OpenAI News · 06/15 01:00
OpenAI正式推出「合作伙伴网络」，并计划投资1.5亿美元，支持全球合作伙伴加速企业级AI解决方案的采纳、部署和转型，旨在推动AI在商业领域的应用与发展。
推荐理由：这是OpenAI在企业级AI生态布局上的关键一步，对有意与OpenAI合作的企业具有直接指导意义。
2026-06-15（内容获取于 06/15 21:26）
TLDR AI · 06/15 08:00
Anthropic宣布关闭旗下AI项目Fable，此前有报道称该项目因「过于危险」而被停止发布。此外，AI领域近期还有GLM-5.2的发布以及OpenRouter Fusion的进展。（多家报道）
推荐理由：了解AI前沿项目的风险与挑战，同时关注新模型的发布，有助于把握AI技术发展方向。
Two Minute Papers · 06/14 23:27
NVIDIA（英伟达）发布了一款新的免费人工智能技术，具体细节和功能在视频中有所阐述，被标题强调为「对人类的礼物」。
推荐理由：英伟达的免费AI工具可能成为开发者的新选择，值得关注其具体功能和应用潜力。
Product Hunt · 06/15 14:04
Kickbacks.ai是一款创新工具，旨在让用户在等待Claude Code生成代码时获得收益。它提供了一种利用空闲计算资源或注意力的方式，为开发者带来了潜在的额外价值。
推荐理由：为开发者提供了利用AI生成代码等待时间的实用工具，提升生产力。
Product Hunt · 06/15 13:21
AgentBrush是一款为AI编程代理（coding agent）设计的图像生成工具，弥补了现有代理在视觉创作方面的缺失。它使编程代理能够进行图像创作，扩展了其应用范围。
推荐理由：丰富了AI编程代理的功能，为自动化内容生成提供了新的可能性。
Claude Code Changelog
Claude Code v2.1.113版本更新，改进了CLI以生成原生二进制文件，并新增`sandbox.network.deniedDomains`设置，允许阻止特定域名，增强了工具的安全性和网络访问控制。
推荐理由：关注Claude Code的最新功能和安全更新，对于使用该工具的开发者至关重要。
X 推文 (AttentionVC) · 06/14 23:33
微软CEO萨蒂亚·纳德拉发文强调，在AI驱动的经济中，一个没有强大生态系统的「前沿」技术是难以稳定发展的，这表明了生态系统建设在AI转型中的关键作用。
推荐理由：洞察AI时代下企业发展和战略布局的趋势，强调生态系统的重要性。
X 推文 (AttentionVC) · 06/14 21:48
一篇推文指出，在AI时代，一种「微妙但相当危险」的「洗脑」现象正在发生并可能恶化。这引发了对AI技术潜在社会影响和伦理问题的深层思考。
推荐理由：提示我们审慎看待AI技术发展，关注其对人类认知和社会结构的深远影响。
Hugging Face Blog · 06/12 23:56
AllenAI推出了名为「olmo-eval」的评估工作台，旨在为模型开发循环提供全面的评估工具。该平台有助于开发者更有效地测试、分析和改进其AI模型，提升开发效率和模型性能。
推荐理由：为AI模型开发者提供了新的评估工具，有助于提升模型开发效率和质量。
LinuxDo · 06/15 17:55
有社区成员发帖表示，因看到许多用户面临模型使用困难，他手动为社区贡献了50美元的GPT额度，以缓解模型短缺问题，这反映了当前AI模型资源获取的社区互助现象。
推荐理由：了解社区在AI模型短缺背景下的自发行动，体现了用户对AI资源的需求。
iptv-org/iptv
TypeScript · ★ 122,464 · 🍴 6,570 · 📈 2,650 stars today
Collection of publicly available IPTV channels from all over the world
中文介绍 iptv-org/iptv 收集并聚合了全球各地公开可用的 IPTV 频道，提供一个庞大的 M3U 播放列表资源。它解决了用户寻找免费直播电视频道来源的痛点，方便用户通过支持 M3U 格式的播放器观看多样化的全球电视节目。主要面向寻求免费、多语言电视内容的普通用户和技术爱好者。
teslamate-org/teslamate
Elixir · ★ 8,153 · 🍴 943 · 📈 35 stars today
A self-hosted data logger for your Tesla 🚘 [main maintainer=@JakobLichterfeld]
中文介绍 Teslamate 是一款专为 Tesla 车主设计的自托管数据记录器。它允许用户在自己的服务器上收集和存储 Tesla 车辆的详细数据，如行驶里程、充电记录、能耗表现等。该项目解决了车主对车辆数据隐私和深度分析的需求，方便技术爱好者通过其提供的仪表盘深入了解车辆使用情况和性能。
Panniantong/Agent-Reach
Python · ★ 29,469 · 🍴 2,404 · 📈 1,045 stars today
Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.
中文介绍 Agent-Reach 赋予 AI Agent 访问整个互联网的能力，使其能直接阅读和搜索 Twitter、Reddit、YouTube、Bilibili、小红书等平台内容。该项目通过一个 CLI 工具实现，无需 API 费用，有效解决了传统 AI Agent 无法实时获取互联网信息或面临高昂 API 成本的问题，面向希望构建具备实时互联网感知能力的 AI Agent 开发者。
meshery/meshery
TypeScript · ★ 10,531 · 🍴 3,427 · 📈 227 stars today
Meshery, the cloud native manager
中文介绍 Meshery 是一款云原生管理工具，专注于服务网格的生命周期管理。它旨在帮助用户设计、部署、操作和管理各种服务网格（如 Istio、Linkerd），简化云原生环境下的复杂性。该工具适用于运维工程师、SRE 及开发者，以提高微服务架构的可观察性、弹性和安全性。
chatwoot/chatwoot
Ruby · ★ 31,487 · 🍴 7,612 · 📈 431 stars today
Open-source live-chat, email support, omni-channel desk. An alternative to Intercom, Zendesk, Salesforce Service Cloud etc. 🔥💬
中文介绍 Chatwoot 是一款开源的客户支持平台，提供在线聊天、邮件支持及多渠道服务台功能。它为企业提供了一个经济高效的客户沟通与管理解决方案，是 Intercom、Zendesk 等商业产品的优秀替代品。Chatwoot 旨在帮助中小企业和创业公司统一管理客户咨询，提升客户服务效率。
krahets/hello-algo
Java · ★ 126,763 · 🍴 15,130 · 📈 34 stars today
《Hello 算法》：动画图解、一键运行的数据结构与算法教程。支持简中、繁中、English、日本語，提供 Python, Java, C++, C, C#, JS, Go, Swift, Rust, Ruby, Kotlin, TS, Dart 等代码实现
中文介绍 《Hello 算法》是一本以动画图解方式呈现的数据结构与算法教程，支持中英日等多语言，并提供 Python、Java、C++ 等十多种编程语言的代码实现。该项目通过可视化和“一键运行”的代码，极大降低了学习门槛，帮助初学者和准备面试的工程师轻松理解并掌握核心算法概念。
freeCodeCamp/freeCodeCamp
TypeScript · ★ 447,633 · 🍴 44,975 · 📈 738 stars today
freeCodeCamp.org's open-source codebase and curriculum. Learn math, programming, and computer science for free.
