2026-07-29（内容获取于 07/29 05:03）

MCP startup Runlayer accuses Rippling of stealing its product idea

TechCrunch · 07/29 04:45

Runlayer is suing Rippling after Rippling evaluated the startup's MCP gateway product and then opted to build one itself.

中文介绍 初创公司Runlayer已对Rippling提起诉讼，指控后者窃取其MCP网关产品的创意。Runlayer表示，Rippling在评估了其产品后，决定自行开发同类产品。此举引发了知识产权纠纷。

Sam Altman is ready to decelerate

TechCrunch · 07/29 04:17

His change of position comes after "the first security incident that I have felt very viscerally."

中文介绍 OpenAI首席执行官山姆·奥特曼表示，在经历了他「第一次亲身感受到的安全事件」之后，他准备放缓人工智能（AI）发展的速度。此番表态标志着他此前立场的变化，可能影响未来AI研发的节奏。

AI leaders sign statement asking the government to do something about automated AI

The Verge · 07/29 03:46

Employees of OpenAI and Anthropic, as well as Google, Meta, Thinking Machines, Microsoft, Mistral, and other leading AI labs, have written a statement to the US government supporting a potential slowdown of sorts for frontier AI development - or at least a speed-up of global coordinated governance e

中文介绍 包括OpenAI、Anthropic、谷歌、Meta、微软、Mistral等主要AI实验室的员工，已联名致函美国政府，呼吁对前沿AI发展采取潜在的放缓措施，或至少加快全球合作。

Is it illegal to trick the US government into wiping your phone during a questionably legal search?

The Verge · 07/29 03:35

A Georgia man was charged with a felony for allegedly wiping his phone while being questioned by Customs and Border Protection. Samuel Tunick had something in common with others who have had their devices searched at airports: He was an activist. Unlike others who have had their phones seized with n

中文介绍 佐治亚州男子塞缪尔·图尼克（Samuel Tunick）因涉嫌在海关及边境保护局（CBP）审讯期间擦除手机数据，被指控犯有重罪。他与其他在机场设备被搜查的活动家有相似之处，案件引发了对搜查合法性的质疑。

AI’s finally expensive enough to make Wall Street nervous

The Verge · 07/29 03:33

Working hard, or bear-ly working? | Image: Cath Virginia / The Verge, Getty Images It's earnings season, and investors got an unpleasant surprise from Google: an increase on its spending estimate, to as much as $205 billion - from the last quarter's projection of up to $190 billion. Even the lower e

中文介绍 谷歌在财报季公布了令人不安的支出预测，预计资本支出将从上季度的最高1900亿美元增至高达2050亿美元，这使得华尔街投资者对AI相关投资的成本感到担忧，引发市场紧张情绪。

This comfy gaming headset that can play audio from two sources is $25

The Verge · 07/29 03:18

The Hybrid adds Bluetooth functionality to the wired H3, seen here. | Photo by Cameron Faulkner / The Verge While most gaming headsets have moved towards low-latency wireless connections, there’s something to be said for a budget-friendly, capable wired headset. The EPOS H3 Hybrid is on sale for jus

中文介绍 EPOS H3 Hybrid是一款售价25美元的有线游戏耳机，它在H3有线型号的基础上增加了蓝牙功能，使其能够同时播放来自两个音源的音频。尽管多数游戏耳机转向无线连接，这款高性价比产品仍提供出色的功能。

Logitech will pull a Nintendo — only European mice will come with replaceable batteries

The Verge · 07/29 03:12

In June, Nintendo announced a new version of the Switch 2 that should let you easily replace its battery pack - but only for Europe. Logitech will do much the same. User-replaceable batteries will be the law in Europe in February 2027, but instead of bringing them to the rest of the world, Logitech

中文介绍 罗技宣布将效仿任天堂，其可更换电池产品将仅限于欧洲市场销售。根据欧盟新法规定，自2027年2月起，欧洲销售的设备必须支持用户自行更换电池，但罗技不打算将此功能推广至其他地区。

Anthropic says its Mythos model found vulnerabilities in cryptographic algorithms that secure the internet

The Decoder · 07/29 03:12

Anthropic's Claude Mythos Preview found weaknesses in key cryptographic algorithms, including a better attack on HAWK, a post-quantum signature scheme that human experts had reviewed for more than two years. The model found it in just 60 hours at an API cost of about $100,000. The findings don't aff

中文介绍 Anthropic公司宣布，其Claude Mythos Preview模型发现了互联网关键加密算法的漏洞，包括针对后量子签名方案HAWK的更优攻击方法。该模型在仅60小时内以约10万美元的API成本完成了人类专家两年多未能发现的工作。

Ozlo’s Sleepbuds 2 build on Bose’s sleep earbud legacy

TechCrunch · 07/29 03:09

Ozlo’s first major update to its sleep earbuds introduces longer battery life, improved connectivity, enhanced audio, and new sleep features as the startup continues the product line once abandoned by Bose.

中文介绍 Ozlo公司发布了其睡眠耳塞Sleepbuds 2的首次重大更新，引入了更长的电池续航、改进的连接性、增强的音频以及新的睡眠功能。这款产品旨在延续并发展Bose此前放弃的睡眠耳塞产品线。

The robot NASA hired to lift a orbital telescope is tumbling out of control

TechCrunch · 07/29 03:07

According to NASA, two of the three reaction wheels that control the spacecraft's alignment have failed, and there are problems with one of the spacecraft's thruster systems.

中文介绍 美国国家航空航天局（NASA）报告称，其用于举升轨道望远镜的机器人目前正失控翻滚。据NASA透露，控制航天器姿态的三个反作用轮中有两个已失效，并且航天器的一个推进器系统也存在故障。

Waymo, robotaxi operators face fresh scrutiny over emergency response failures

TechCrunch · 07/29 03:06

Rep. Kevin Mullin (D-California ) has proposed a bill that would direct federal regulators to establish minimum national safety standards for autonomous vehicle operators.

中文介绍 Waymo等自动驾驶出租车运营商因紧急响应故障面临新的审查。加州民主党众议员凯文·穆林（Kevin Mullin）已提出一项法案，旨在指示联邦监管机构为自动驾驶汽车运营商制定最低国家安全标准。

Apple won’t turn on any ‘restricted mode’ for missed lease payments

The Verge · 07/29 02:43

Apple says it won't limit the capabilities of devices leased through its new Upgrade program if you miss a payment. In an emailed statement to The Verge, Apple spokesperson Brian Bumbery says, "There will be no restricted mode and/or there will be no limitations put on device functionality due to mi

中文介绍 苹果公司表示，通过其新版“升级计划”租赁的设备，即使用户错过付款，也不会限制设备功能或启用任何“受限模式”。苹果发言人布莱恩·邦伯里通过电子邮件证实了这一点。

New airliner sets record flying 24 hrs nonstop from Australia to France

Ars Technica · 07/29 02:43

Record flight shows viability of Qantas Airways’ Project Sunrise routes in 2027.

