both Deepseek and Grok

Awesome day in terms of new model releases from both Deepseek and Grok. Both of these new model updates represent huge jumps in capability at insanely low costs. This is truly Jevons paradox for AI playing out. By bringing down the cost of AI, demand will continue to accelerate. When I talk to enterprises, they have vastly more use-cases for what they’d like agents to handle than generally what they’re able to budget for today. They want agents to scan their codebases for security issues. They want agents to review all their documents. They want agents to process information in a large portion of their workflows. And so on. When we bring down the cost of AI or drive up capability levels, it opens up all new use cases in the enterprises, which drives even greater demand. This is also great for the applied AI layer. The more choice of models - especially those tuned for different types (or costs) of work - the more value there is in being at the layer that can route and optimize based on the task. Great for the entire stack right now.｜@levie · 热度 259 · 2026-08-13

Claude: Browser agents can be tricked by instructions hidden in a page

Browser agents can be tricked by instructions hidden in a page. We build defenses against this, and we still recommend a few habits of your own: https://t.co/aHIHLNkOnk｜@claudeai · 热度 258 · 2026-08-12

Madhu Guru: The biggest alpha in AI products over the next few years is in the applicatio...

The biggest alpha in AI products over the next few years is in the application layer. Models will keep getting cheaper, better and more local. The differentiation will come from deeply understanding specific user workflows and having the imagination to redesign the experience around them. The denominator of people who can build AI products is about to get enormous. So being in the top 0.1% of builders will matter more than ever.｜@realmadhuguru · 热度 224 · 2026-08-12

Peter Yang: I believe the way we use computers will soon change forever

I believe the way we use computers will soon change forever. We’re moving from doing work manually with a keyboard, mouse, and laptop to directing agents in the cloud using our voices. In my new essay, I cover why: 1. Voice is becoming the orchestration layer 2. Personal computers will move to the cloud 3. Trust will be the key differentiator Read it here (it’s 100% free): https://t.co/cPvIqq0mge｜@petergyang · 热度 126 · 2026-08-12

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 · ⭐44879 · Search 增补 · 近快照 +2187星 · 创建 2026-03-17

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 · ⭐102130 · Search 增补 · 近快照 +1340星 · 创建 2026-06-12

Leonxlnx/taste-skill

Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop｜JavaScript · ⭐76254 · Search 增补 · 近快照 +829星 · 创建 2026-02-19

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 · ⭐71444 · Search 增补 · 近快照 +735星 · 创建 2026-02-24

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 · ⭐106019 · Search 增补 · 近快照 +709星 · 创建 2026-04-03

DeepSeek Harness developer preview

https://github.com/deepseek-ai/deepseek-harness https://deepseek-harness.github.io/deepseek-harness/en/guide...

Gemini 3.7 Flash

https://ai.google.dev/gemini-api/docs/models/gemini-3.7-flas...

Spaghettifying DRAM

Gloomberb

Accelerating GPT-5.6 Sol Ultrafast

法拉第未来 2026 财年半年报归母亏损 7488.5 万美元，同比亏损收窄 44.51%

IT之家 8 月 14 日消息，法拉第未来今日发布 2026 财年（2026 年 1 月～2026 年 12 月）半年报（2026 年 1 月～2026 年 6 月）及第二财季（2026 年 4 月～2026 年 6 月）报告： 2026 财年半年报 营业总收入 ：134.8 万美元 （IT之家注：现汇率约合 910.83 万元人民币） ，同比下降 99.93% 毛利润 ：-2208 万美元 （现汇率约合 -1.49 亿元人民币） ，同比亏损收窄 53.93% 毛利率 ：-1638%，同比亏损扩大 342.78 个百分点 归母净利润 ：-7488.5 万...