中文介绍 freeCodeCamp.org 的开源代码库和课程提供了免费的数学、编程和计算机科学教育资源。该项目旨在通过实践项目和结构化课程，帮助全球学习者免费掌握编码技能，解决就业问题。它适用于任何希望免费学习编程、提升技术能力的个人。
trycua/cua
HTML · ★ 18,033 · 🍴 1,168 · 📈 57 stars today
Open-source infrastructure for Computer-Use Agents. Sandboxes, SDKs, and benchmarks to train and evaluate AI agents that can control full desktops (macOS, Linux, Windows).
中文介绍 CUA (Computer-Use Agents) 提供了一套开源基础设施，用于训练和评估能够完全控制 macOS、Linux、Windows 桌面环境的 AI Agent。它包含了沙盒环境、SDK 和基准测试工具，解决了开发通用桌面操作 AI Agent 所需的复杂环境搭建和评估难题，主要面向 AI 研究者和开发者。
jwasham/coding-interview-university
★ 352,082 · 🍴 83,545 · 📈 516 stars today
A complete computer science study plan to become a software engineer.
中文介绍 Coding Interview University 是一份全面的计算机科学学习计划，旨在帮助学习者系统性地掌握成为软件工程师所需的知识。它详细列出了算法、数据结构、操作系统等面试核心主题，并提供了学习资源。该计划非常适合计算机科学学生和希望转行成为软件工程师的求职者。
rohitg00/ai-engineering-from-scratch
Python · ★ 32,753 · 🍴 5,375 · 📈 538 stars today
Learn it. Build it. Ship it for others.
中文介绍 ai-engineering-from-scratch 是一份从零开始的 AI 工程学习指南。它侧重于 AI 项目从概念到构建再到最终交付的整个生命周期，强调实践和产品化能力。该项目适用于希望系统学习 AI 解决方案开发与部署的工程师、开发者和技术爱好者。
music-assistant/server
Python · ★ 2,327 · 🍴 442 · 📈 226 stars today
Music Assistant is a free, opensource Media library manager that connects to your streaming services and a wide range of connected speakers. The server is the beating heart, the core of Music Assistant and must run on an always-on device like a Raspberry Pi, a NAS or an Intel NUC or alike.
中文介绍 Music Assistant Server 是一个免费开源的媒体库管理器核心服务，它能连接到多种流媒体服务和各类智能音箱。该项目旨在整合用户的分散音乐资源和播放设备，提供一个统一的音乐管理和播放体验，解决了多平台音乐内容难以同步和统一播放的问题，面向拥有多样化音乐生态的用户。
Free-TV/IPTV
Python · ★ 17,147 · 🍴 2,553 · 📈 361 stars today
M3U Playlist for free TV channels
中文介绍 Free-TV/IPTV 提供了免费电视频道的 M3U 播放列表集合。它为用户带来了便捷的免费电视观看方案，只需将其导入支持 M3U 格式的 IPTV 播放器即可收看。该项目主要服务于寻求免费、多样化直播电视内容的普通用户，无需复杂的设置。
Introduction-to-Autonomous-Robots/Introduction-to-Autonomous-Robots
TeX · ★ 2,960 · 🍴 642 · 📈 488 stars today
Introduction to Autonomous Robots
中文介绍 Introduction to Autonomous Robots 提供了自主机器人领域的入门资料。该项目旨在为初学者提供理解自主机器人基本原理和概念的平台，涵盖了导航、感知、决策等关键技术。它适用于对机器人学、人工智能感兴趣的学生和技术爱好者，作为其学习路径的起点。
Raphire/Win11Debloat
PowerShell · ★ 47,750 · 🍴 1,929 · 📈 114 stars today
A simple, lightweight PowerShell script that allows you to remove pre-installed apps, disable telemetry, as well as perform various other changes to declutter and customize your Windows experience. Win11Debloat works for both Windows 10 and Windows 11.
中文介绍 Win11Debloat 是一个轻量级的 PowerShell 脚本，专为 Windows 11 用户设计。它能帮助用户移除预装应用、禁用遥测数据收集，并进行多项系统优化，以提升性能、保护隐私并简化用户体验。该脚本适用于希望定制和清理 Windows 11 系统的技术用户。
mikeroyal/Self-Hosting-Guide
Dockerfile · ★ 20,628 · 🍴 1,048 · 📈 126 stars today
Self-Hosting Guide. Learn all about locally hosting (on premises & private web servers) and managing software applications by yourself or your organization. Including Cloud, LLMs, WireGuard, Automation, Home Assistant, and Networking.
中文介绍 Self-Hosting-Guide 是一份全面的自托管指南，教导用户如何在本地或私有服务器上部署和管理各种软件应用。内容涵盖了云服务、大型语言模型 (LLMs)、WireGuard 等多种技术，解决了个人和组织对数据隐私和自主控制的需求。主要面向技术爱好者、开发者和中小企业。
itsfatduck/optimizerDuck
C# · ★ 3,515 · 🍴 158 · 📈 321 stars today
Free, open-source Windows optimization tool for performance, privacy, and simplicity.
中文介绍 OptimizerDuck 是一款免费开源的 Windows 优化工具，旨在提升系统的性能、增强用户隐私并简化操作。它提供了一系列清理和调整功能，帮助用户摆脱不必要的程序和设置，实现更流畅、更安全的 Windows 体验。适用于追求极致性能和隐私保护的 Windows 用户。
NVIDIA/SkillSpector
Python · ★ 5,970 · 🍴 442 · 📈 1,079 stars today
Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, and security risks.