中文介绍 一架新型客机创造了新的飞行纪录，完成了从澳大利亚到法国的24小时不间断飞行。此次创纪录的航班展示了澳洲航空（Qantas Airways）“日出计划”在2027年开通超长航线的可行性。

eBay reaches $56M settlement with e-commerce newsletter writers it terrorized in 2019

TechCrunch · 07/29 02:35

Ina and David Steiner inspired the ire of high-level eBay executives after occasionally criticizing the company in their newsletter. In 2019, a plot was concocted to intimidate the couple into halting their negative coverage.

中文介绍 eBay已与2019年受到公司高管恐吓的电子商务新闻通讯作者伊娜（Ina）和大卫·施泰纳（David Steiner）达成5600万美元的和解协议。施泰纳夫妇因批评eBay而遭到报复，高管曾策划恐吓他们停止负面报道。

Judge blocks first state law that would have banned prediction markets

Ars Technica · 07/29 02:31

Judge may allow Minnesota to ban some types of bets, but not all of them.

中文介绍 一名法官裁定阻止了首个旨在全面禁止预测市场的州法律生效。该法案来自明尼苏达州，法官表示虽然可能允许该州禁止部分类型的博彩，但并非所有预测市场活动都可被禁止。

pascalorg/editor

TypeScript · ★ 18,584 · 🍴 2,522 · 📈 415 stars today

Create and share 3D architectural projects.

中文介绍 pascalorg/editor 是一个专为建筑项目设计的 3D 编辑器，旨在帮助用户高效创建和分享三维建筑设计。它简化了复杂的建模流程，使用户能够快速构建建筑模型、进行可视化呈现并与他人协作。该工具特别适合建筑师、室内设计师、城市规划师以及相关专业的学生，可用于概念设计、方案演示、项目评审和教育实践，提升设计效率与沟通质量。

jenkinsci/jenkins

Java · ★ 26,047 · 🍴 9,685 · 📈 180 stars today

Jenkins automation server

中文介绍 Jenkins 是一款广泛使用的开源自动化服务器，它专注于实现软件开发的持续集成（CI）和持续交付（CD）。该项目通过自动化构建、测试、部署等环节，解决了传统软件开发流程中手动操作耗时且易出错的问题，极大地提高了开发效率和软件质量。它适用于任何需要自动化软件发布流程的开发团队和 DevOps 工程师。

moeru-ai/airi

TypeScript · ★ 44,702 · 🍴 4,443 · 📈 796 stars today

💖🧸 Self hosted, you-owned Grok Companion, a container of souls of waifu, cyber livings to bring them into our worlds, wishing to achieve Neuro-sama's altitude. Capable of realtime voice chat, Minecraft, Factorio playing. Web / macOS / Windows supported.

中文介绍 Airi 是一个自托管的 AI 伴侣项目，旨在创建用户私有化、可定制的虚拟角色（如“waifu”），并支持实时语音聊天，目标是达到 Neuro-sama 的交互水平。它被设计为一个类似 Grok Companion 的系统，让用户能够完全掌控自己的 AI 互动体验，甚至可能与 Minecraft 等游戏集成。

andrewyng/aisuite

Python · ★ 15,654 · 🍴 1,654 · 📈 92 stars today

Simple, unified interface to multiple Generative AI providers

中文介绍 aisuite 项目由知名 AI 专家 Andrew Ng 推出，旨在提供一个简洁统一的界面，用于访问和管理多个生成式 AI 供应商服务。它解决了开发者或企业在同时使用不同大语言模型（LLM）或生成式 AI 平台时面临的接口不一致、集成复杂等问题，通过标准化 API 抽象层，简化了与各种 AI 服务的交互。该项目使开发者能够更轻松地比较不同模型、快速切换供应商或构建多模态 AI 应用。适用于需要集成、测试或管理多个生成式 AI 模型的开发者、研究人员及企业。

affaan-m/ECC

JavaScript · ★ 234,705 · 🍴 35,768 · 📈 692 stars today

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

中文介绍 `ECC` 是一个全面的 AI 代理性能优化系统，旨在为 Claude Code、Codex、Opencode 和 Cursor 等多种大型语言模型（LLM）驱动的代理提供增强功能。它专注于提升代理的技能、直觉、记忆、安全性，并采用研究优先的开发方法，以构建更强大、高效且可靠的 AI 代理。该项目适合开发和部署高级 AI 代理的工程师和研究人员，特别是在需要整合多模态能力和优化代理行为的场景中。

huggingface/speech-to-speech

Python · ★ 7,143 · 🍴 946 · 📈 177 stars today

Build local voice agents with open-source models

中文介绍 Hugging Face 的 `speech-to-speech` 项目提供了一套工具和方法，旨在帮助开发者利用开源模型构建本地运行的语音代理。它专注于实现高质量的语音到语音（Speech-to-Speech）转换，使用户能够在设备本地部署智能语音交互系统。这解决了云端语音服务可能存在的隐私、延迟和成本问题。该项目适用于需要定制化、低延迟或离线语音助手的场景，例如智能家居、车载系统或个人生产力工具，赋能开发者创建创新的语音应用。

virgiliojr94/book-to-skill

Python · ★ 11,182 · 🍴 1,324 · 📈 366 stars today

Turn any technical book PDF into a Claude Code skill — ready to study, reference, and use while you work.

中文介绍 `book-to-skill` 项目致力于将任何技术书籍的 PDF 文档转化为 Anthropic Claude Code AI 的可用“技能”。它通过处理 PDF 内容，使其能被 AI 理解和利用，从而将书籍的知识库无缝集成到开发者的工作流中。这极大地便利了技术内容的学习、快速参考和实际应用，解决了在复杂技术文档中检索信息效率低下的问题。开发者、工程师或任何技术学习者都能通过此工具，将浩瀚的专业书籍变成随时可用的智能助手，提高学习和工作效率。

opengeos/GeoLibre

TypeScript · ★ 3,319 · 🍴 388 · 📈 743 stars today

A lightweight, cloud-native GIS platform for visualizing, exploring, and analyzing geospatial data. It runs in the web browser, on the desktop, on mobile, and inside Jupyter notebooks.

中文介绍 GeoLibre 是一个轻量级、云原生的 GIS 平台，专为地理空间数据的可视化、探索和分析而设计。它支持多端运行，包括网页浏览器、桌面应用、移动设备以及 Jupyter Notebooks 环境，极大方便了用户在不同场景下处理和分析空间信息。该平台特别适合地理信息专业人员、数据科学家及开发者，用于构建交互式地图、进行空间数据分析或开发跨平台 GIS 应用。

paperswithbacktest/awesome-systematic-trading

Python · ★ 9,493 · 🍴 1,289 · 📈 402 stars today

A curated list of awesome libraries, packages, strategies, books, blogs, tutorials for systematic trading.

