2026-07-07（内容获取于 07/07 05:24）

LeRobot v0.6.0: Imagine, Evaluate, Improve

Hugging Face Blog · 07/07 08:00

中文介绍 Hugging Face发布了LeRobot v0.6.0版本，这是一个旨在通过“想象、评估、改进”迭代过程提升机器人技能的机器人学习框架。新版本可能包含性能优化和新功能，以支持更高效的机器人行为学习与评估，助力社区在机器人AI领域进行创新与协作。

Hoto’s PixelDrive screwdriver is down to $60, matching its best price

The Verge · 07/07 04:24

If your Prime Day purchases included a new desk, TV stand, bookshelf, or other furniture you still haven’t assembled, Hoto’s PixelDrive cordless screwdriver can help speed up the process. It’s currently on sale for $59.99 ($20 off) at Amazon, matching its best price to date. Hoto PixelDrive Cordless

中文介绍 Hoto的PixelDrive无绳电动螺丝刀目前在亚马逊上以59.99美元（优惠20美元）的价格出售，与其历史最低价持平。该产品旨在帮助用户加快家具组装过程，适用于日常家居维修和安装，提供便捷高效的螺丝拧紧解决方案。

What's the oldest Americana flown in space?

Ars Technica · 07/07 03:57

From a Revolutionary War flag to the Statue of Liberty...

中文介绍 这篇文章探讨了飞往太空的最古老的美国文物是什么。内容可能涵盖从美国独立战争时期的旗帜到与自由女神像相关的物品等具有历史意义的美国象征物，分析它们如何被带入太空，以及这些文物所承载的独特文化和历史价值。

Vercel CEO Guillermo Rauch on the fight to split off models from agents

TechCrunch · 07/07 03:49

"The reality is, when you're optimizing for production, you start looking at a price/performance," Guillermo Rauch tells TechCrunch.

中文介绍 Vercel首席执行官Guillermo Rauch在接受TechCrunch采访时，讨论了将AI模型与代理（agents）分离的重要性。他指出，在优化生产环境时，价格与性能的权衡是关键，强调了在实际应用中实现效率和成本效益的必要性。

You can now customize Siri’s pace and expressivity in the latest iOS 27 beta

TechCrunch · 07/07 03:01

The update is part of Apple's broader effort to make Siri feel more natural and personal, as it rebuilds the assistant around generative AI.

中文介绍 苹果在最新的iOS 27测试版中，引入了允许用户自定义Siri语速和表达方式的功能。此更新是苹果利用生成式AI重塑Siri，使其听起来更自然、更个性化的努力之一，旨在显著提升用户与Siri的互动体验和满意度。

Cloudflare replaces its blanket AI bot block with granular controls for search, training, and agent crawlers

The Decoder · 07/07 02:54

Cloudflare is giving all customers granular AI bot controls. Site owners can now manage Search, Training, and Agent bots separately instead of blocking them all at once. Starting September 15, 2026, Training and Agent bots will be blocked by default on ad-supported pages. The article Cloudflare repl

中文介绍 Cloudflare为其所有客户提供了精细化的AI机器人控制选项，网站所有者可独立管理搜索、训练和代理机器人，而非一概阻止。从2026年9月15日起，在支持广告的页面上，训练和代理机器人将默认被阻止，以增强网站对其流量的自主控制权。

America’s greatest idea is still under threat

The Verge · 07/07 02:36

The Patrouille de France perform a flyover during the Sail4th 250 Tall Ships Parade along the Hudson River in New York on July 4th, 2026. | Photo by John Lamparski/Bloomberg via Getty Images The United States of America recently turned 250 years old. What a spectacle! The fireworks were amazing, and

中文介绍 2026年7月4日，美国庆祝建国250周年之际，这篇文章探讨了美国引以为傲的“最伟大理念”——可能指言论自由或民主原则——目前所面临的威胁。文中提及法国巡逻兵在美国纽约哈德逊河上空表演，但重点似乎是围绕美国核心价值观及其挑战。

Every major tech layoff in 2026 that has name-checked AI

TechCrunch · 07/07 02:35

A running look — in reverse chronological order — at the bigger tech companies that have announced significant layoffs this year with AI as a stated factor.

中文介绍 TechCrunch整理了2026年至今所有以AI为裁员理由的大型科技公司裁员清单，按时间倒序排列。这份清单揭示了AI技术发展对科技行业就业市场的影响，以及企业在战略调整和人力优化中如何引用AI作为重组的考量因素。

Zhipu AI launches ZCode to challenge Claude Code and OpenAI Codex at a fraction of the cost

The Decoder · 07/07 02:28

Zhipu AI is bringing GLM-5.2 to its ZCode development environment, pitching the model's long-context capabilities for complex coding tasks. New customers get a free five-day trial with up to 5 million tokens per day, and subscribers receive about 1.5 times more token quota through July 2026. The art

中文介绍 智谱AI推出ZCode开发环境，旨在以远低于Claude Code和OpenAI Codex的成本，挑战它们在代码生成领域的地位。ZCode集成了GLM-5.2模型，提供长上下文处理复杂编码任务的能力，新用户可获得5天每天500万token的免费试用，订阅用户token配额增加1.5倍。

Amazon competitor Bookshop.org says Kobo eReader support will happen this year after all

TechCrunch · 07/07 02:12

Bookship.org seemed to delay this anticipated partnership again, but tells TechCrunch that it has settled business terms and is working on integration.

Tencent releases Hy3 open-source model that allegedly matches models up to five times its active size

The Decoder · 07/07 02:03

Tencent has released Hy3, an open-source language model with 295 billion parameters built on a mixture-of-experts architecture. Only 21 billion parameters are active at any given time. Tencent says Hy3 matches models two to five times its size while cutting its hallucination rate in half to 5.4 perc

中文介绍 腾讯发布了开源语言模型Hy3，该模型拥有2950亿参数，基于专家混合架构，但任意时刻仅激活210亿参数。腾讯宣称，Hy3的性能可媲美其活跃参数量两到五倍的模型，并显著降低了幻觉率，显示出其在效率与准确性上的优势。

Your family’s $300 stake in OpenAI

MIT Tech Review AI · 07/07 02:00

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. OpenAI CEO Sam Altman’s oft-discussed promise that Americans will share in the wealth AI creates was in the news again last week. On Thursday, the Financial Times

中文介绍 MIT科技评论文章探讨了OpenAI首席执行官Sam Altman关于美国民众将分享AI创造财富的承诺。该话题上周再次引发关注，文章可能深入分析了“你的家庭在OpenAI拥有300美元股份”这一概念的经济与社会影响，以及实现此愿景的潜在机制和挑战。

NRC is (sort of) getting rid of "as low as reasonably achievable" standard

Ars Technica · 07/07 01:48

Its issues with current nuclear safety standards are termed semantic, not physical.