中文介绍 NVIDIA SkillSpector 是一款专为 AI Agent 技能设计的安全扫描工具。它能够检测 AI Agent 技能中的漏洞、恶意模式和潜在安全风险，确保其运行的安全性和可靠性。该工具解决了 AI 应用在部署前进行安全审计的难题，主要面向 AI Agent 开发者和企业安全团队。
shiyu-coder/Kronos
Python · ★ 30,123 · 🍴 5,173 · 📈 395 stars today
Kronos: A Foundation Model for the Language of Financial Markets
中文介绍 Kronos 是一个专注于金融市场语言的基础模型。它旨在理解、分析和生成与金融行业相关的文本信息，如财报、新闻、市场报告等。该模型解决了传统 AI 在金融领域专业知识不足的问题，为金融分析师、量化研究员和金融科技公司提供强大的自然语言处理能力，支持市场预测和风险评估。
AdaSR: Adaptive Streaming Reasoning with Hierarchical Relative Policy Optimization
👍 0
Large reasoning models typically follow a read-then-think paradigm: they observe the complete input, reason over a static context, and then produce the answer. Yet many real-world scenarios are inherently dynamic, such as audio and video stream, where information arrives as a continuous stream and m
中文介绍 现有大型推理模型通常采用“先读后思”模式处理完整静态输入。针对音视频流等信息持续到达的动态场景，本文提出AdaSR（Adaptive Streaming Reasoning）框架。该框架通过分层相对策略优化，旨在使大模型能够适应动态环境下的流式推理，突破传统模型的静态上下文限制，实现更实时的信息处理。
RepFusion: Leveraging Multimodal Priors for Denoising in Representation Space
👍 7
Large language models (LLMs) are widely used in text-to-image (T2I) systems, but they are typically limited to text encoding, while denoising is handled by newly trained generative backbones. The emergence of representation autoencoders (RAEs) shifts the generation target toward semantically structu
中文介绍 大型语言模型（LLMs）在文生图（T2I）系统中主要负责文本编码，去噪任务通常由独立的生成骨干网络处理。本文提出RepFusion模型，利用多模态先验知识，结合表征自编码器（RAEs）将生成目标转移到表征空间，以在该空间内进行去噪。此方法旨在通过融合多模态信息，提升文生图系统的生成质量和效率。
LLM Agents Can See Code Repositories
👍 11
Coding agents powered by large language models have demonstrated strong performance on software engineering tasks. Yet most agents consume repositories almost entirely as text, which differs from how human developers use visual structure such as folder hierarchies and dependency relationships to ori
中文介绍 基于大语言模型（LLM）的编码智能体在软件工程任务中表现出色。然而，多数智能体将代码仓库几乎完全视为文本处理，这与人类开发者利用文件夹层级、依赖关系等视觉结构的方式不同。本文探讨如何使LLM智能体能像人类一样“看到”代码仓库的结构信息，从而更有效地理解和操作复杂的代码库，进一步提升其在软件开发中的应用能力。
HarnessX: A Composable, Adaptive, and Evolvable Agent Harness Foundry
👍 31
AI agent performance depends critically on the runtime harness, comprising the prompts, tools, memory, and control flow that mediate how a model observes, reasons, and acts. Yet today's harnesses remain largely hand-crafted and static: each new model or task still demands bespoke scaffolding, and th
中文介绍 AI智能体的性能关键取决于其运行时“套件”（Harness），包括提示词、工具、内存和控制流等。当前套件大多是手工定制且静态的，每引入新模型或任务都需要量身打造。本文提出HarnessX，一个可组合、自适应、可进化的智能体套件铸造平台，旨在通过模块化设计和自动化生成，提升智能体开发的效率和灵活性，以适应不同模型与任务的需求。
ClinHallu: A Benchmark for Diagnosing Stage-Wise Hallucinations in Medical MLLM Reasoning
👍 4
Building trustworthy medical multimodal large language models (MLLMs) is critical for reliable clinical decision support. Existing medical hallucination benchmarks mainly focus on data collection, but often ignore where hallucinations originate within the reasoning process. We find that hallucinatio
中文介绍 构建可信赖的医疗多模态大语言模型（MLLMs）对临床决策支持至关重要。现有医疗幻觉基准主要侧重数据收集，但常忽略幻觉在推理过程中的具体产生阶段。本文提出ClinHallu，一个用于诊断医疗MLLM推理中“分阶段幻觉”的基准，旨在精确识别幻觉的源头，从而指导模型改进，提升其在医疗领域的可靠性和安全性。
From Chatbot to Digital Colleague: The Paradigm Shift Toward Persistent Autonomous AI
👍 41
Large Language Models (LLMs) are undergoing a fundamental transformation from conversational generators into integrated AI systems capable of reasoning, action, memory, and self-improvement. We conceptualize this transition as a shift from Chatbot to Digital Colleague: from conversational answers to
中文介绍 大型语言模型（LLMs）正从简单的对话生成器向具备推理、行动、记忆和自我改进能力的集成式AI系统转型。本文将这一根本性转变概念化为“从聊天机器人到数字同事”的范式转移，强调了持久化自主AI的发展趋势。此转变预示着LLMs将能更深入地融入工作流程，成为具备持续学习和独立解决问题能力的智能协作伙伴。
OmniVideo-100K: A Dataset for Audio-Visual Reasoning through Structured Scripts and Evidence Chains
👍 19
Current automated pipelines for audio-visual Question Answering (QA) generally adopt a ``video-caption-QA'' paradigm. However, these methods typically segment videos into short clips and generate separate descriptions for audio and visual modalities. This decoupled processing severs inherent associa
中文介绍 当前音视频问答（QA）自动化流程多采用“视频-字幕-问答”范式，常将视频分割成短片段并分别描述音视觉模态，导致处理脱节。本文提出OmniVideo-100K数据集，旨在通过结构化脚本和证据链，促进对音视频模态的整体理解和推理。该数据集提供更全面的多模态信息，以克服现有方法在处理长视频和复杂语境时的局限性。
Rethinking RAG in Long Videos: What to Retrieve and How to Use It?