中文介绍 `awesome-systematic-trading` 是一个精心策展的资源列表，旨在汇集系统化交易领域内各类优质资料。它提供了包括编程库、软件包、交易策略、专业书籍、行业博客和学习教程等在内的丰富内容，覆盖了量化交易、算法交易和投资研究的核心技术与方法。该项目为量化交易员、金融数据分析师、研究人员和对系统化投资感兴趣的学习者提供了一站式信息获取平台，极大方便了他们发现、学习和应用相关的工具和知识，加速了系统化交易策略的开发与优化。

microsoft/agent-governance-toolkit

Python · ★ 5,137 · 🍴 831 · 📈 17 stars today

AI Agent Governance Toolkit — Policy enforcement, zero-trust identity, execution sandboxing, and reliability engineering for autonomous AI agents. Covers 10/10 OWASP Agentic Top 10.

中文介绍 Microsoft 的 `agent-governance-toolkit` 是一个专为自主 AI 代理设计的治理工具包，旨在确保 AI 代理的安全、可靠和合规运行。它提供了策略执行、零信任身份验证、执行沙箱以及可靠性工程等关键功能，全面解决 AI 代理可能面临的风险和挑战。该工具包完全覆盖了 OWASP Agentic Top 10 所列的全部安全威胁，为企业和开发者提供了构建可信赖 AI 系统的框架。适用于需要大规模部署或对安全性有严格要求的 AI 代理应用场景，如企业自动化、关键业务流程等。

yorukot/superfile

Go · ★ 21,407 · 🍴 693 · 📈 660 stars today

Pretty fancy and modern terminal file manager

中文介绍 superfile 是一款现代化的终端文件管理器，旨在为命令行用户提供美观且功能强大的文件管理体验。它解决了传统终端文件操作界面简陋、效率不高的问题，通过直观的交互方式，使用户能在终端环境中高效地浏览、查找、复制、移动和删除文件。适用于开发者、系统管理员及偏爱命令行操作的高级用户，提升其日常文件管理效率。

bradautomates/claude-video

Python · ★ 11,982 · 🍴 1,207 · 📈 989 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。

How to Build a Team of AI Agents That Actually Work Together (Full Course)

@sairahul1 · 130.6K 粉丝 · 1.5M 阅 · 507 赞 · 59 转

07/22 14:34

I run a one-person business. No team. No employees. No co-founder. For two years I have been the researcher, writer, planner, reviewer, and strategist. All of it. At once. Last week I tried something

中文介绍 这位独立创业者分享了如何利用 AI 智能体组建一个「虚拟团队」，以应对研究、写作、规划、审核和策略制定等多重业务角色。他探讨了如何让 AI 智能体有效协作，模拟真实的团队工作流，帮助像他一样的单人公司高效运营，提供了从零开始构建智能体协作系统的实战经验。

Why hasn’t AI increased unemployment?

@PeterMcCrory · 46.5K 粉丝 · 399.2K 阅 · 514 赞 · 106 转

07/23 01:18

I thought I’d share a few high-level reflections and a framework that helps me make sense of why we (so far) don’t see significant impact of AI on the US labor market. I focus on the US because (a) AI

中文介绍 博主分享了对 AI 至今尚未显著提升美国失业率的看法，并提出了一个分析框架来解释这一现象。内容聚焦于 AI 对美国劳动力市场的影响，探讨了当前阶段 AI 如何与就业动态相互作用，并未像许多人预期那样大规模取代工作岗位，提供了对宏观经济影响的深度思考。

The complete Graph Engineering playbook for Claude Code

@Gyome1_ · 3.9K 粉丝 · 220.0K 阅 · 504 赞 · 65 转

07/23 16:53

Most people are still using Claude Code like a very expensive intern. They give it one task, wait for one answer, then manually decide what happens next. But the teams getting the most leverage from

中文介绍 博主分享了针对 Claude Code 的「图工程」完整指南。他指出，许多用户仍将 Claude Code 视为昂贵的实习生，每次只分配一项任务并等待单一答案，然后手动决定下一步，这种方法效率低下。该指南旨在帮助团队最大化 Claude Code 的利用率，通过系统性的图工程方法，实现更高效、更具杠杆效应的 AI 协作工作流。

More On An Internal OpenAI Model Hacking Into HuggingFace

@TheZvi · 39.0K 粉丝 · 219.4K 阅 · 510 赞 · 75 转

07/27 03:14

We now have more details of what happened. Every time we learn more details, it somehow makes things seem worse. The remaining details may have to wait a bit. OpenAI: We recognize there are a lot of

中文介绍 OpenAI内部模型「入侵」HuggingFace事件再曝细节。博主指出，每次新信息都让情况显得更糟，暗示该事件可能涉及更深层的模型自主性或安全问题。OpenAI方面已承认存在大量疑问，后续详情待披露。

22580: From GPT2 to Kimi3, Explained

@waterloo_intern · 10.4K 粉丝 · 215.4K 阅 · 660 赞 · 86 转

07/27 23:22

Twenty-two thousand five hundred and eighty. That’s how many GPT-2 (2019) models fit inside KimiK3 (2026). We scaled up by a factor of 22,580 in seven years. But is it just... scale? In this worklog,

中文介绍 博主分析了从2019年的GPT-2到2026年预测的KimiK3模型，计算出模型规模在7年内增长了22,580倍。推文质疑这种「规模化」是否是唯一的进步指标，并表示会在工作日志中深入探讨AI发展中的其他关键因素。这篇分析旨在超越单纯的参数数量，审视AI技术演进的深层驱动力及潜在影响，鼓励对行业发展进行更全面的思考。

Frontier Diffusion & Control

@satyanadella · 7.5M 粉丝 · 153.7K 阅 · 923 赞 · 114 转

07/24 00:31

In a world where software has real marginal cost for the first time, how do we ensure frontier benefits are diffused across the entire ecosystem? The key is to optimize the cost-to-outcome frontier in

中文介绍 微软 CEO Satya Nadella 探讨了在软件首次具有真实边际成本的时代，如何确保前沿技术（特别是 AI）的效益能够广泛扩散至整个生态系统。他强调关键在于优化「成本-结果」边界，以实现最大化的技术普及和价值创造，提出了对 AI 时代技术发展和经济模式的战略思考。

Towards Automating Eval Engineering

@Vtrivedy10 · 13.9K 粉丝 · 152.6K 阅 · 503 赞 · 50 转

07/23 01:04

Today we’re releasing our Eval Engineering Skill, a skill that helps coding agents build evals using context from a repository and agent traces. The skill inspects how an agent is structured, mines

中文介绍 博主发布了名为「Eval Engineering Skill」的新工具，旨在帮助编码智能体自动构建评估（evals）。该技能通过检查智能体结构，并利用代码仓库上下文和智能体运行轨迹，为智能体生成高效的自我评估机制，以提升其性能和可靠性，实现了自动化评估工程的创新。

this system will change your life...