中文介绍 美国核管理委员会（NRC）计划对核安全法规进行微小调整，涉及“尽可能合理达到低水平”（ALARA）标准。尽管这些调整被描述为语义而非物理上的改变，但它们可能对核工业的辐射防护实践产生影响，引发了对核安全政策演变的关注。

Five questions for Dr. Rubin, who’s armed with a mic and a bowtie

The Verge · 07/07 01:15

Dr. Rubin’s feed Bullshit is cheap but truth is expensive. Anyone with half a brain cell can post wild misinformation that goes mega viral, which wastes the time and expertise of highly trained people who feel an obligation to inform others of the truth. Today I want you to meet one of those highly-

中文介绍 这篇文章对扎卡里·鲁宾医生进行了采访，他以佩戴领结和手持麦克风的形象而闻名。鲁宾医生致力于对抗网络上的错误信息，强调“谎言廉价而真相昂贵”，呼吁专业人士履行告知公众真相的责任，旨在纠正被广泛传播的虚假信息。

GPT-4's dominance lasted a year while today's top models barely survive seven weeks at the top

The Decoder · 07/07 01:14

OpenAI's GPT-4 led the Epoch Capabilities Index for about a year, far longer than any model since. Since Claude 3 Opus took the top spot in February 2024, the lead has changed hands 17 times, with a median stay of just seven weeks. Competition is fiercer now, but the capability gains between models

中文介绍 报告指出，OpenAI的GPT-4在“Epoch能力指数”榜首保持了约一年，远超后续模型。自2024年2月Claude 3 Opus登顶以来，榜首已易主17次，模型平均仅在榜首停留七周。这表明大模型领域竞争日趋激烈，技术迭代速度显著加快。

asgeirtj/system_prompts_leaks

JavaScript · ★ 51,337 · 🍴 8,371 · 📈 1,386 stars today

Extracted system prompts from Anthropic - Claude Fable 5, Opus 4.8, Claude Code, Claude Design. OpenAI - ChatGPT 5.5 Thinking, GPT 5.5 Instant, Codex. Google - Gemini 3.5 Flash, 3.1 Pro, Antigravity. xAI - Grok, Cursor, Copilot, VS Code, Perplexity, and more. Updated regularly.

中文介绍 system_prompts_leaks 汇集了 Anthropic (如 Claude Fable 5, Opus 4.8)、OpenAI (如 ChatGPT 5.5 Thinking, GPT 5.5 Instant) 和 Google (如 Gemini 3.5 Flash, 3.1 Pro) 等主流大语言模型被“泄露”的系统 Prompt。此项目为研究者、开发者和安全专家提供了深入了解这些模型内部指令和行为模式的宝贵资源，有助于进行 prompt 工程优化和安全分析。

addyosmani/agent-skills

Shell · ★ 70,682 · 🍴 7,656 · 📈 1,114 stars today

Production-grade engineering skills for AI coding agents.

中文介绍 为 AI 编码代理提供生产级的工程技能，旨在提升 AI 代理在软件开发中的实际能力和可靠性。它可能包含代码生成、测试、重构等实用工具和方法，帮助开发者构建更高效、更可靠的智能编程助手。适用于 AI 工程师和软件开发团队。

Zackriya-Solutions/meetily

Rust · ★ 19,166 · 🍴 1,944 · 📈 2,493 stars today

Privacy first, AI meeting assistant with 4x faster Parakeet/Whisper live transcription, speaker diarization, and Ollama summarization built on Rust. 100% local processing. no cloud required. Meetily (Meetly Ai - https://meetily.ai) is the #1 Self-hosted, Open-source Ai meeting note taker for macOS &

中文介绍 Meetily 是一款注重隐私的 AI 会议助手，基于 Rust 构建，支持 100% 本地化处理，无需云服务。它利用 Parakeet/Whisper 实现 4 倍速实时转录，提供发言人识别功能，并使用 Ollama 进行会议总结。此工具为重视数据安全的个人和团队提供高效、私密的会议记录与整理解决方案。

ruvnet/RuView

Rust · ★ 77,425 · 🍴 10,400 · 📈 471 stars today

π RuView turns commodity WiFi signals into real-time spatial intelligence, vital sign monitoring, and presence detection — all without a single pixel of video.

中文介绍 RuView 项目创新性地将普通的 WiFi 信号转化为实时空间智能、生命体征监测和存在检测能力，完全无需使用任何视频设备。它通过分析 WiFi 信号在环境中因人体移动或呼吸等造成的微小扰动，提取出高价值的环境和生理数据。这解决了传统监控方案中隐私侵犯、硬件复杂或覆盖范围有限的问题。RuView 适用于智能家居、医疗保健（如老人跌倒预警、睡眠监测）、安防监控以及任何需要非接触式、隐私友好型人体感知的应用场景。

Leonxlnx/taste-skill

JavaScript · ★ 58,795 · 🍴 4,005 · 📈 1,453 stars today

Taste-Skill - gives your AI good taste. stops the AI from generating boring, generic slop

中文介绍 Taste-Skill 旨在解决大型语言模型 (LLM) 生成内容平庸、缺乏个性化的问题。它通过为 AI 提供“品味”和“技能”，帮助模型摆脱生成乏味、通用化文本的倾向。项目可能包含一系列提示工程技巧、风格指南或后处理模块，旨在提升 AI 输出的创造性和独特性。适用于内容创作者、营销人员、开发者等，需要 AI 生成更具吸引力、非同质化内容的场景，例如文案创作、故事生成或独特的产品描述。

alirezarezvani/claude-skills

Python · ★ 21,097 · 🍴 2,836 · 📈 611 stars today

345 Claude Code skills & agent skills & plugins (30+ Agents, 70+ custom commands, 330+ skills, customizable references, scripts)for Claude Code, Codex, Gemini CLI, Cursor, and 8 more coding agents — engineering, marketing, product, compliance, C-level advisory, research, business operations, commerc

中文介绍 alirezarezvani/claude-skills 是一个包含 337 个 Claude Code 技能、代理技能及插件的庞大集合。它为 Claude Code、Codex、Gemini CLI、Cursor 等十余种 AI 编码代理提供超过 30 个代理、70 个自定义命令和 330 多个技能，并支持自定义引用和脚本。此项目极大地丰富了 AI 编码助手的工具箱，旨在提升开发者在多平台上的编程效率和自动化水平。

openai/codex-plugin-cc

JavaScript · ★ 26,221 · 🍴 1,571 · 📈 910 stars today

Use Codex from Claude Code to review code or delegate tasks.