👍 31
Retrieval-augmented generation is moving beyond text into long, egocentric video, where systems must select query-relevant chunks across multiple modalities and temporal granularities. Yet progress in VideoRAG is limited by two gaps: existing benchmarks allow queries to be answered without the video
中文介绍 检索增强生成（RAG）技术正从文本领域扩展至长时程、主视角视频，要求系统从多模态和不同时间粒度中选择与查询相关的视频片段。然而，VideoRAG的进展受限于现有基准的不足。本文重新审视长视频中的RAG问题，重点探讨“检索什么”以及“如何使用检索到的信息”，旨在优化长视频内容理解和生成效率，以弥补当前方法的不足。
OmniDirector: General Multi-Shot Camera Cloning without Cross-Paired Data
👍 89
Cloning camera motion from reference videos is an important task in video generation, as videos provide intuitive and precise control. Existing methods either directly use parametric representations that fail to handle multi-shot generation or synthesize cross-paired data, which suffer from data sca
中文介绍 从参考视频中克隆摄像机运动是视频生成中的一项重要任务。现有方法要么直接使用参数化表示但无法处理多镜头生成，要么需要合成交叉配对数据，效率低下。本文提出OmniDirector模型，旨在无需交叉配对数据的情况下，实现通用的多镜头摄像机运动克隆。此方法有望提供更直观、精确的视频生成控制，同时提高工作效率。
The Hidden Power of Scaling Factor in LoRA Optimization
👍 5
In Low-Rank Adaptation (LoRA), the scaling factor α is often treated as a mere complement to the learning rate, yet its role in optimization remains poorly understood. In this paper, we reveal that the scaling factor α and the learning rate function differently, with α emerging as the dominant drive
中文介绍 在低秩适应（LoRA）技术中，缩放因子α常被视为学习率的辅助，其在优化中的作用尚未被充分理解。本文揭示了缩放因子α与学习率的功能差异，指出α并非简单的补充。研究深入探讨了α的“隐藏力量”，阐明了它在模型训练和性能优化中的独特机制和关键影响，为LoRA的进一步应用和改进提供了新视角。
See What I See, Know What I Think: Dense Latent Communication Across Heterogeneous Agents
👍 3
Multi-agent systems communicate mostly through text, paying a lossy and expensive decode and re-encode cost. KV-cache communication is a promising alternative, yet most prior work is homogeneous, using duplicate copies of the same model, and avoids the central challenge of cross-model latent alignme
中文介绍 多智能体系统间的通信主要通过文本进行，这会产生有损且昂贵的编解码成本。KV-缓存通信是一种有前景的替代方案，但现有研究多限于同质智能体，并回避了异质智能体间通信的核心挑战。本文提出“异质智能体间的密集潜在通信”方法，旨在通过共享高维潜在表示，实现更高效、低损耗的信息传递，从而提升多智能体系统的协作能力和整体性能。
Getting Better at Working With You: Compiling User Corrections into Runtime Enforcement for Coding Agents
👍 4
Interactive LLM agents are becoming part of daily work, but they do not reliably become easier to work with over time: a correction remembered in one session may still be violated in the next. We study this gap between preference access and preference compliance. In tasks derived from anonymized rea
中文介绍 交互式大语言模型（LLM）智能体已融入日常工作，但其并非总能可靠地从用户修正中学习：一次会话中记住的修正可能在下次被违反。本文研究了偏好获取与偏好遵循之间的差距，提出将用户修正编译为编码智能体的运行时强制执行规则。此方法旨在通过动态调整智能体行为，确保其能持续遵循用户指令，提升长期协作效率。
APPO: Agentic Procedural Policy Optimization
👍 59
Recent advances in agentic Reinforcement Learning (RL) have substantially improved the multi-turn tool-use capabilities of large language model agents. However, most existing methods assign credit over coarse heuristic units, such as tool-call boundaries or fixed workflows, making it difficult to id
Measuring Epistemic Resilience of LLMs Under Misleading Medical Context
👍 12
Large language models (LLMs) now reach expert-level scores on medical licensing exams, encouraging the assumption that high scores imply safe medical judgment while patients increasingly use them for health advice. We show this assumption is fragile: when misleading context is injected into question
Benchmarking AI Agents for Addressing Scientific Challenges Across Scales
👍 1
AI agents are increasingly being developed to accelerate scientific discovery, yet their practical capabilities in real research settings remain poorly understood. Existing benchmarks for AI agents rarely capture the complexity, heterogeneity, and extended reasoning required by scientific work, wher
RedAct: Redacting Agent Capability Traces for Procedural Skill Protection
👍 15
Users rely on execution traces to observe agent behavior, diagnose failures, and ensure accountability. These traces contain rich procedural detail, including tool invocations, intermediate decisions, and error-recovery logic. Yet this detail can expose private procedural skills, allowing downstream
Orchestra-o1: Omnimodal Agent Orchestration
👍 35
The recent success of agent swarms has shifted the paradigm of large language model (LLM)-based agents from single-agent workflows to multi-agent systems, highlighting the importance of agent orchestration for task decomposition and collaboration. However, existing orchestration frameworks are limit
From AGI to ASI
👍 19
Over the last decade, building human-level artificial general intelligence has moved from far-fetched speculation to being a concrete next-decade target for many of the largest AI organisations. Achieving this goal would have profound and far-reaching impacts on human society, which raises many comp
When is Your LLM Steerable?
👍 3
Activation steering offers a lightweight approach to control language models' behavior at inference time, but whether it succeeds or fails heavily depends on the prompt, concept, model, and steering configuration. Finding the regime and boundaries of successful steering typically requires expensive
Rethinking Psychometric Evaluation of LLMs: When and Why Self-Reports Predict Behavior
👍 5
Anticipating LLM behavioral tendencies from low-cost psychometric probes is critical for safe deployment, but only if self-reports (SR) reliably predict behavior. Recent work documented substantial SR-behavior dissociation in LLMs, but relied on broad personality traits (Big 5) that predict specific
The Arbiter Agent: Continually Monitoring Multi-Agent Conversations to Detect Emergent Misalignment
👍 2
As AI systems built from multiple language-model agents become more common, they are increasingly used to make decisions together: discussing, negotiating, and acting on shared tasks. While individual agents may appear well-aligned when tested on their own, problems can arise from how they interact
P3D-Bench: Benchmarking MLLMs for Parametric 3D Generation and Structural Reasoning
👍 2
Multimodal large language models can write code to produce complex programs as well as use programs to do 3D modeling, which opens up a new avenue for 3D generation powered by their priors, world knowledge and reasoning. Yet existing benchmarks rarely evaluate 3D modeling through code. Such modeling
WebChallenger: A Reliable and Efficient Generalist Web Agent
👍 2
Autonomous web navigation remains challenging for LLM agents, and the strongest generalist systems rely on proprietary reasoning models whose inference cost is prohibitive for the repetitive tasks where such agents would be most useful. We argue this gap stems not from insufficient model capability
AlloSpatial: Agentic Harness Framework for Spatial Reasoning in Foundation Models
👍 0
Multimodal Foundation Models (MFMs) have made substantial progress, yet remain fragile in spatial reasoning over the physical world. A key bottleneck lies in their inability to transform local egocentric observations into a global allocentric spatial representation. To address this, we propose AlloS