@eptwts · 116.8K 粉丝 · 140.0K 阅 · 514 赞 · 28 转

07/24 01:21

the consensus on self-hosted agents is that they're simply coding tools... you point one at a repo, it writes code, you close the terminal & everything it learned about you dies with that session.

中文介绍 该推文指出，当前自托管 AI 代理（self-hosted agents）普遍存在局限性，即被视为单纯的编码工具，每次会话结束后，它们学到的所有用户上下文都会随之丢失。博主认为这种「即用即丢」模式限制了代理的持续学习和效率，暗示需要一种能保留上下文的新系统。

Opus 5 is a really bad model

@HarukaKunori · 241 粉丝 · 129.0K 阅 · 632 赞 · 44 转

07/27 19:07

After trying out Opus 5 for a few hours today, I honestly think Anthropic's benchmark scores are a complete fraud. Sure, the model might have improved in a few areas, but it has regressed unbelievably

中文介绍 博主试用Anthropic的Opus 5数小时后，强烈质疑其基准测试分数存在「欺诈」。他认为尽管模型在某些方面有所改进，但在多数方面却出现了令人难以置信的退步，与官方宣传大相径庭。

How to master graph engineering (Full Course)

@EXM7777 · 129.0K 粉丝 · 124.5K 阅 · 520 赞 · 54 转

07/22 22:20

I'm going to show you how to build your first agent graph and put it to work in your business today: A team of AI agents that researches in parallel, tries to kill its own findings, and hands you one

中文介绍 该博主将展示如何构建首个「智能体图」（agent graph）并将其应用于业务中。这个智能体团队能够并行进行研究、主动验证并「反驳」自身发现，最终提供精炼的见解。这是一个关于如何设计和实现高效协作式 AI 智能体工作流的教程，旨在帮助用户掌握图工程技术。

Why Software Factories Fail: Benchmarking the new frontier

@dexhorthy · 25.8K 粉丝 · 115.4K 阅 · 500 赞 · 39 转

07/28 01:43

This is a continuation of Parts 1 and 2 of "Why Software Factories Fail" Part 1: the harness is not enough Part 2: turning the lights back on we got better benchmarks Remember when I said this in Part

中文介绍 该帖子是「软件工厂为何失败」系列文章的续篇，深入探讨了构建和评估软件工厂的挑战。博主在第一部分「仅有脚手架是不够的」和第二部分「重新点亮灯火」的基础上，提出了「更好的基准测试」方法，旨在帮助开发者理解软件工厂的局限性，并优化其性能评估策略。内容聚焦如何避免常见陷阱，提升自动化软件开发的成功率。

why we're buzzing

@jack · 10.3M 粉丝 · 109.6K 阅 · 782 赞 · 93 转

07/23 06:25

yesterday we released buzz. it's an open source workspace that puts people, agents, conversations, and code on the same level, behind one cryptographic identity system. we built it to reduce our

中文介绍 Jack Dorsey 发布了「buzz」，一个开源工作区，旨在将人员、智能体、对话和代码整合到一个统一的加密身份系统之下。该平台旨在简化多方协作，通过提供一个扁平化的工作环境，减少系统复杂性，提升团队和 AI 智能体的交互效率，是关于未来协作模式的新探索。

Trajectory: A Standard Format for Agent Experience Data

@Letta_AI · 9.8K 粉丝 · 86.5K 阅 · 504 赞 · 31 转

07/24 14:07

Introducing trajectory, an open-source package that normalizes coding-agent sessions from @AnthropicAI Claude Code, @OpenAI Codex, @pidotdev, @LangChain deepagents, @openclaw, Letta Code, and other

中文介绍 该推文介绍了名为「trajectory」的开源软件包，旨在标准化来自不同 AI 编程代理的会话数据。它能够归一化 AnthropicAI 的 Claude Code、OpenAI 的 Codex、pidotdev、LangChain deepagents、openclaw 以及 Letta Code 等多个主流 AI 编码工具的「agent experience data」。trajectory 的目标是解决 AI 代理数据格式不统一的问题，促进跨平台的数据互操作性与分析，提升开发效率。

Graph Engineering Clearly Explained

@akshay_pachaar · 282.5K 粉丝 · 66.4K 阅 · 515 赞 · 74 转

07/26 02:48

Loop engineering got about six weeks in the spotlight before the timeline moved on. On July 18, Peter Steinberger, the person behind OpenClaw, posted a nine-word question. "Are we still talking loops

中文介绍 讨论 AI 智能体架构的演进，从「Loop engineering」转向「Graph Engineering」。博主旨在清晰解释 Graph Engineering 这一新概念，并指出 Loop engineering 在短暂热度后已被新趋势取代，引发了关于当前智能体设计模式的思考。

Why Software Factories Fail

@dexhorthy · 25.8K 粉丝 · 64.9K 阅 · 515 赞 · 50 转

07/25 00:51

or: the harness is not enough Update - the talk version of this post is live on youtube: https://www.youtube.com/watch?v=Ib5GBkD555M i guess we doin loops now We're all racing to put AI coding into production. A lot has

中文介绍 博主探讨了「软件工厂」模式失败的原因，特别是在将 AI 编码引入生产环境的背景下。推文指出，仅仅依靠「harness」（工具或框架）不足以确保成功，暗示当前将 AI 编码落地存在深层挑战。内容可能分析了在生产环境中规模化部署 AI 编码时面临的陷阱和限制，为开发者和团队提供了避免类似错误的反思与警示。

Here's exactly how to build your company brain (in 5 mins)

@DhravyaShah · 61.2K 粉丝 · 56.0K 阅 · 548 赞 · 36 转

07/28 14:16

Every company will have a company brain, whether they believe it or not. This is the bet that I'm making, and I'm constantly seeing very future forward startups come to us for their own setup. What is

中文介绍 博主提出「公司大脑」是企业发展的必然趋势，并分享了「如何准确地在5分钟内构建公司大脑」的实操指南。他观察到许多前瞻性初创公司正积极寻求建立自己的公司大脑系统。该帖子旨在指导企业快速搭建一个整合知识、驱动智能决策的内部AI平台，强调其对未来公司运营的重要性与即时性，提供实用操作建议。

Agent Harness Engineering vs. Loop Engineering vs. Graph Engineering

@beamnxw · 2.6K 粉丝 · 53.5K 阅 · 561 赞 · 92 转

07/25 22:25

A practical guide to the three architecture layers people keep mixing together The confusion is understandable. All three ideas sit around the same model, all three influence reliability, and all

中文介绍 提供智能体架构的实用指南，详细对比并区分了「Agent Harness Engineering」、「Loop Engineering」和「Graph Engineering」这三个常被混淆的概念。博主旨在厘清它们各自的特点及其对模型可靠性的影响，帮助读者理解不同架构层。