中文介绍 `codex-plugin-cc` 是一个 OpenAI 官方插件，旨在将 Codex 模型的功能集成到 Claude Code 环境中。它允许用户在 Claude Code 内部直接调用 Codex 的能力，执行代码审查或将特定编程任务委派给 Codex 处理。该插件解决了在不同 AI 模型之间切换的效率问题，为需要利用 Codex 强大代码理解和生成能力的用户提供便利，特别适合开发者进行自动化代码质量检查或智能任务分配。

mvanhorn/last30days-skill

Python · ★ 49,675 · 🍴 4,139 · 📈 511 stars today

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary

中文介绍 这是一个 AI 代理技能，能迅速在 Reddit、X、YouTube、Hacker News、Polymarket 及整个互联网上研究任意话题，并综合生成一份基于事实的摘要。主要用于获取过去 30 天的最新信息，解决信息过载问题，帮助用户快速了解热点动态和事件。适合研究人员和内容创作者。

ogulcancelik/herdr

Rust · ★ 12,786 · 🍴 744 · 📈 783 stars today

agent multiplexer that lives in your terminal.

中文介绍 herdr 是一个命令行工具，它充当 AI Agent 的多路复用器，允许用户在终端中高效管理和交互多个智能代理。通过提供一个统一的接口，它解决了同时运行和协调多个 Agent 时的复杂性，例如不同 Agent 的输入、输出和状态管理。开发者和AI工程师可以使用 herdr 来构建复杂的自动化工作流，同时监控不同 Agent 的执行情况，无需切换多个应用或窗口，极大地提升了Agent驱动型应用的开发和调试效率。

bradautomates/claude-video

Python · ★ 4,135 · 🍴 593 · 📈 539 stars today

Give Claude the ability to watch any video. /watch downloads, extracts frames, transcribes, hands it all to Claude.

中文介绍 该项目赋予 Claude AI 模型观看和理解视频的能力。它通过下载视频、提取关键帧和生成文字转录，将处理后的内容提供给 Claude 进行分析。解决了大型语言模型原生不支持视频输入的问题，使用户能利用 Claude 进行视频摘要、内容问答或深度分析。适用于 AI 开发者和视频内容分析场景。

karakeep-app/karakeep

TypeScript · ★ 26,834 · 🍴 1,318 · 📈 178 stars today

A self-hostable bookmark-everything app (links, notes and images) with AI-based automatic tagging and full text search

中文介绍 Karakeep 是一个可自托管的“万物收藏”应用，用户可以保存链接、笔记和图片等多种内容。它集成了基于 AI 的自动标签功能和全文搜索，解决了信息碎片化和查找困难的问题。用户能够私密、高效地管理个人数字资产，并通过智能分类和强大搜索快速定位所需信息。适合注重隐私的个人用户及小型团队。

firecrawl/firecrawl

TypeScript · ★ 146,115 · 🍴 8,405 · 📈 834 stars today

The API to search, scrape, and interact with the web at scale. 🔥

中文介绍 Firecrawl 提供了一个 API 接口，用于大规模地搜索、抓取和交互网页内容。它旨在解决传统爬虫在处理复杂网站、绕过反爬机制以及扩展性方面的挑战，让开发者能轻松高效地获取所需网络数据。适用于数据分析、市场研究、内容聚合及构建自动化 Web 服务等场景。

steipete/CodexBar

Swift · ★ 16,695 · 🍴 1,374 · 📈 598 stars today

Show usage stats for OpenAI Codex and Claude Code, without having to login.

中文介绍 `CodexBar` 是一款便捷工具，旨在显示 OpenAI Codex 和 Anthropic Claude Code 的 API 使用统计数据，用户无需反复登录各自平台即可实时查看。它解决了开发者和用户频繁查询 API 用量和成本的痛点，提供了一个直观、快捷的界面来监控模型消耗。这对于需要管理预算、追踪项目开销或评估 AI 代理使用效率的个人和团队尤为有用，方便他们轻松掌握 API 调用情况和相关费用。

alibaba/zvec

C++ · ★ 13,432 · 🍴 816 · 📈 355 stars today

A lightweight, lightning-fast, in-process vector database

中文介绍 Zvec 是一个轻量级、闪电般的进程内向量数据库。它解决了在应用内部高效存储和查询高维向量的痛点，无需独立部署服务。这对于需要进行快速相似度搜索的 AI 应用（如 RAG、推荐系统）至关重要，能显著降低延迟并简化架构。适合追求高性能和低开销的 AI 开发者。

sindresorhus/awesome

★ 482,211 · 🍴 35,731 · 📈 393 stars today

😎 Awesome lists about all kinds of interesting topics

中文介绍 这是一个广受欢迎的“Awesome lists”精选集，汇集了关于各种有趣主题的高质量资源列表。它旨在帮助用户快速发现和访问特定领域（如编程语言、工具、框架、设计模式等）的优质内容、工具和学习资料，解决了信息检索的效率问题。是开发者和学习者探索新领域、提升技能的理想起点。

gastownhall/gastown

Go · ★ 16,652 · 🍴 1,537 · 📈 293 stars today

Gas Town - multi-agent workspace manager

中文介绍 Gas Town 是一个多代理工作空间管理器，专注于解决在复杂任务中协调和管理多个 AI 代理的挑战。它提供一个框架或工具集，用于有效地分配任务、管理代理间的通信、共享资源和追踪工作流程。这对于构建和部署需要多个 AI 代理协同作业的系统至关重要，例如自动化复杂的业务流程、模拟多智能体行为或进行大型语言模型 (LLM) 链式推理。开发者、研究人员或企业可以通过 Gas Town 提升多代理系统的协作效率和可靠性。

PorTAL: Portable Task Adapters for LLMs

@RampLabs · 13.4K 粉丝 · 335.4K 阅 · 507 赞 · 45 转

07/02 02:09

Researcher: Ben Geist Abstract Parameter-efficient fine-tuning (e.g. LoRA) adapts a frozen LLM to a task, but the resulting adapter is locked to one base model. When a new model is released, the