英伟达 GeForce NOW 原生 Linux 应用发布：结束 Beta 测试，支持 Ubuntu 24.04 及 Flatpak 安装

IT之家 8 月 13 日消息，英伟达于当地时间 8 月 13 日宣布，其云游戏服务 GeForce NOW 的原生 Linux 应用已正式结束测试阶段，面向所有用户发布。 该应用于 2026 年 1 月以 Beta 版形式首次登陆 Linux 平台，经过数月的社区反馈与优化，现已推出正式版本。 Linux 版 GeForce NOW 客户端将继续支持 Ubuntu 24.04 及更新版本。英伟达同时新增了官方 Flatpak 软件仓库，以简化安装和后续更新流程。用户可通过英伟达的.bin 安装程序或添加 Flatpak 仓库直接安装 com.nvidia.geforc...

全球最大电池纯电动飞机 Heart Aerospace X1 完成首飞：超 1 兆瓦功率输出 + 超 11 吨起飞重量

IT之家 8 月 13 日消息，当地时间 8 月 12 日，瑞典航空航天公司 Heart Aerospace 宣布其 X1 演示机在美国纽约州普拉茨堡国际机场完成首次载人飞行。 这架飞机翼展达 106 英尺 （IT之家注：约 3,231 厘米） ，机长 76 英尺 （约 2,316 厘米） ，起飞重量超过 25,000 磅 （约 11,340 千克） ，这也是迄今为止完成飞行的最大电池的电动飞机。 此次试飞在美国联邦航空管理局颁发的实验类特殊适航证书（SAC-EC）框架下进行。飞行持续约 27 分钟，最高爬升至离地高度 1,100 英尺 （约 335.3 米） 。 测试...

从“开放模型”到“开放生态”，AI 开源进入下半场

点击查看原文>

微软 AI Gateway 新层级引热议：统一治理背后的权限隐忧

点击查看原文>

软件不是文件：KDC 的知识工程主张

点击查看原文>

python's pre-declared constants are kinda weird

I want extern "fil-c"

EuroZig

AI is removing the middle class of software engineering

Unlocking _everything_ on the CPU with DRAM scrambling

Roundtables: Inside the “Censorship-Industrial Complex” Idea Shaping US Policy

Listen to the session or watch below The “censorship-industrial complex” is an idea that a network of government, tech, and research groups is collaborating to suppress conservative online speech. This was fodder for the right-wing information sphere for years—then it began making its way into US...

Building a practical path to post-quantum cryptography

Quantum computing has alternated between breakthrough darling and overhyped promise in technology circles. Its powerful new capabilities come with a threat to break current cryptography, but for business leaders navigating the noise, the signal should be clear: post-quantum cryptography (PQC) is ...

Flock is tightening its rules in response to a growing surveillance backlash

The police-tech giant Flock is announcing today that it will change officers’ access to its nationwide network of license plate readers, in an apparent effort to quell a growing backlash and win back contracts lost amid concerns about mass surveillance and police abuse. Several changes aim direct...

Investors sue Selena Gomez alleging fraud tied to her mental health startup

The plaintiffs say they invested nearly $1.2 million in the company, and are accusing Gomez of failing to build and market the startup.

Flock says its new tool will help identify police abuse, but hasn’t explained how it works

The surveillance company announced it's making a tool called "Audit Assistance" mandatory for all customers, claiming it's already helped catch abuse. But the company has yet to explain how the tool works in detail, raising questions about its effectiveness.

If Apple sends you a push notification alerting you to a spyware attack, take it seriously

Apple now sends out push notifications to iPhone lock screens when the company identifies government spyware targeting someone's devices.

‘That is not acceptable’: Judge orders Google to make rival app store installs easier

One month after Epic Games and Google seemingly stopped fighting over the future of Android app distribution, they were back in a San Francisco courtroom today - where Judge James Donato just ordered Google to make it easier to install rival app stores on Android. It's been nearly three years sin...

The fight over Flock and other ALPRs

There are over 120,000 of Flock’s automatic license plate reader (ALPR) cameras installed all over the US. Flock’s cameras, and others like them, use AI to identify and track vehicles based on their license plate number, make, model, color, and other info, networked together to track vehicles and...

Microsoft’s Clippy-like Mico character is no longer the face of Copilot

Microsoft Copilot will no longer show its emotive yellow blob, Mico, when you use the chatbot's voice mode. In a support page, Microsoft says it's going to move Mico to its Learn Live platform, where the avatar will have "more to react to," as reported earlier by GeekWire. Mico launched in Copilo...