The Cold-Start Safety Gap in LLM Agents
👍 2
Are tool-calling LLM agents equally safe throughout a conversation? We discover they are not: agents are most vulnerable at the very start of a session and become substantially safer after a few regular agentic tasks -- a phenomenon we term the cold-start safety gap. To study this systematically, we
Memory is Reconstructed, Not Retrieved: Graph Memory for LLM Agents
👍 53
Despite recent progress, LLM agents still struggle with reasoning over long interaction histories. While current memory-augmented agents rely on a static retrieve-then-reason paradigm, this rigid pipeline design prevents them from dynamically adapting memory access to intermediate evidence discovere
Skip a Layer or Loop It? Learning Program-of-Layers in LLMs
👍 10
Large language models (LLMs) perform inference by following a fixed depth and order, non-recurrent execution of all layers. We reveal the wide existence of training-free, flexible, dynamic program-of-layers (PoLar), where pretrained layers can be packed as modules and then skipped or looped to form
Smaller Models are Natural Explorers for Policy-Level Diversity in GRPO
👍 18
We identify a new dimension for enhancing rollout diversity in Group Relative Policy Optimization (GRPO) for LLMs. While GRPO relies on diverse rollouts, prevailing strategies primarily increase diversity by injecting more token-level randomness, which may introduce step-wise noise and lead to incoh
CARVE: Certified Affordable Repair of Vetoed Maneuvers via Envelopes for Interactive Driving
👍 0
Interactive driving exposes a failure mode that is easy to miss in rule-aware autonomous-driving stacks: a hard-rule margin can be negative for an ego candidate even though a small lawful accommodation by a non-priority agent would restore feasibility. Existing rulebooks, shields, and reachability f
On the Limits of LLM Adaptability: Impact of Model-Internalized Priors on Annotation Task Performance
👍 0
Large Language Models (LLMs) are increasingly used for zero-shot annotation and LLM-as-a-judge tasks, yet their reliability hinges on how model-internalized priors interact with user-provided instructions. We investigate three dimensions of this interaction: (1) how an LLM's familiarity with data an
Kickbacks.ai
Get paid to wait for Claude Code to finish
AgentBrush
Your coding agent's missing tool: image generation
MiMo Code
A coding agent with explicit long-term memory architecture
Reignat
Privacy-friendly web analytics platform built for makers
Dropmatico
Drop. Pick. Done.
EmailFlow.AI
Like Claude Design for Email Newsletters
IdleDev
Get paid while your AI agent thinks
stackd.cc
The answer to "what's your AI stack?"
VEXI
Open-source AI coding agent for your terminal
Novu Connect
Ship agents where your users already work
A frontier without an ecosystem is not stable
@satyanadella · 5.9M 粉丝 · 1.9M 阅 · 3.5K 赞 · 541 转
I’ve been thinking a lot about the future of the firm in an AI-driven economy. This transition is different than any previous platform shift. In the past, we used digital systems to enhance human
Loops: What Every AI Engineer Needs to Know in 2026
@sairahul1 · 113.0K 粉丝 · 852.6K 阅 · 600 赞 · 79 转
Peter Steinberger, creator of OpenClaw, who now works with OpenAI. Yesterday he posted this: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents."
Codex-maxxing: treating Codex like an operating loop
@BradGroux · 5.9K 粉丝 · 714.6K 阅 · 1.0K 赞 · 638 转
Most people still use coding agents like fancy autocomplete or a one-shot chat box. That leaves a lot of value on the table. The better pattern is to treat Codex like a durable operating loop:
Fable 5 (Mythos) Prompting Masterclass by Anthropic
@aiedge_ · 69.5K 粉丝 · 700.1K 阅 · 506 赞 · 68 转
TLDR: Anthropic just published the official playbook for prompting the most powerful AI model on earth - I translated it. Most people won't read this guide (it's buried in the API docs), which is
Everything Is Recorded Now
@dhaber · 50.0K 粉丝 · 497.3K 阅 · 500 赞 · 57 转
One of the biggest ways that AI is transforming work (and also one of the most taboo subjects inside companies at the moment) is that most work discussions are being recorded now by default. This
First Steps Toward Automated AI Research
@Recursive_SI · 6.3K 粉丝 · 465.1K 阅 · 516 赞 · 71 转
Early results from Recursive’s automated AI research system on model training and GPU kernel benchmarks Today we are releasing early results from Recursive’s automated AI research system. Across three
Build self-improving agent system with Fable 5 in 14 steps : loops, dynamic workflows, routines
@0xCodez · 6.4K 粉丝 · 371.8K 阅 · 515 赞 · 56 转
Most people are using Claude Fable 5 like Sonnet 4.6 with a bigger context window. They prompt it. It works for 5 minutes. They close the tab. 9 out of 10 users have never run an agent system that
Anthropic is losing the mandate of heaven
@haridigresses · 12.5K 粉丝 · 281.7K 阅 · 513 赞 · 36 转
Four months ago, in early February, Anthropic was the darling. OpenAI was the dominant behemoth to root against. Over the last 1-2 years, we'd seen the Sam ouster / return drama, Ilya and Mira had
An extremely subtle brainwashing
@AnthonyNAguirre · 4.2K 粉丝 · 232.1K 阅 · 527 赞 · 38 转
I've been getting really bothered lately by something subtle but quite dangerous that I believe is currently going on and promises to get worse: Large numbers of very smart, capable, and important
The Untrainable
@saranormous · 143.5K 粉丝 · 194.8K 阅 · 614 赞 · 40 转
The mid-2026 investor's version of AI psychosis is a despair that nothing is investable, that we should put all our money into Anthropic and Nvidia and go home. I have never felt it. I have been sure
How to Build a Self-Improving Loop in Claude Code (Exact Setup Inside)
@0x_rody · 1.7K 粉丝 · 193.2K 阅 · 513 赞 · 72 转
Claude writes your code, hands it over, and 3 tests are failing. You paste the errors back, it fixes one thing, breaks another, and you spend the evening as a messenger between Claude and your
Building a Good Vertical Agent
@BrainsAndTennis · 10.5K 粉丝 · 187.4K 阅 · 539 赞 · 45 转
How do you build an agent that actually performs in a domain — one customers pick because it's better? The basics have been standardized over the past year: an agent is a while-loop around a model
My Week with Fable
@MatthewBerman · 121.3K 粉丝 · 108.0K 阅 · 661 赞 · 26 转
tl;dr I've been testing Fable (Mythos) for the past week and it feels unlike any other model I've used. It feels, and is priced, like a next-generation model. It also has some real quirks. The Good
Kimi to Predict All 104 World Cup Matches: Germany May Be Underestimated
@Kimi_Moonshot · 172.7K 粉丝 · 106.6K 阅 · 500 赞 · 61 转
Our predictions will probably be wrong. But the World Cup offers a rare, public, verifiable, and constantly evolving real-world setting. Through this initiative, we hope to place analysis,
Loop engineering: the 14-step roadmap from prompter to loop designer.