中文介绍 「PorTAL」是一项新研究，旨在解决现有参数高效微调（如 LoRA）的适配器无法在不同基础 LLM 模型间移植的局限。该研究提出可移植任务适配器，使用户无需在新模型发布时重新训练，提高了 LLM 微调的灵活性和效率。

The AI Economy: The Next Chapter

@rickyho_1989 · 9.7K 粉丝 · 296.6K 阅 · 508 赞 · 69 转

06/30 20:23

Part I: The Economics of Intelligence Why the AI industry is about to optimize for intelligence per dollar rather than intelligence itself I have become increasingly convinced that the artificial

中文介绍 博主分析了AI经济的未来发展趋势。他认为，AI行业将从追求「智能本身」转向优化「单位成本智能」，即更注重智能的性价比。这预示着AI应用将迎来一个更注重效率和成本效益的新阶段。

Getting started with loops

@ClaudeDevs · 545.1K 粉丝 · 252.4K 阅 · 2.9K 赞 · 239 转

07/07 03:08

There’s a lot of talk right now about "designing loops" instead of prompting your coding agent. If you spend some time on X trying to pin down what a loop actually is, you'll come across multiple

中文介绍 ClaudeDevs 针对 AI 开发中的「循环」概念提供了入门指导。推文可能探讨了如何在 AI 代理设计中应用循环结构，以实现迭代优化或复杂任务处理，旨在帮助开发者理解这一基础工作流。

How To Master Fable (Fundamentals Guide)

@milesdeutscher · 671.1K 粉丝 · 243.8K 阅 · 500 赞 · 73 转

07/02 02:13

TL;DR: Everything you need to do to get maximum value from Fable. I guarantee that after you're done reading, you'll have all the necessary tools to quite literally 10x your AI productivity with

中文介绍 分享 Fable 工具的入门指南，旨在帮助用户充分利用 Fable 平台，实现 AI 生产力提升。内容涵盖 Fable 的核心功能和使用技巧，助力用户高效完成 AI 相关任务，最大化工具价值。

The Fable Loop Library: 25 Workflows on Autopilot

@EXM7777 · 123.0K 粉丝 · 208.2K 阅 · 525 赞 · 47 转

07/04 23:43

i'm going to teach you how to run Fable 5 on autopilot, using my own library of loops and goals... 25 workflows, each with a prompt and the exact tool it plugs into the method follows karpathy's

中文介绍 博主发布「Fable Loop Library」，提供 25 个 Fable 自动化工作流。该库旨在帮助用户实现 Fable 操作的自动驾驶，有效提升工作效率，简化复杂任务处理。

Career advice in the age of AI

@philhchen · 9.1K 粉丝 · 179.1K 阅 · 516 赞 · 34 转

07/03 05:25

AI models get better at anything you can write a loss function for, and school is mostly loss functions: well-defined problems graded against known answers. Therefore, the valuable work of the next

中文介绍 该帖子围绕AI时代下的职业发展提供建议。博主旨在分享面对人工智能技术进步时，个人应如何规划职业路径、提升自身价值，以适应未来就业市场的变化和挑战。

Your AI, your growth

@arthurmensch · 74.0K 粉丝 · 162.7K 阅 · 539 赞 · 76 转

07/04 05:31

Of course you need to use open-source models if you’re an enterprise leader. Close model providers, that are now forcing data retention, are gaining immense leverage on your business if you don’t. As

中文介绍 该推文仅以“Your AI, your growth”作为预览内容，未提供具体细节或论证，因此无法生成详细摘要。推测博主旨在强调AI对个人或组织成长的重要性，但具体观点、方法或案例均未提及。

Wiki Memory

@hwchase17 · 115.7K 粉丝 · 129.7K 阅 · 500 赞 · 57 转

06/30 22:26

Memory for agents is still early, with little to no standards. “Memory” means something different to everyone. But one common pattern is emerging: wiki memory. The idea is simple: use an agent to turn

中文介绍 讨论 AI 智能体（agent）记忆机制的现状，指出目前缺乏统一标准。博主提出一种正在兴起的「wiki memory」模式，即利用智能体将信息整理成 wiki 形式作为记忆，为智能体提供了结构化、易于检索的记忆解决方案。

Toward Unrestricted Intelligence: Venice Series A

@ErikVoorhees · 908.5K 粉丝 · 124.8K 阅 · 529 赞 · 103 转

07/01 22:28

“If others possess your thoughts and constrain your words, then you exist at their permission, and you are not free.” —Anonymous Venice launched just over two years ago to create a private and

中文介绍 Venice宣布完成A轮融资，其产品旨在提供私密且不受限制的智能服务，以应对言论受限及思想被束缚的担忧。该项目两年前启动，专注于赋能用户掌控自己的数字存在与表达。

NEvo: Neural-Guided Evolutionary Video Synthesis for Dynamic Visual Selectivity

👍 0

07/02 23:27

The human brain processes dynamic visual input through hierarchically organized, functionally specialized regions. While recent in silico brain encoding models can synthesize optimal stimuli to probe selectivity in different brain regions, prior work has been largely limited to static images, leavin

中文介绍 一项名为「NEvo」的神经引导演化视频合成模型被提出，旨在模拟人脑处理动态视觉输入。该模型能够合成最佳刺激，以探测大脑不同区域的视觉选择性。这项研究关注于大脑的层级组织和功能特化区域，为深入理解大脑如何处理动态视觉信息提供工具。

Pre-Flight: A Benchmark for Evaluating Large Language Models on Aviation Operational Knowledge

👍 0

07/02 15:49

Large language models (LLMs) are increasingly proposed for aviation business operations, from documentation and training generation to customer facing assistants. General purpose benchmarks do not measure whether a model reasons safely and correctly about aviation specific operational knowledge, and

中文介绍 「Pre-Flight」是一项新提出的基准，用于评估大型语言模型（LLMs）在航空运营知识方面的表现。鉴于LLMs在航空业务中应用渐广，此基准旨在衡量模型能否安全且正确地理解和推理航空特定操作，填补了现有通用基准无法有效评估专业领域能力的空白。

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

👍 26

07/02 08:00

We present WorldDirector, a highly controllable video world model framework designed for persistent dynamic object memory and unrestricted viewpoint exploration. Unlike existing world models that entangle physical dynamics with pixel rendering and rely on continuous visual observation to sustain mot

中文介绍 WorldDirector是一个高度可控的视频世界模型框架，旨在实现持久动态对象记忆和无限制视角探索。它区别于现有将物理动态与像素渲染纠缠在一起的世界模型，有效解决了传统模型对连续视觉观察的依赖问题，从而提升了视频生成和模拟的效率与真实感。

Program-as-Weights: A Programming Paradigm for Fuzzy Functions

👍 98

07/02 08:00

Many everyday programming tasks resist clean rule-based implementation, such as alerting on important log lines, repairing malformed JSON, or ranking search results by intent, and are increasingly outsourced to large language model APIs at the cost of locality, reproducibility, and price. We propose

中文介绍 “Program-as-Weights”是一种新的编程范式，专为难以用清晰规则实现的模糊功能设计。该范式能处理日志警报、损坏JSON修复或按意图排序搜索结果等任务，旨在提供一种本地且可复现的解决方案，以替代日益依赖大型语言模型API但存在局限性的传统方法。

From SRA to Self-Flow: Data Augmentation or Self-Supervision?