不用改签名，给 Mac 微信聊天记录「搬个家」

对长期把 Mac 放在固定位置、外置硬盘常年连接的用户来说，它可以让体积庞大的聊天记录不再挤占有限的内置存储。 查看全文

从原理出发，高效搞定夏日营养冷食

不知道多少人和我一样，夏天吃饭的纠结项往往从「今晚到底吃什么」，变成了「太热了到底要不要吃呢」。天气太闷太热、厨房又没有空调，在想到开火的那一刻又心生畏惧，最后还是点外卖或者随便吃点东西凑合了事。但也 ... 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。

TDS REVIEW | 铁三角 ATH-SQ1TW2NC 降噪真无线耳机体验

SQ1TW2NC 的声音素质、编码配置和降噪综合能力，应该都算是在国内同价位甚至更低价位主流新产品面前不太拿得出手的水平，没有特别明显的功能层面的长板，但你会发现即使如此，它依然有一些点值得国内品牌学习。 查看全文

邹市明辟谣

胖东来闭店一天整条街都没人出摊

紧急预拨8000万支持3省抢险救灾

林一新歌七夕首发

这3种面包其实是热量刺客

田曦薇胡一天吻戏是我能看的吗

我人生的存在感和意义

逃出绝命街OMG恐龙吓到我了

DeepSeek 昨晚刚开源了 Harness：附万少的2 万字保姆级教程

DeepSeek 昨晚刚开源了 Harness：附万少的2 万字保姆级教程 就在昨天（8 月 13 日），DeepSeek 正式发布首款 Agent 产品 DeepSeek Harness（dsh），

我为 DeepSeek Harness 写了一个 dsh-apple-mode 插件

DeepSeek Harness 终于在千呼万唤中推出了公开测试版。我也是第一时间下载体验了一下，并最终开发了一个 dsh-apple-mode 插件，以更好的适配 iOS/Apple 开发。

连 agent loop 自己都是插件，拆开 DeepSeek 新开源的 harness 底座

DeepSeek 开源的 dsh 把模型适配、工具、会话日志甚至 agent 循环本身都做成可替换插件，底层是借用机器人生态的 Cordis 框架，拆开看它的可逆效果和能力缝隙设计。

8月14日隔夜要闻：标普500指数收盘创历史新高 原油结束连涨走势 存储板块同步走强

市场 8月14日收盘：标普500指数收盘创历史新高 受通胀数据低于预期及油价下跌提振 8月14日美股成交额前20：闪迪收涨13.67% 存储板块同步走强 8月14日热门中概股多数下跌...

大宗商品综述：原油结束连涨走势 黄金和铝价走低

原油价格结束连续上涨，投资者正在研判目前究竟有多少原油通过霍尔木兹海峡。受美国通胀缓和的迹象影响，黄金价格走低并跌破每盎司4，400美元。

Reddit将被纳入标普500指数 盘后股价飙升

根据新闻稿，Reddit将于8月18日开盘前纳入标普500指数。该公司股价周四盘后一度大涨10%。 Reddit将取代AvalonBay Communities。

库什纳预计下周赴开罗与埃及、卡塔尔和土耳其调解人会谈

一位知情人士称，美国总统特朗普的特使贾里德·库什纳预计下周前往开罗，与埃及、卡塔尔和土耳其的调解人举行会谈。 此前有报道称，库什纳还计划下周访问以色列...

原油：WTI涨势停滞 霍尔木兹海峡僵局之际交易员聚焦石油运量

原油结束近期连涨走势，旨在结束伊朗战争的协议进展有限之际，交易员正在研判目前究竟有多少原油通过霍尔木兹海峡运输。

深度体验DeepSeek Harness，我原谅它涨价了

欢迎，为后人种点树

4.8亿美元砸向端侧算力！Agent芯片新贵冲出重围

首颗AI芯片已进入量产

马斯克Grok 4.6重回一梯队！更低价格反超Fable 5，这Cursor是真没白收购

马斯克版Workbuddy也来了