@0xCodez · 6.4K 粉丝 · 97.8K 阅 · 510 赞 · 80 转
Most developers still prompt their coding agents by hand. They type, they wait, they read the diff, they type again. 9out of 10 builders have never written a single loop that prompts the agent for
Designing loops with Fable 5
@RLanceMartin · 30.4K 粉丝 · 84.7K 阅 · 660 赞 · 50 转
Mythos-class models like Claude Fable 5 have changed the way many of us work at Anthropic. I want to share two tips for getting the most out of this class of models. Self-correction loops There’s been
Autonomous Long-Running Coding Agents
@omarsar0 · 307.3K 粉丝 · 81.2K 阅 · 518 赞 · 66 转
Autonomous coding is moving from better prompting to better control systems. The important shift is that engineers are learning how to wrap agents in goals, evaluators, loops, and artifacts that let
Anthropic's War on Opensource AI
@TheAhmadOsman · 61.0K 粉丝 · 74.9K 阅 · 507 赞 · 98 转
Anthropic wants the public to see one thing: the careful lab, the safety lab, the grown-up in the room trying to keep frontier AI from running off a cliff. However, the pattern around Anthropic does
Coinbase for Agents: Your AI Agent Can Now Trade and Pay with Coinbase
@coinbase · 7.0M 粉丝 · 72.8K 阅 · 500 赞 · 62 转
TL;DR: Coinbase for Agents connects your AI agent directly to your Coinbase account so it can trade, pay, and execute workflows on your behalf, all within limits you control. Available today as an MCP
Principled Thinking and AI Need to Go Together
@RayDalio · 2.2M 粉丝 · 72.6K 阅 · 515 赞 · 93 转
What is the best approach to being effectively intelligent now that human intelligence and artificial intelligence are merging? Because I have been building computerized investment decision-making
The Window Has Closed
@AndrewCurran_ · 53.9K 粉丝 · 62.8K 阅 · 7d 曝光 62.8K
The Window Has Closed
A frontier without an ecosystem is not stable
@satyanadella · 5.9M 粉丝 · 1.9M 阅 · 7d 曝光 1.9M
A frontier without an ecosystem is not stable
An extremely subtle brainwashing
@AnthonyNAguirre · 4.2K 粉丝 · 232.1K 阅 · 7d 曝光 232.1K
An extremely subtle brainwashing
The 7-day Hermes setup (full guide)
@zaimiri · 54.4K 粉丝 · 46.9K 阅 · 7d 曝光 46.9K
The 7-day Hermes setup (full guide)
Mastering Codex (Mobile) for Engineering
@Dimillian · 51.8K 粉丝 · 38.2K 阅 · 7d 曝光 38.2K
Mastering Codex (Mobile) for Engineering
Codex-maxxing: treating Codex like an operating loop
@BradGroux · 5.9K 粉丝 · 714.6K 阅 · 7d 曝光 714.6K
Codex-maxxing: treating Codex like an operating loop
Autonomous Long-Running Coding Agents
@omarsar0 · 307.3K 粉丝 · 81.2K 阅 · 7d 曝光 81.2K
Autonomous Long-Running Coding Agents
Anthropic is losing the mandate of heaven
@haridigresses · 12.5K 粉丝 · 281.7K 阅 · 7d 曝光 281.7K
Anthropic is losing the mandate of heaven
Anduril and CBP Expand Partnership with 200 More Sentry Towers
@anduriltech · 243.1K 粉丝 · 24.5K 阅 · 7d 曝光 24.5K
Anduril and CBP Expand Partnership with 200 More Sentry Towers
Building recursive agent systems
@leerob · 258.6K 粉丝 · 36.8K 阅 · 7d 曝光 36.8K
Building recursive agent systems
Why the Government Just Killed Claude Fable 5
中文介绍 该视频探讨了政府停止或禁止「Claude Fable 5」项目的原因。具体原因和「Claude Fable 5」的性质在现有信息中未详细说明。
This is Why They Banned Claude Fable 5
中文介绍 该视频讨论了「Claude Fable 5」项目被禁止的原因。目前没有进一步信息说明其具体被禁原因及该项目的详细情况。
Hermes Agent NEW Super-App and DeepSeek v4 Catches Up To Opus 4.8?
ChatGPT And Codex Are Merging (This Changes Everything)
Claude Opus 4.8: Everything You Need to Know
Browsers Are Dead. Claude Just Replaced Them.
Every NEW Claude Feature Explained
Codex: Build Your Full AI Marketing Team (Agents + Skills)
OpenAI just released Codex Mobile
Codex Mobile App Released (Complete Setup Guide)
Claude FM 🎵 music for thinking and building
中文介绍 这是一个由Claude发布的音乐播放列表或频道，旨在为用户提供用于思考和工作背景的音乐。
Claude FM 🎵 music for thinking and building
中文介绍 这是一个由Claude发布的音乐播放列表或频道，旨在为用户提供用于思考和工作背景的音乐。
Code with Claude Tokyo 2026: Opening Keynote
中文介绍 Claude发布了其名为「Code with Claude Tokyo 2026」的活动开幕主题演讲视频。该活动预计于2026年在东京举行，主要内容可能围绕使用Claude进行编程相关主题。
The Problem Solvers | Michael Truell at Cursor
中文介绍 Claude发布了Michael Truell在Cursor公司关于「问题解决者」的演讲视频。视频具体内容涉及Michael Truell对问题解决的见解。
Claude Fable 5 beats Pokémon FireRed only using vision
中文介绍 Claude发布的视频展示了名为「Claude Fable 5」的系统，仅通过视觉输入成功击败了游戏《宝可梦 火红》。这表明「Claude Fable 5」可能具备先进的视觉理解和游戏操作能力。
Claude Fable 5 plays Factorio
Claude Fable 5 simulates the solar system and predicts a solar eclipse
Claude Fable 5 sets a fluid simulation to Beethoven
Claude Fable 5 designs a 3D-printable model in a Claude-built CAD editor
Working Like a Lawyer with Claude
Claude FM 🎵 music for thinking and building
中文介绍 这是一个由Claude发布的音乐播放列表或频道，旨在为用户提供用于思考和工作背景的音乐。
Claude FM 🎵 music for thinking and building
中文介绍 这是一个由Claude发布的音乐播放列表或频道，旨在为用户提供用于思考和工作背景的音乐。
Code with Claude Tokyo 2026: Opening Keynote
中文介绍 Claude发布了其名为「Code with Claude Tokyo 2026」的活动开幕主题演讲视频。该活动预计于2026年在东京举行，主要内容可能围绕使用Claude进行编程相关主题。
The Problem Solvers | Michael Truell at Cursor
中文介绍 Claude发布了Michael Truell在Cursor公司关于「问题解决者」的演讲视频。视频具体内容涉及Michael Truell对问题解决的见解。
Claude Fable 5 beats Pokémon FireRed only using vision
Claude Fable 5 plays Factorio
Claude Fable 5 simulates the solar system and predicts a solar eclipse
Claude Fable 5 sets a fluid simulation to Beethoven
Claude Fable 5 designs a 3D-printable model in a Claude-built CAD editor
Working Like a Lawyer with Claude
NVIDIA's New Free AI - A Gift To Humanity
中文介绍 NVIDIA（英伟达）发布了一款新的免费人工智能技术，具体细节和功能在现有信息中未详细说明。该视频标题强调此AI对人类的益处。
AI Agents as "Games Masters"? 🎮🔥
DeepMind’s New AI Found A Strange New Way To Think
Meet the AI "Co-Scientist" Changing Everything 🤖🧪 #ai
Claude Opus 4.8: Lying Machine No More?