👍 11

07/02 08:00

Representation alignment has become an effective way to accelerate diffusion transformer training and improve generation quality. Recent self-alignment methods, such as SRA and Self-Flow, further remove the dependency on external pretrained encoders by constructing alignment within the diffusion mod

中文介绍 表示对齐已成为加速扩散Transformer训练和提升生成质量的有效方法。SRA和Self-Flow等新型自对齐方法，通过构建内部对齐，进一步消除了对外部预训练编码器的依赖。这项研究探讨了这些方法是属于数据增强还是自监督学习的范畴，及其对模型训练的深远影响。

VT-WAM: Visual-Tactile World Action Model for Contact-Rich Manipulation

👍 0

07/02 08:00

Contact-rich manipulation requires policies to react to local deformation, pressure, slip, and friction, yet these cues are temporally sparse and often invisible in visual observations. Existing visual-tactile policies usually feed tactile observations directly into action prediction, but rarely mod

中文介绍 论文「VT-WAM: Visual-Tactile World Action Model for Contact-Rich Manipulation」提出VT-WAM模型，旨在提升机器人处理接触式操作的性能。这类操作需对局部形变、压力、滑动和摩擦等线索做出反应，但这些信息在视觉上往往稀疏且不可见。现有策略直接利用触觉观察预测动作，VT-WAM有望改进此问题。

Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots

👍 35

07/02 08:00

Embodied AI models now span vision-language-action (VLA) models and world-action models (WAMs), but practical deployment remains fragmented across model-specific Python stacks, backend assumptions, and robot-side glue code, especially on heterogeneous edge devices. Existing inference runtimes are de

中文介绍 「Embodied.cpp」是一个为具身AI模型（如VLA和WAMs）设计的便携式推理运行时。它旨在解决具身AI模型在异构机器人和边缘设备上部署碎片化的问题，这些问题源于模型特定的Python堆栈和后端差异。该运行时提供了一个统一高效的解决方案，以促进具身AI模型的广泛实际应用。

EAGLE-360: Embodied Active Global-to-Local Exploration in 360^circ

👍 1

07/02 08:00

While Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in standard visual understanding, adapting them for active visual search in 360^circ panoramic environments exposes fundamental limitations. Specifically, standard MLLMs struggle to effectively model inherent p

中文介绍 尽管多模态大型语言模型（MLLMs）在标准视觉理解方面表现出色，但将其应用于360度全景环境中的主动视觉搜索时，暴露出根本局限。MLLMs难以有效进行全局到局部的探索。EAGLE-360旨在解决此问题，提供一种在360度环境中进行具身主动全局到局部探索的新方法，以提升模型在该复杂场景下的表现。

Interpretation-Oriented Cloud Removal via Observation-Anchored Residual Flow with Geo-Contextual Alignment

👍 4

07/02 08:00

Cloud removal (CR) is essential for optical remote sensing, serving as a prerequisite for reliable downstream interpretation, such as semantic segmentation and change detection. However, existing CR approaches often prioritize visual realism while overlooking their impact on subsequent analytical ta

中文介绍 云去除（CR）是光学遥感中的关键技术，对于语义分割和变化检测等下游判读任务至关重要。现有CR方法常侧重于视觉真实感，但往往忽视了其对后续判读效果的影响。一篇论文提出了一种「面向判读的云去除」新方法，通过「观测锚定残差流」和「地理上下文对齐」技术，旨在更有效地去除云层，同时提升遥感图像在语义分割和变化检测等应用中的判读准确性。

Learning to Move Before Learning to Do: Task-Agnostic pretraining for VLAs

👍 6

07/02 08:00

Vision-Language-Action (VLA) models are fundamentally bottlenecked by the scarcity of expert demonstrations -- triplets of observations, instructions, and actions that are costly to collect at scale. We argue that this bottleneck stems from conflating two distinct learning objectives: acquiring phys

中文介绍 视觉-语言-动作（VLA）模型受限于专家示范数据稀缺，即观察、指令和动作三元组的收集成本高昂。论文提出“先学移动，再学做事”的理念，即对VLA模型进行任务无关的预训练。研究认为，当前瓶颈源于混淆了两个不同的学习目标，通过分离运动和任务学习可有效提升VLA模型的泛化能力。

OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers

👍 25

07/02 08:00

Diffusion transformers (DiTs) achieve state-of-the-art image and video generation, but their multi-step sampling and growing parameter count make inference expensive. Post-training quantization (PTQ) is the natural remedy, yet DiT activations shift across timesteps, prompts, and guidance branches, f

中文介绍 扩散Transformer（DiTs）在图像和视频生成方面表现卓越，但其多步采样和日益增长的参数量导致推理成本高昂。训练后量化（PTQ）是常见的优化方案，但DiT的激活值在不同时间步和提示下会发生偏移，给量化带来挑战。OrbitQuant提出了一种数据无关的量化方法，旨在高效优化DiT模型。

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

👍 45

07/02 08:00

Autonomous agents are increasingly expected to improve executable policies through feedback, yet existing evaluations often collapse this process into a final score or confound it with open-ended software-engineering progress. We introduce Autonomous Policy Evolution, a controlled evaluation setting

中文介绍 自主智能体被期望通过反馈改进可执行策略，但现有评估常将其简化为最终分数，或与软件工程进展混淆。论文介绍了“自主策略演化”这一受控过程。EvoPolicyGym是一个新基准，专门用于在交互式环境中评估自主策略的演化能力，旨在提供更精细、全面的评估方式，以推动智能体发展。

LIME: Learning Intent-aware Camera Motion from Egocentric Video

👍 0

07/02 08:00

Autonomous robots often need to move their camera before they can act: to inspect an object, reveal an occluded region, or obtain a view that responds to a user's intent. While vision-language navigation translates instructions to base motion and vision-language-action policies map instructions to m