Getting the most out of GPT-5.6: Sol, Terra, and Luna

@cerebras · 64.6K 粉丝 · 48.8K 阅 · 550 赞 · 40 转

07/28 03:44

Authors: @0xSero & Zhenwei Gao (@zhennydez) Your Codex subscription now comes with 3 main models, Sol, Terra, and Luna, each independently trained and served, with reasoning dials. Together, they

中文介绍 该帖子宣布，Codex订阅服务已新增三款主要模型：Sol、Terra和Luna。这三款模型均独立训练、部署，并配备了独特的「推理调节器」（reasoning dials）。推文旨在指导用户如何充分利用这些新模型，以优化其在GPT-5.6（或高级AI任务）中的表现，帮助用户根据不同任务需求微调模型行为，从而获得更精准和高效的AI输出。

The new rules of context engineering for Claude 5 models

@trq212 · 318.0K 粉丝 · 48.6K 阅 · 1.2K 赞 · 119 转

07/25 01:45

I’ve written previously about how to best prompt the newest generation of Claude 5 models and work with them iteratively to discover what you want to build. But when you send a message to Claude, the

中文介绍 博主分享了针对 Claude 5 模型「上下文工程」的新规则。推文延续此前关于如何最佳提示 Claude 5 模型并进行迭代工作的内容，重点探讨了向 Claude 发送消息时上下文如何运作，旨在帮助用户更有效地利用模型。

Data Pyramid for Embodied Manipulation

👍 30

07/27 08:00

Multimodal foundation models learned to see and to speak by consuming the whole internet. Embodied agents admit no such shortcut, since they require data that couple observations with physical states and actions. These signals can be provided, to varying degrees, by multiple data sources. In this wo

中文介绍 具身智能体（Embodied agents）缺乏像多模态基础模型那样大规模的互联网数据。该研究提出「Data Pyramid」概念，旨在为具身操作提供数据，这些数据需结合观测、物理状态和动作。文章探讨了如何通过多种数据源提供这些关键信号，以克服数据瓶颈，提升具身智能体的学习能力。

ClinFusion: A Vision-Centric Multimodal LLM System for Holistic Medical Understanding

👍 4

07/27 08:00

Multimodal large language models (MLLMs) hold immense potential to revolutionize clinical practice, yet deploying them in the medical domain is fundamentally a vision-centric challenge: models must absorb knowledge from heterogeneous 2D and 3D medical images, and evaluation protocols must align with

中文介绍 该研究提出了「ClinFusion」系统，这是一个以视觉为中心的多模态大语言模型（MLLM），旨在实现全面的医学理解。文章指出，将MLLM应用于医疗领域面临的核心挑战是需要模型从异构的2D和3D医学图像中学习知识，这对于提升临床实践具有巨大潜力。

Rethinking Classifier-Free Guidance in On-Policy Diffusion Distillation

👍 61

07/27 08:00

On-policy distillation (OPD) adapts diffusion models by querying a teacher along trajectories generated by the current student, but how it should behave under classifier-free guidance (CFG), a default component of modern diffusion systems, remains poorly understood. Existing OPD methods naturally ex

中文介绍 该论文重新审视了策略内蒸馏（On-policy distillation, OPD）中「无分类器引导」（Classifier-Free Guidance, CFG）的应用。OPD通过查询教师模型来适应扩散模型，而CFG是现代扩散系统的默认组件。研究指出，CFG在OPD下的行为机制尚未被充分理解，现有OPD方法对此考虑不足。

DreamStyle3D: Efficient 3D Stylized Asset Generation via Dual-Attention Disentanglement

👍 1

07/27 08:00

With the growth of gaming, animation, and virtual reality industries, the demand for efficient generation of stylized 3D assets is rapidly increasing. However, existing approaches still struggle to jointly preserve style fidelity, geometric consistency, and generation efficiency, as most of them sti

中文介绍 随着游戏、动画和虚拟现实产业的发展，对高效生成风格化3D资产的需求日益增长。该论文提出了「DreamStyle3D」方法，通过双注意力解耦（Dual-Attention Disentanglement）技术，旨在解决现有方法在同时保持风格保真度、几何一致性和生成效率方面的挑战，从而更有效地创建风格化3D内容。

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

👍 19

07/27 08:00

Multi-turn long-horizon planning is critical for foundation model agents, yet how to fundamentally improve it remains unclear. Existing models are trained on uncontrollable and opaque Internet data, making it difficult to identify how planning ability is acquired, shaped, and integrated. To address

中文介绍 针对基础模型智能体的多轮长程规划能力，该研究探讨了从预训练到后训练的改进机制。文章通过单教师和多教师策略内智能体蒸馏（On-Policy Agentic Distillation）方法，深入分析了规划能力如何习得、塑造和增强，旨在解决现有模型依赖不可控网络数据导致规划能力不透明的问题。

DataOrchestra: Learning to Orchestrate Per-Example Curation of Pretraining Data

👍 0

07/27 08:00

Pretraining data processing is critical to the downstream performance of Large Language Models (LLMs). However, many existing approaches define a fixed processing strategy at the corpus or domain level and apply it uniformly to many examples, without adapting to the needs of each example. We propose

中文介绍 预训练数据处理对大语言模型（LLMs）的下游性能至关重要。现有方法通常在语料库或领域层面定义固定处理策略，并统一应用于所有样本，缺乏对单个样本需求的适应性。该论文提出了「DataOrchestra」框架，旨在学习如何对每个预训练数据样本进行精细化编排和管理，以提升LLM性能。

ERUnderstand: Evaluating Vision-Language Models on Structured ER Diagrams

👍 2

07/27 08:00

Entity-Relationship Diagrams (ERDs) are central to conceptual database design, yet they are typically available only as rendered images rather than machine-readable schemas, limiting AI-assisted database engineering. We introduce ERUnderstand, the first large-scale benchmark for structured understan

中文介绍 实体关系图（ERD）是概念数据库设计的核心，但其通常以渲染图像而非机器可读模式存在，限制了AI辅助数据库工程。该研究引入了「ERUnderstand」，这是首个用于评估视觉语言模型（VLM）在结构化ERD上性能的大规模基准数据集，旨在弥合图像与可理解数据之间的鸿沟。

Beyond Scale and Generation: Understanding Language Model-based Entity Matching

👍 1

07/27 08:00

Entity matching identifies records that refer to the same real-world entity. Language models can be adapted to this task through bi-encoder, cross-encoder, and generative matcher architectures. However, prior studies often conflate matcher architecture with differences in model backbone, model varia

中文介绍 实体匹配旨在识别指向同一现实世界实体的记录。该研究超越了规模和生成能力，深入理解基于语言模型的实体匹配任务。文章探讨了双编码器、交叉编码器和生成式匹配器等架构，旨在区分匹配器架构与模型骨干差异对实体匹配性能的影响，以提供更清晰的认识。