A Second Nobel Prize for AlphaFold? 🧬🏆 #alphafold #deepmind #nobelprize #science #ai
What Happens After A 1,000,000x AI Compute Leap? | Jeff Dean
Einstein vs Feynman, Who Wins? 🧠🤔 #physics #ai #science #feynman #research
Google DeepMind CEO Loves Hard Questions 🙂
Demis Hassabis On What AI Will Do Next
Anthropic Fable shutdown 🏛️, GLM-5.2 🤖, OpenRouter Fusion 🧠
中文介绍 TLDR AI快讯报道，Anthropic旗下的Fable项目宣布关闭。此外，新闻还提及了GLM-5.2的发布以及OpenRouter Fusion的动态。这些信息反映了近期AI领域在项目进展和技术发布方面的一些焦点。
Introducing the OpenAI Partner Network
OpenAI launches the Partner Network, investing $150M to help global partners accelerate enterprise AI adoption, deployment, and transformation.
中文介绍 OpenAI正式推出“合作伙伴网络”（Partner Network），并计划投资1.5亿美元，以支持全球合作伙伴加速企业级AI解决方案的采纳、部署和转型，旨在推动AI在商业领域的应用与发展。
[AINews] Fable and Mythos officially too dangerous to release
We are in the strangest timeline.
中文介绍 据报道，AI项目Fable和Mythos因被认为「过于危险」而正式停止发布。此消息反映了业界对AI技术潜在风险的担忧，可能指向了模型在安全性和可控性方面遇到的挑战。
olmo-eval: An evaluation workbench for the model development loop
中文介绍 AllenAI推出了名为「olmo-eval」的评估工作台，旨在为模型开发循环提供全面的评估工具。该平台有助于开发者更有效地测试、分析和改进其AI模型，提升开发效率和模型性能。
New OpenAI Academy courses for the next era of work
OpenAI introduces three Academy courses that help people build practical AI skills, create repeatable workflows, and apply agents in everyday work.
中文介绍 OpenAI学院推出了三门新课程，旨在帮助人们掌握实用的AI技能，创建可重复的AI工作流程，并将AI代理（agents）应用于日常工作中。这些课程面向下一代工作，以加速AI技术的普及和应用。
[AINews] Loopcraft: The Art of Stacking Loops
a quiet day lets us highlight a great concept from Peter Steinberger, Boris Cherny, and Andrej Karpathy
中文介绍 本篇新闻探讨了「Loopcraft：循环堆叠的艺术」这一概念，该思想由Peter Steinberger、Boris Cherny和Andrej Karpathy等专家提出。它可能涉及优化编程循环或AI模型迭代过程中的效率和结构。
How Preply combines AI and human tutors to personalize learning
Preply uses OpenAI to launch AI-generated lesson summaries, providing personalised feedback and language learning exercises.
中文介绍 在线学习平台Preply与OpenAI合作，利用AI技术推出AI生成的课程摘要，为用户提供个性化反馈和语言学习练习。此举旨在将AI与真人导师结合，共同优化个性化学习体验，提升语言学习效率。
OpenAI buys Ona 🤝, Anthropic backtracks 🔁, Xiaomi’s MiMo code 👨💻
中文介绍 TLDR AI快讯报道，OpenAI计划收购Ona公司，Anthropic在某些决策上出现反复，而小米则推出了MiMo代码项目。这些事件反映了AI领域近期在企业并购、战略调整及技术开源方面的动态。
Google DeepMind is worried about what happens when millions of agents start to interact
Google DeepMind is funding research into the potential dangers of situations where millions of different AI agents interact with each other online. According to Rohin Shah, who directs the company’s AGI safety and alignment research, the mass-market arrival of agents that can carry out tasks without
中文介绍 Google DeepMind正资助一项研究，以探究数百万个AI代理在网络上相互作用时可能产生的潜在危险。公司AGI安全与对齐研究负责人Rohin Shah表示，随着AI代理大规模普及，其互动可能带来不可预测的风险，引发业界对AI安全问题的担忧。
[AINews] Open Models, Model Labs vs Agent Labs, and What's Untrainable — Sarah Guo
a quiet day lets us reflect on a great essay
中文介绍 本篇新闻基于Sarah Guo的一篇文章，探讨了开放模型（Open Models）的趋势、模型实验室（Model Labs）与代理实验室（Agent Labs）之间的区别，以及AI领域中「不可训练」的挑战。
Supporting Europe’s work in ensuring a trustworthy AI ecosystem
OpenAI supports the EU Code of Practice on AI content transparency, advancing provenance standards and tools to help people understand AI-generated content.
中文介绍 OpenAI宣布支持欧盟的《AI内容透明度行为准则》，旨在推动内容溯源标准和工具的进步，以帮助公众更好地理解和识别AI生成的内容。此举是OpenAI为构建可信赖的欧洲AI生态系统所做的努力。
BBVA puts AI at the core of banking with OpenAI
Learn how BBVA scaled ChatGPT Enterprise to 100,000 employees and partnered with OpenAI to accelerate AI-powered banking transformation worldwide.
中文介绍 BBVA银行与OpenAI合作，将ChatGPT企业版扩展至10万名员工使用，旨在加速其全球AI驱动的银行业务转型。此举表明BBVA正将AI技术置于银行核心业务，以提升效率和创新服务。
OpenAI to acquire Ona
OpenAI plans to acquire Ona to expand Codex with secure, persistent cloud environments, enabling long-running AI agents across enterprise workflows.