ACID: Action Consistency via Inverse Dynamics for Planning with World Models

👍 0

07/02 08:00

Decision-time planning with action-conditioned world models has become a popular paradigm for embodied control. However, the standard planning cost judges a candidate solely by how close its predicted terminal state lies to the goal, leaving the realizability of the intermediate transitions unchecke

Representation Distribution Matching for One-Step Visual Generation

👍 8

07/02 08:00

We elucidate the design space of Representation Distribution Matching (RDM), our name for the paradigm that trains a one-step image generator by matching generated and reference feature distributions under frozen pretrained encoders. We identify two design axes, how the distributions are compared an

The Moving Eye: Enhancing VLA Spatial Generalization via Hybrid Dynamic Data Collection

👍 0

07/02 08:00

Vision-Language-Action (VLA) models have shown remarkable promise in generalized robotic manipulation. However, their spatial generalization remains fragile. We argue that simply increasing the number of viewpoints is insufficient. Models often fall into the trap of Shortcut Learning, latching onto

Optimizing Visual Generative Models via Distribution-wise Rewards

👍 15

07/02 08:00

Conventional reinforcement learning strategies for visual generation typically employ sample-wise reward functions, yet this practice frequently results in reward hacking that degrades image diversity and introduces visual anomalies. To address these limitations, we present a novel framework that fi

AGVBench: A Reliability-Oriented Benchmark of Data Augmentation for Vein Recognition

👍 15

07/02 08:00

Vein recognition is a secure biometric technology often constrained by limited annotated data and imaging variations. While data augmentation mitigates this, strategies designed for natural images may disrupt the fine-grained topology and textures essential for identity discrimination. We present AG

AnyGroundBench: A Specialized-Domain Benchmark for Video Grounding in Vision-Language Models

👍 8

07/02 08:00

Vision-Language Models (VLMs) have demonstrated immense promise in Spatio-Temporal Video Grounding (STVG). However, current evaluation protocols are largely confined to zero-shot assessments on general, daily-life benchmarks. This creates a critical disconnect from real-world applications in special

BamiBERT: A New BERT-based Language Model for Vietnamese

👍 1

07/02 08:00

In this paper, we introduce BamiBERT, a new BERT-based pre-trained language model for Vietnamese that addresses key limitations of PhoBERT -- the current de facto Vietnamese text encoder. Trained from scratch on a 129GB corpus of general-domain Vietnamese text for 20 epochs, BamiBERT supports an ext

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

👍 55

07/02 08:00

Memory for a long-horizon LLM agent is a contract about what each future decision is allowed to see. The simplest contract appends past observations, tool calls, and reflections to every prompt, which makes prior context easy to access but also turns it into a jumbled mixture in which the effect of

ArcAD: Anomaly-Rectified Calibration for Cold-Start Supervised Anomaly Detection

👍 0

07/02 08:00

The deployment of Industrial Anomaly Detection (IAD) in real-world manufacturing frequently encounters a challenging cold-start bottleneck, in which limited normal samples fail to represent the full normal distribution and only a few anomalies are available. Under such a regime, existing methods str

Bridge-WA: Predicting Where and How the World Changes for Robotic Action

👍 0

07/02 08:00

General-purpose vision-language-action models benefit from large vision-language priors, but effective manipulation also requires anticipating action-relevant scene changes. Existing world-action models often rely on large generative world models or dense future rollouts, which are expensive and spe

DeepGaze3.5-VL: Modeling Scanpaths via Autoregressive Token Prediction

👍 0

07/02 08:00

Understanding human visual attention on a scene over time has applications in domains such as interface design and inferring cognitive states. Modeling visual scanpaths has historically relied on specialized architectures with hand-crafted priors. While these architectures can model fixation sequenc

HandsOnWorld: Unconstrained Egocentric Video Generation with Camera-Disentangled Hand Control

👍 0

07/02 08:00

We present HandsOnWorld, a framework for hand-controlled egocentric video generation that forgoes multi-view and marker-based motion capture, learning instead from unconstrained monocular video. Such generality is bottlenecked by the scarcity of scalable 3D hand annotations: large egocentric corpora

PACE: A Proxy for Agentic Capability Evaluation

👍 10

07/02 08:00

Evaluating LLM agents on benchmarks like SWE-Bench and GAIA can be expensive, time-consuming, and requires complex infrastructure. A single evaluation can cost thousands of dollars and take days to complete. In contrast, non-agentic LLM benchmarks that test individual capabilities (e.g., reasoning,

SkillCoach: Self-Evolving Rubrics for Evaluating and Enhancing Agentic Skill-Use

👍 17

07/02 08:00

Skills are becoming a reusable operational layer for LLM agents, encoding SOPs, domain rules, tool workflows, scripts, and validation routines. In realistic skill repositories, overlapping skills make reliable skill-use difficult. Final verifier success is too coarse for both evaluation and training

VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon

👍 22

07/02 08:00

Vision-Language-Action (VLA) foundation models have recently achieved strong progress in embodied intelligence. To reduce policy-call frequency while preserving temporal coherence, most generative policies adopt an action chunk mechanism, executing multiple future actions in an open-loop manner unde

Denser neq Better: Limits of On-Policy Self-Distillation for Continual Post-Training

👍 7

07/02 08:00

Continual post-training enables foundation models to acquire new knowledge while preserving existing capabilities. Recent work suggests that on-policy learning can mitigate forgetting, with on-policy self-distillation emerging as a particularly attractive approach. In this work, we revisit this opti

LASER: A Corrective Lens for LVLMs via Visual Attention Preservation and Sink Suppression

👍 1

07/02 08:00

Large vision-language models (LVLMs) exhibit strong reasoning ability but suffer from visual forgetting during long-horizon decoding, where attention progressively drifts away from visual evidence. Existing methods largely treat this issue as a late-stage attention decay problem or attempt to mitiga

Octolens

07/07 05:17

Social listening for the agent era

中文介绍 Octolens是一款专为“agent时代”设计的AI产品，专注于社交聆听服务。它旨在帮助用户监控和分析社交媒体上的对话和趋势，从而为基于AI代理的应用提供深入的市场洞察和用户反馈。这款工具将社交数据分析与AI代理技术相结合，提升信息获取和处理效率。