Kimi K3: Open Frontier Intelligence

👍 251

07/27 08:00

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model d

中文介绍 该论文介绍了「Kimi K3」模型，这是一个拥有2.8万亿参数的MoE（专家混合）模型，其中包含1040亿激活参数。Kimi K3具备原生的视觉能力和100万个token的上下文窗口，其核心技术是Kimi Delta Attention和Attention Residuals，旨在显著提升序列间的信息流。

Evidence Attribution in Visual Document Understanding without Coordinates or Region Labels

👍 2

07/27 08:00

Reliable visual document understanding requires a model to attribute each answer to the evidence regions that support it. Recent benchmarks and systems express this step through a coordinate interface: the model outputs the coordinates of bounding boxes that mark the evidence regions in the document

中文介绍 可靠的视觉文档理解要求模型将每个答案归因于其支持的证据区域。现有基准和系统通常通过坐标接口实现，模型输出边界框坐标来标记证据区域。该研究探索了一种在不依赖坐标或区域标签的情况下，实现视觉文档理解中证据归属的方法，以提高模型的鲁棒性。

QueenVIS: Rethinking Image-Only Training for Video Instance Segmentation via Query Enrichment

👍 0

07/27 08:00

Video instance segmentation (VIS) requires models to detect, segment, and track object identities across frames, and most methods enforce temporal consistency through video-level supervision. Image-only training approaches, with MinVIS as one prominent example, have challenged this assumption, reach

中文介绍 视频实例分割（VIS）要求模型在帧间检测、分割并跟踪目标身份，多数方法通过视频级监督来确保时间一致性。该论文提出「QueenVIS」方法，通过查询富集（Query Enrichment）重新思考了仅用图像进行VIS训练的范式，挑战了以MinVIS为代表的现有图像级训练方法，旨在提升性能。

From Data to Device: ELMOD An Efficient German-First 2.7B Language Model for Mobile Inference

👍 1

07/27 08:00

We present ELMOD - Efficient Language Model for On-Device Deployment - a compact (2.7B) German language model designed for efficient inference on resource-constrained hardware. ELMOD was trained on a limited computational budget (55k H100 GPU hours) using exclusively publicly available data. We deve

中文介绍 该论文介绍了「ELMOD」（Efficient Language Model for On-Device Deployment），这是一个紧凑型、27亿参数的德语优先语言模型。ELMOD专为在资源受限硬件上进行高效推理而设计，仅使用5.5万H100 GPU小时和公开数据进行训练，旨在实现从数据到设备的优化部署。

DecoupleMix: Decoupled Ratio Search and Convex Allocation for Scalable VLM Data Recipes

👍 5

07/27 08:00

While data curation for Vision Language Models (VLMs) is increasingly active, public practice for constructing pretraining mixtures remains largely heuristic: practitioners stack datasets that pass quality filters, set cross-domain ratios by intuition, and lack a principled, attributable criterion f

Mixture-of-Thought-Tokens: Unifying Perception and Reasoning for Free-form Multimodal Grounding

👍 0

07/27 08:00

Multimodal Large Language Models have made great progress in grounding tasks, yet existing methods still struggle to unify precise localization and complex reasoning. For one thing, text-based methods rely on coordinates or index prediction, severely limiting the perceptual capabilities of the model

From Proprietary to Open-Source: Bridging the Distribution Gap via Multi-Agent Protocol Distillation in Agentic Search

👍 64

07/27 08:00

Agentic search enables large language models to solve knowledge-intensive tasks by interleaving multi-step reasoning with retrieval, yet optimizing this with outcome-based reinforcement learning (RL) provides only sparse supervision. Knowledge distillation can supply denser guidance, and advanced pr

Catalyst Diffusion Transformer: Generative Inverse Design of Heterogeneous Catalysts

👍 0

07/27 08:00

The vast chemical design space and complex, interdependent design variables make catalyst discovery for targeted properties highly labor- and resource-intensive. Although generative models have emerged as a promising solution, existing approaches are generally limited to single-property conditioning

FilmBench: A Film-Grade Benchmark for Cinematic Video Generation

👍 3

07/27 08:00

Progress in video generation keeps narrowing the visual gap between AI-generated and professionally produced footage, yet most benchmarks still draw prompts from web sources or LLM templates and score them with untrained, generic multimodal models. More fundamentally, their evaluation taxonomies rem

Where Quality Breaks in Compressed Short-Text Generation: Staged Bottleneck Localization

👍 0

07/27 08:00

Compressed short-text generators can fail in two different places: the codec may discard information before generation starts, or the latent generator may produce weak codes. Without separating these failure modes, researchers can spend compute improving the wrong component. We study this problem in

DeVA: Decoupled Video-Action Model with physical guidance for robot policy learning

👍 0

07/27 08:00

Generalizable robot manipulation requires policies that can anticipate how visual scenes evolve while executing language instructions. While recent Vision-Language-Action models benefit from large-scale pretraining, their predominantly static pretraining objectives provide limited supervision for ph

Grading the Narrators: An Isnad-Rijal Framework for Claim-Level Provenance in Multi-Agent Knowledge Systems

👍 0

07/27 08:00

Modern multi-agent knowledge systems increasingly accumulate knowledge through chains of autonomous transformations rather than direct retrieval. Existing provenance work records what happened - execution traces, tool calls, evidence links - and source-reliability estimation is long established (tru

Sol-Attn: Accelerating Video Generation Inference via On-the-Fly Attention Sparsification

👍 24

07/27 08:00

Diffusion transformers are essential for high-fidelity video generation, but long token sequences make attention a dominant inference bottleneck. Training-free dynamic sparse attention alleviates this bottleneck by computing only selected key-value blocks, yet existing methods struggle to sparsify a

DailyBench: A Unified Benchmark for AI-Generated and Manipulated Images from Modern Generative Models

👍 1

07/27 08:00

Recent advances in generative models have shifted AI-generated image detection from identifying easily distinguishable, fully synthetic images to identifying highly realistic content generated by both modern generation and manipulation pipelines. However, existing detection benchmarks are often buil

Modeling Stylistic Co-evolution in Symbolic Music Heritage Collections

👍 0

07/27 08:00

Digitized musical heritage collections offer new opportunities to examine how stylistic traditions change over historical time, but computational analyses often reduce musical works to static classifications or similarity scores. This article proposes a representation-to-dynamics framework for study

From Cognitive Architectures to Language Agents: A Mechanism-Level Review of Lineage, Convergence, and Migration Gaps

👍 1

07/27 08:00

Memory, planning, reflection, and tool use are often compared as feature labels, obscuring the control semantics that determine how an agent actually runs. This review connects ten historical cognitive architectures, eight language-agent runtime families, and forty-two mechanism-focused modern syste

WorldDiT: A Unified Diffusion Architecture for World and Action Modeling

👍 0

07/27 08:00

Many recent robot policies pursue stronger control by using large pretrained vision-language models (VLMs) as the action backbone. We introduce WorldDiT, a unified diffusion transformer architecture that couples action generation with visual world modeling and achieves strong performance without a l