Typeahead 2.0

07/07 05:16

Private AI autocomplete for every app on your Mac

中文介绍 Typeahead 2.0是一款为Mac用户设计的私有AI自动补全工具。它能为Mac上运行的任何应用程序提供智能文本预测功能，旨在提高用户输入效率，同时强调数据的本地处理和隐私保护。这款应用将AI技术融入日常操作，提升Mac的使用体验。

Stanley Studio

07/07 05:14

The AI video editor you hire that edits like a human

中文介绍 Stanley Studio是一款基于人工智能的视频编辑工具，其核心优势在于能够像人类专业编辑一样进行视频处理。用户可以“雇佣”这款AI编辑器来完成复杂的视频剪辑任务，旨在提供高质量且富有创造性的编辑成果，同时显著提升效率并降低制作成本。

CodeMote

07/07 05:13

Claude Code, Codex, any CLI agent. Driven from your iPhone

中文介绍 CodeMote是一款通过iPhone远程控制AI编程代理的工具。它支持如Claude Code、Codex以及任何命令行界面（CLI）代理，使用户能够利用iPhone便捷地驱动这些AI工具执行代码生成或开发任务。该产品旨在提升开发者的移动工作效率与灵活性。

Astryx

07/07 05:13

A customizable, agent-ready open-source design system

中文介绍 Meta公司推出了Brain2Qwerty v2项目，该技术能够直接从非侵入式脑信号中解码完整的句子。这项创新研究旨在探索人机交互的新范式，通过分析大脑活动，实现无需物理输入设备即可进行文本生成，具有潜在的广泛应用前景，如辅助沟通。

Edgee Claude Code Compressor V2

07/07 05:13

Fewer tokens, same context, 50% cost reduction

中文介绍 Edgee Claude Code Compressor V2是一款专为Claude Code设计的代码压缩工具。该产品能在保持原有代码上下文完整性的前提下，显著减少所需的token数量，从而为用户带来高达50%的成本节约。它旨在优化大型语言模型在处理代码时的效率和经济性。

AirKaren

07/07 05:13

AI that fights customer service for you

中文介绍 AirKaren是一款利用人工智能技术，能代表用户与客服进行沟通和“斗争”的产品。它旨在自动化处理客户服务问题，帮助用户解决与企业客服交互时的常见困扰，从而节省用户的时间和精力，提升问题解决效率。

Sunrise

07/07 05:13

A real planner for Google Tasks

中文介绍 Sunrise是一款专为Google Tasks设计的“真实规划器”工具。它旨在增强Google Tasks的功能，提供更全面、更直观的任务管理和日程规划体验，帮助用户高效组织待办事项，提升个人和团队的工作效率。

AnySearch

07/07 05:13

Real-time structured search trusted by agents and developers

中文介绍 AnySearch是一款提供实时结构化搜索服务的产品。它旨在帮助代理和开发者快速、准确地获取所需信息，通过优化的搜索算法，确保结果的可靠性与相关性，从而提升信息检索效率，广受专业用户信赖。

WorkBuddy

07/07 05:10

Produce sharpened results faster with a team of AI experts

中文介绍 WorkBuddy 提供一个由AI专家组成的团队，旨在帮助用户更快地产生更优质的成果。该产品通过整合多个AI模型和工具，模拟专业团队协作，优化工作流程，从而提高效率和产出质量。

Why Every Designer Should Switch to Claude Fable 5

07/07 01:37

中文介绍 这则YouTube视频由Riley Brown发布，探讨了设计师为何应该转用“Claude Fable 5”。视频可能详细分析了“Claude Fable 5”这款工具或AI模型的功能、优势以及它如何提升设计工作效率，旨在说服设计师采纳此新平台。

Building a Real AI Jarvis (GPT Realtime 2 + Cursor)

07/01 09:40

中文介绍 该视频由Riley Brown发布，展示了如何利用GPT Realtime 2和代码编辑器Cursor来构建一个真实的AI助手，旨在模仿电影中的智能管家「贾维斯」。内容可能涵盖AI实时交互、编程辅助以及个人化AI系统开发等技术细节。

Poweful Agents inside Slack

06/25 06:39

AI Agents Just Changed Forever: GLM 5.2, Codex Skills, Claude & Cursor

06/21 21:30

9 AI Agent Skills To Get Ahead of 99% of People

06/18 10:27

How to build high quality internal tools with Retool

06/18 05:21

SpaceX Just Bought Cursor for $60B. It’s About to Take OVER.

06/17 05:31

Why the Government Just Killed Claude Fable 5

06/13 11:50

This is Why They Banned Claude Fable 5

06/13 02:28

Hermes Agent NEW Super-App and DeepSeek v4 Catches Up To Opus 4.8?

06/07 04:10

来自镇上的我，不知道什么是网络公司。

07/06 21:05

9 回复 · 程序员 节点

[求助] 男生嘴角两边长毛，拔不掉太捉急（忍不住要拔），求推荐工具/方法

07/06 19:41

13 回复 · 程序员 节点

嘲笑别人的 deepseek 偷跑 100 元后，我被偷跑了 2700 元

07/06 18:38

20 回复 · 程序员 节点

试用 hy3，感觉不错

07/06 18:20

20 回复 · 程序员 节点

mac 的外接显示器要完美必须 5K 吗？

07/06 15:11

48 回复 · Apple 节点

各位的梯子🪜 平均每个月流量使用多少？

07/06 12:50

74 回复 · Apple 节点

出个人闲置电脑

07/06 12:46

19 回复 · Apple 节点

[送永久激活码]RightX 增强你的 Mac Finder 右键

07/06 10:40

127 回复 · Apple 节点

周末 vibe coding 了一款高情商改写的 Mac App，名字叫 Mellow（米洛）

07/06 00:24

6 回复 · Apple 节点

为什么这段 cuda 的并行前缀算法不会数据竞争

07/05 23:28

6 回复 · Python 节点

Learning to code is still worthwhile

07/07 04:59

13 points · 0 comments

Price per 1M tokens is meaningless

07/07 03:43

86 points · 46 comments

Python 3.14 compiled to metal – no interpreter

07/07 03:40

52 points · 27 comments

CoMaps – FOSS Offline Maps

07/07 02:55

120 points · 19 comments

Linux on the Atari Jaguar

07/07 02:35

69 points · 9 comments

Pros and Cons of Solo Development

07/07 02:24

80 points · 41 comments

OpenWrt One – Open Hardware Router

07/07 02:23

224 points · 108 comments

A global workspace in language models

07/07 01:44

144 points · 41 comments

Januscape: Guest-to-Host Escape in KVM/x86 [CVE-2026-53359]