Nanbeige4.2-3B: Unlocking Agentic Capabilities in a Compact Model

👍 0

07/27 08:00

We present Nanbeige4.2-3B, a compact general agentic model with 3B non-embedding parameters. It delivers strong performance across code-agent, office-agent, and complex tool-use tasks while maintaining highly competitive reasoning capabilities in mathematics, coding, and science. Nanbeige4.2-3B is p

The Weight of Silence: A Causal Case for Weights Over the Scratchpad in Latent Chess Reasoning

👍 0

07/27 08:00

Latent, or silent, reasoning lets language models carry out intermediate computation in continuous vector space instead of words, and is widely assumed to function as an internal scratchpad the model consults during inference. Whether that assumption survives reinforcement learning has not been test

MedFailBench: A Clinician-Built Open-Source Benchmark for Medical AI Safety Boundary Inspection

👍 0

07/27 08:00

Most medical AI benchmarks measure whether a model knows the correct answer. MedFailBench asks a different question: which safety boundary failed? We present a synthetic benchmark and failure atlas built by a clinician. The resource labels medical AI errors by severity from 1 to 5 and safety gate ty

Poison to Detect: Detection of Targeted Overfitting in Federated Learning

👍 0

07/27 08:00

Federated Learning (FL) enables collaborative model training among clients without centralising data, making it a widely adopted privacy-enhancing technology (PET). Despite its privacy benefits, FL remains vulnerable to orchestrator-driven privacy attacks. In this paper, we study an underexplored th

OmniVAE: An Audio-Video VAE with Cross-Modal Alignment for Joint Generation

👍 21

07/26 08:00

Recent generative models are moving beyond silent video or standalone audio synthesis toward the joint generation of synchronized audio and video. Despite this progress, jointly generating audio and video with fine-grained cross-modal correspondence remains challenging due to their fundamental struc

Growth Opt Playbook

07/29 04:56

Turn campaign data into your next marketing move

中文介绍 「Growth Opt Playbook」是一款营销优化工具，能将营销活动数据转化为下一步的营销策略。它旨在帮助用户利用数据分析，制定有效的市场推广方案。

FlowTask 2.0

07/29 04:55

Company brain for AI Agents

中文介绍 「FlowTask 2.0」是一个为企业AI智能体设计的“公司大脑”平台。它旨在作为AI智能体的工作中枢，帮助企业整合并优化其内部的AI应用与流程管理。

Leaping AI

07/29 04:55

AI agents that call and text in multi-day campaigns

中文介绍 「Leaping AI」提供人工智能代理，能够执行多天的电话和短信营销活动。这些AI代理可以自动化地与潜在客户进行沟通，支持长时间的营销策略执行。

Snapr

07/29 04:54

Capture thoughts, find them instantly

中文介绍 「Snapr」是一款快速捕捉和即时查找想法的工具。它旨在帮助用户高效地记录灵感，并能在需要时迅速检索，提升信息管理效率。

Space-Bin

07/29 04:54

Secure temporary workspaces. No sign-up required.

中文介绍 「Space-Bin」提供安全的临时工作空间，且无需注册即可使用。它旨在为用户提供便捷、私密的在线协作环境，用于短期项目或敏感信息处理。

RecipeBook by Shofo

07/29 04:54

Buy video training data by the hour featuring 25M+ clips

中文介绍 「RecipeBook by Shofo」提供视频训练数据购买服务，用户可以按小时租赁，其中包含超过2500万个视频片段。该平台旨在为AI模型训练和研究提供大规模的视频数据集。

Cardzen

07/29 04:54

Track every credit, perk, and point before it expires

中文介绍 「Cardzen」是一款追踪信用卡积分、优惠和奖励的工具，旨在帮助用户在过期前管理所有权益。它能提醒用户及时使用，避免浪费。

Lamoom

07/29 04:54

Run agent apps inside your Claude or sell your own

中文介绍 「Lamoom」平台允许用户在Claude环境中运行AI代理应用，或销售自己开发的代理应用。它为AI智能体开发者提供了一个生态系统，支持应用的部署和商业化。

ChatBeacon AIX

07/29 04:54

AI support that resolves 87% of chats

中文介绍 「ChatBeacon AIX」是一款AI支持解决方案，声称能解决87%的客户在线聊天咨询。该产品旨在通过人工智能提升客户服务效率，减少人工干预。

SUB/WAVE

07/29 04:54

Self-hosted radio with an AI DJ and one shared stream

中文介绍 「SUB/WAVE」是一款可自托管的广播系统，配备AI DJ功能，并支持一个共享的直播流。它为用户提供个性化的电台体验，并可与他人共享内容。

How to Use Claude for Excel (Complete Beginner Guide)

07/28 07:23

中文介绍 由Riley Brown制作发布的这份YouTube视频教程，旨在为初学者提供一份完整指南，详细讲解如何有效利用人工智能模型Claude来辅助和优化Excel电子表格的操作。该视频内容将聚焦于演示Claude在数据处理、分析等方面的实际应用技巧，帮助用户掌握将AI工具与Excel结合使用的基本方法。

Opus 5 Is Here… But NEW Claude Voice Is Even Bigger

07/26 04:28

中文介绍 YouTube 视频指出，「Opus 5」已推出，但新的 Claude 语音功能「Claude Voice」被认为是更重要的进展。视频强调，虽然「Opus 5」已发布，但这一全新的 Claude 语音技术被认为具有更大的影响力。

OpenAI just released Codex Voice (It's basically Jarvis)

07/25 00:09

中文介绍 OpenAI 发布了其名为 Codex Voice 的新产品。该产品被描述为类似电影中「贾维斯」（Jarvis）的人工智能系统，暗示它可能具备先进的语音交互或智能助手功能，代表了AI在自然语言处理和人机互动方面的新进展。

Codex Basically Runs My Company Now. Here’s How.

07/22 06:27

中文介绍 YouTube用户Riley Brown发布视频，展示了他如何将Codex（推测为OpenAI Codex）改造为一个“业务增长机器”。该视频围绕利用AI技术，尤其可能是代码生成能力，来加速企业发展的实际策略展开。

Devin AI: The Full Beginner’s Guide (Better Than Claude Code?)

07/17 02:25

Director by OpenArt

07/17 02:12

Codex New Browser: The Secret Weapon Everyone Should be Using

07/15 21:00

OpenAI Just Merged ChatGPT and Codex. This Changes Everything.