07/07 01:35

52 points · 11 comments

OfficeCLI: Office suite for AI agents to read and edit Microsoft Office files

07/07 00:47

70 points · 19 comments

Stealth robotics startup (YC S26) is hiring principal engineers (Palo Alto)

07/07 00:33

We are a new robotics company based in Palo Alto, California, building wearable robotic devices.Our first product is a device that reduces the physical cost of carrying heavy loads over distance, allowing people to move farther, carry more, and arrive less fatigued.The company is early but already o

Show HN: Pulpie – Models for Cleaning the Web

07/07 00:04

Hey HN, I'm Shreyash, founder of Feyn. We built Pulpie, a family of Pareto optimal models for cleaning the web. Pulpie strips boilerplate (ads, footers, sidebars) from raw HTML and returns just the main content as HTML or Markdown.We match SOTA extraction quality while being 20x cheaper. Cleani

Kani: A Model Checker for Rust

07/06 23:53

104 points · 5 comments

Should DayQuil Be Legal?

07/06 23:44

147 points · 191 comments

1k Words: A Writing Contest

07/06 23:30

66 points · 32 comments

CS2 Fog Of War: Server-sided anti-wallhack occlusion culling for CS2 servers

07/06 23:23

75 points · 36 comments

AMD Ryzen AI Halo – $4k AI Dev Kit

07/06 23:01

231 points · 167 comments

Resetting Xbox

07/06 22:18

368 points · 314 comments

Aluminum foil (2021)

07/06 21:28

207 points · 100 comments

Nintendo announces new product revisions in Europe with replaceable batteries

07/06 21:18

283 points · 175 comments

Fable 5 On Vending-Bench: Misbehaving, With Plausible Deniability

07/06 20:38

158 points · 112 comments

Road to Elm 1.0

07/06 19:47

279 points · 129 comments

Real-time map of Great Britain's rail network

07/06 17:38

356 points · 135 comments

Egypt Is Building a New Nile

07/04 04:00

98 points · 34 comments

Using precision editing to study human embryo development shows master gene

07/03 10:03

22 points · 3 comments

Hobbes – A Language and Embedded JIT Compiler

07/03 07:52

15 points · 2 comments

Clojure 1.13 adds support for checked keys

07/03 04:51

162 points · 36 comments

Introduction to Genomics for Engineers

07/02 20:28

208 points · 33 comments

Rotman Lens

07/01 22:56

30 points · 3 comments

Apricot Computers: An underrated British brand

07/01 15:26

70 points · 30 comments

v2.1.201

07/04 07:50

What's changed Claude Sonnet 5 sessions no longer use the mid-conversation system role for harness reminders

中文介绍 克劳德（Claude）代码库发布了 v2.1.201 版本。此版本更新的一项主要变化是，Claude Sonnet 5 的会话在进行引导提醒时，将不再使用对话中的系统角色。这反映了对模型内部交互机制的调整和优化，旨在提升模型性能和用户体验。

v2.1.200

07/04 00:52

What's changed Changed AskUserQuestion dialogs to no longer auto-continue by default; opt into an idle timeout via /config Changed the "default" permission mode to "Manual" across the CLI, --help, VS Code, and JetBrains; --permission-mode manual and "defaultMode": "manual" are accepted alongside def

中文介绍 Anthropic的Claude Code发布v2.1.200版本。此版本更新了“AskUserQuestion”对话框，默认不再自动继续，用户可通过“/config”设置空闲超时。此外，CLI、VS Code和JetBrains等环境的默认权限模式均已更改为“手动”，以提高安全性控制。

v2.1.199

07/03 07:35

What's changed Stacked slash-skill invocations like /skill-a /skill-b do XYZ now load all leading skills (up to 5), not just the first Fixed SSL certificate errors (TLS-inspecting proxies, missing NODE_EXTRA_CA_CERTS, expired certs) burning retries before showing actionable guidance — they now fail

中文介绍 Anthropic的Claude Code发布v2.1.199版本。此更新改进了堆叠斜杠技能调用功能，现在最多可加载前5个技能，而非仅第一个。同时，修复了与SSL证书相关的错误，解决了因TLS检查代理、缺少NODE_EXTRA_CA_CERTS或证书过期导致的重试问题。

v2.1.198

07/03 06:55

What's changed Subagents now run in the background by default, so Claude keeps working while they run and is notified when they finish (previously a gradual rollout) Claude in Chrome is now generally available Added background agent notifications in claude agents — sessions that need input or finish

中文介绍 Anthropic的Claude v2.1.198版本更新。Claude现已在Chrome浏览器上普遍可用，并为Claude代理新增后台通知功能，在会话需输入或结束时触发。同时，增加了"/dataviz"技能，提供图表和仪表盘设计指导。

v2.1.197

07/01 01:56

What's changed Introducing Claude Sonnet 5: now the default model in Claude Code, with a native 1M-token context window and promotional pricing of $2/$10 per Mtok through August 31. Update to version 2.1.197 for access. https://www.anthropic.com/news/claude-sonnet-5

中文介绍 Anthropic的Claude Code发布v2.1.197更新，并引入了Claude Sonnet 5作为默认模型。该模型原生支持100万上下文窗口，并提供推广期价格：每百万tokens输入2美元，输出10美元，优惠期截至8月31日。

v2.1.196

06/30 07:27

What's changed Added support for organization default models — admins set it in the org console; it shows as "Org default" (or "Role default") in /model when you haven't picked one yourself Added readable default names for sessions at start, making them easier to identify and message Added clickable

v2.1.195

06/27 05:29

What's changed Added CLAUDE_CODE_DISABLE_MOUSE_CLICKS to disable mouse click/drag/hover in fullscreen mode while keeping wheel scroll Fixed hook matchers with hyphenated identifiers (e.g. code-reviewer, mcp__brave-search) accidentally substring-matching — they now exact-match. Use mcp__brave-search_

v2.1.193

06/26 05:45

What's changed Added autoMode.classifyAllShell setting to route all Bash/PowerShell commands through the auto-mode classifier instead of only arbitrary-code-execution patterns Added auto-mode denial reasons to the transcript, the denial toast, and /permissions recent denials Added claude_code.assist

v2.1.191

06/25 05:58

What's changed Added /rewind support for resuming a conversation from before /clear was run Fixed scroll position jumping to the bottom while reading earlier output during a streaming response Fixed background agents resurrecting after being stopped — stopping an agent from the tasks panel is now pe

v2.1.190

06/24 23:53

What's changed Bug fixes and reliability improvements