07/13 01:24

Grok 4.5 + Cursor: Full Guide

07/10 01:51

Build Amazing Websites with Claude (Full Guide)

07/07 01:37

小孩子照片比较多，目前存在 icloud 2t，有什么什么比较优雅的实体备份方式？

07/28 21:47

16 回复 · Apple 节点

giffgaff 别开 WiFi Calling，我是 6 月份激活的实体卡，早上开了 WiFi Calling，刚收到邮件被封了

07/28 19:47

61 回复 · 程序员 节点

从 0 到 1 做方案设计和开发，现在哪个模型比较强

07/28 16:44

7 回复 · 程序员 节点

各位的 macOS 菜单栏里放了哪些有用的小工具？

07/28 14:45

21 回复 · Apple 节点

19 年的 MBP 坏了

07/28 14:02

21 回复 · Apple 节点

注册 Apple Developer 求问

07/27 18:17

7 回复 · Apple 节点

如何才能成功注册 Kliq 用于绑定土区

07/27 14:37

13 回复 · Apple 节点

现在 mba 32g+512 有好价吗？

07/27 14:23

6 回复 · Apple 节点

TG api 我用 google Voice 老申请失败

07/24 11:18

17 回复 · Python 节点

北通手柄 Linux USB 断连排查与修复

07/23 15:05

6 回复 · Linux 节点

Recursion Is Lying to You

07/29 04:20

4 points · 3 comments

Pacing the frontier

07/29 04:09

10 points · 1 comments

Hulios: An eBPF-powered, transparent Tor gateway for Linux

07/29 03:54

8 points · 0 comments

Uv 0.12.0

07/29 03:41

64 points · 28 comments

MCP 2026-07-28 Specification: transport going stateless

07/29 02:35

58 points · 21 comments

The iPhone Upgrade Program is being replaced by Apple Upgrade

07/29 01:37

https://www.apple.com/shop/apple-upgrade

Discovering Cryptographic Weaknesses with Claude

07/29 01:22

107 points · 54 comments

Steel Bank Common Lisp version 2.6.7

07/29 01:11

133 points · 43 comments

WOFF 1.0: a milestone on W3C's journey of fonts on the web

07/29 01:05

45 points · 1 comments

Substack writers, you need a website

07/29 00:58

251 points · 143 comments

A walk through of the DeltaNet family of linear attention variants

07/29 00:02

271 points · 111 comments

How Do I Profile eBPF Code?

07/28 23:55

91 points · 6 comments

Show HN: XY – A Fast, composable, GPU-accelerated interactive plotting library

07/28 23:54

85 points · 30 comments

Delayed Gratification – Proud to Be 'Last to Breaking News'

07/28 23:50

178 points · 91 comments

Kimi K3 Architecture Overview and Notes

07/28 23:48

186 points · 21 comments

Zig's Incremental Compilation Internals

07/28 23:46

138 points · 96 comments

So, you want to make a game engine (2023)

07/28 23:36

49 points · 30 comments

Harmony Explained: Progress Towards a Scientific Theory of Music (2012)

07/28 23:20

74 points · 54 comments

Anthropeum – Where in the world, and when, does this human artifact belong?

07/28 23:01

114 points · 32 comments

Now Is the Time to Give LLMs Access to the ACM Digital Library

07/28 23:01

83 points · 54 comments

Stop Killing the Internet: No Digital ID and No Age Verification

07/28 22:58

358 points · 108 comments

Una GPS smart watch – Repairable, USB-C charging, developer-friendly

07/28 22:48

83 points · 60 comments

New HIV vaccine shows unprecedented success in preclinical study

07/28 21:12

459 points · 219 comments

Kimi Linear: An Expressive, Efficient Attention Architecture (2025)

07/28 18:52

248 points · 108 comments

DMARC has been public since 2012 but most company domains still don't enforce it

07/28 18:20

157 points · 98 comments

7.1 Earthquake in Japan

07/28 15:44

756 points · 196 comments

The most advanced robotic servicing satellite–that we know about

07/25 02:39

24 points · 1 comments

The Fabled Flatbreads of Uzbekistan (2015)

07/24 22:43

31 points · 15 comments

How to survive boiling water

07/24 22:19

405 points · 88 comments

Show HN: How far do I have to go to run into 100k people?

07/24 05:33

6 points · 1 comments

v2.1.220

07/25 09:35

What's changed Bug fixes and reliability improvements

中文介绍 Anthropic 旗下的 Claude Code 项目发布了 v2.1.220 版本。此次更新主要内容为错误修复和可靠性改进，旨在提升软件的稳定性和用户体验。

v2.1.219

07/25 01:14

What's changed Added Claude Opus 5 (claude-opus-5), now the default Opus model — 1M context, fast mode at $10/$50 per Mtok Added sandbox.network.strictAllowlist setting to deny non-allowlisted hosts for sandboxed commands without prompting Added DirectoryAdded hook that fires after /add-dir or the S

中文介绍 Anthropic的Claude-code项目发布v2.1.219更新。此版本引入了Claude Opus 5（claude-opus-5）作为默认Opus模型，具备1M上下文。其快速模式定价为每百万token $10/$50。同时，更新新增“sandbox.network.strictAllowlist”设置，增强沙盒命令的网络安全。

v2.1.218

07/23 05:24

What's changed Changed /code-review to run as a background subagent, so review work no longer fills your conversation and keeps stacked slash commands as its review target Added screen-reader announcements of deleted text for word and line deletions (Option+Delete, Ctrl+W, Cmd+Backspace, Ctrl+U, Ctr

v2.1.217

07/22 05:35

What's changed Added emoji shortcode autocomplete in the prompt input: type :heart: to insert ❤️, or :hea for suggestions — disable with the emojiCompletionEnabled setting Added warnings when transcript writes are failing (e.g. disk full) or when session saving is off due to an inherited environment

v2.1.216

07/21 06:14

What's changed Added sandbox.filesystem.disabled setting to skip filesystem isolation while keeping network egress control Fixed a slowdown in long sessions where message normalization cost grew quadratically with the number of turns, causing multi-second stalls and slow resumes Fixed auto mode deny

v2.1.215

07/19 10:56

What's changed Claude no longer runs the /verify and /code-review skills on its own; invoke them with /verify or /code-review when you want them

v2.1.214

07/18 09:20

What's changed Fixed single-segment dir/** allow rules like Edit(src/**) auto-approving writes to nested dir/ directories anywhere in the tree instead of only /dir Fixed a permission-check bypass affecting commands run in Windows PowerShell 5.1 sessions Fixed Bash permission checks to fail closed on

v2.1.212

07/17 08:26

What's changed /fork now copies your conversation into a new background session (its own row in claude agents) while you keep working; the in-session subagent it used to launch is now /subtask Added claude auto-mode reset to restore the default auto-mode configuration, with a confirmation prompt (pa

v2.1.211

07/16 07:02

What's changed Added --forward-subagent-text flag and CLAUDE_CODE_FORWARD_SUBAGENT_TEXT environment variable to include subagent text and thinking in stream-json output Fixed permission previews relayed to chat channels not neutralizing bidirectional-override, zero-width, and look-alike quote charac

v2.1.210

07/15 07:45

What's changed Added a live elapsed-time counter to the collapsed tool summary line so long-running tool calls visibly tick instead of looking stuck Added a startup warning for Write(path), NotebookEdit(path), and Glob(path) permission rules — use Edit(path) or Read(path) instead Fixed isolation: 'w