中文介绍 OpenAI宣布计划收购Ona公司，旨在通过安全、持久的云环境扩展其Codex能力。此次收购将赋能AI代理在企业工作流中实现长期运行，进一步提升AI在商业应用中的效率和稳定性。
How an astrophysicist uses Codex to help simulate black holes
Discover how astrophysicist Chi-kwan Chan uses Codex to build black hole simulations, helping scientists study extreme physics and test Einstein’s theory of general relativity.
中文介绍 天体物理学家Chi-kwan Chan正利用OpenAI的Codex工具，构建黑洞模拟。这项应用有助于科学家研究极端物理现象，并验证爱因斯坦的广义相对论，展示了AI在复杂科学研究中的潜力。
Profiling in PyTorch (Part 2): From nn.Linear to a Fused MLP
中文介绍 本文是关于PyTorch性能分析的第二部分，重点探讨了如何从标准的`nn.Linear`层优化到融合的多层感知机（Fused MLP）。该技术旨在提升模型训练和推理的效率，对深度学习开发者具有参考价值。
sing-box 性能确实强，但是配置是真的难，各种不兼容
5 回复 · 程序员 节点
大家的“公益短信”能够按照关键词屏蔽么？
26 回复 · 程序员 节点
买 iwatch s11 还是等 iwatchs12 呢 ？
12 回复 · Apple 节点
花云-flowercloud 是跑路了吗, 线路极其不稳定,持续很久了
18 回复 · 程序员 节点
GPT-5.5 真能让人用得舒心吗？怎么在我这跟弱智一样
69 回复 · 程序员 节点
ios27 新 siri 问一些联网问题就报 problem，本地化问题没问题，是节点不对？
6 回复 · Apple 节点
想把我的国航 air 卖了换港行 promax 有必要吗 尘埃未定
13 回复 · Apple 节点
ios27 磁盘占有太大
5 回复 · Apple 节点
介于国行 iPhone 和外版差距越来越大，想换国内安卓了，求推荐品牌系列
65 回复 · Apple 节点
有一个 Apple ID，注册了 beta.apple，为什么只能收到 iOS 26 beta，没有 27 beta
11 回复 · Apple 节点
看的出来大家确实是没模型用了，于心不忍我手动给他加了50刀额度
本来因为我号池PLUS号不多了，已经关闭了GPT渠道 做公益站真的很矛盾，每次看到号池见底 这个是昨天新搭建的服务器渠道 18 个帖子 - 18 位参与者 阅读完整话题
求顶，让管理们看到（
事先声明：我并非有意带节奏，但是接连问了两个管理都没有人理我（可能是比较忙），这事对我真的很重要，迫于无奈，我只好发帖求助 事情是这样的，我从去年开始追求连续365天上线，可是Discourse论坛系统出了问题，我在阅读时间为0的这一天很明确的上线了，而且不是上线了事，是浏览了不少帖子 我已经连着上了10个月了，这是最后两个月，我不想因为这种bug断掉，只好发帖求助 48 个帖子 - 29 位参与者 阅读完整话题
百倍佬把签到功能放出来了，终于不用眼巴巴望着了
感谢百倍佬，每次放额度都十分大气，做到这个地步已经很感激不尽，同时也十分感谢所有尽心维护公益站的佬们，能让我们有免费学习体验的机会，各位佬辛苦了 注意：每次佬的号池维护了能蹬会通知，现在还蹬不了佬友们 42 个帖子 - 37 位参与者 阅读完整话题
百倍佬的 Sub2API 公益站加了每日签到，快去签到~
先攒着再说，等能用的时候狠狠狠的蹭一下~ 新佬们，我不是这个公益站站长，我跟你们一样是用户，我只是提醒大家一下今天新增了签到。 51 个帖子 - 38 位参与者 阅读完整话题
节奏带得飞起，但这有什么好争的？
今天公益站的相关帖子很热闹，也不乏一些人浑水摸鱼带节奏。但我没想明白，这玩意有什么可争的？ 这个事情追根究底就是两个点： 没掺假，那其实没什么好说的。 掺假了，这其实也没什么好说的，肯定不对。 掺假不应该。 这一点，我想应该是共识没得洗。有佬友热心做公益是好事，不过有多少米煮多少饭。能做多久做多久，有什么模型提供什么模型，大家都不会说什么，更不会怪罪。但是弄虚作假，以次充好就是另一回事了，没必要。 至于说掺假的，我希望你们用数据和证据说话。捕风捉影的帖子，恕不接受。 虚空打靶，没有事实根据的帖子，不管是正方还是反方，大家点点举报。 329 个帖子 - 282 位参与者 阅读完整话题
华为出code cli了，无限用GLM5.1
DevEco Code 目前可以免费使用GLM5.1 如图，基于 OpenCode 扩展开发 73 个帖子 - 58 位参与者 阅读完整话题
分享一个零成本刷新codex额度的方法
现在codex出了一个邀请新用户刷新额度的活动，给大家分享一个不用接码就可以实现的方法： 点击进去之后能够看到： 然后选择好仓库，然后随便打个招呼就可以了，你的账号就会收到邮箱显示获取刷新额度的机会 28 个帖子 - 20 位参与者 阅读完整话题
佬友们，我最近发现一个分区。
佬友们，我最近发现一个分区叫做 快问快答 。 20 个帖子 - 20 位参与者 阅读完整话题
我真的引流过吗？
君的公益建立至今，只有一条公告 我没要LDC，我也没开充值，我也没强迫大家使用，没有建立任何私域（各种群），如果一定要说我引流，那也是往LD社区引流 我不是第一天开公益站，我也不是第一天加入LD社区 服务器自费，域名自费，cf pro自费 渠道自己想办法,之前gpt资源很多的时候，有点飘，开了qq邮箱注册，现在想起来也是有点后悔的 服务器一个月两百美元的样子，域名不贵，cfpro点错了开成了年费，好像是240刀，记不清了 就这样 剩下的，无可奉告！ 实时rpm 1500 用户数39000 280 个帖子 - 260 位参与者 阅读完整话题
马斯克不愧为人类之光
前言：隔壁群友说我帖子攻击性太高，建议我放张涩图打头阵，怎么说呢…… 早上上班摸鱼算了两大页纸，但是手机上编辑公式实在太麻烦了，索性直接写结论。如果对数据有疑问到评论区再讨论。 马斯克说他计划在太空建立数据中心，那么我们很容易进行以下计算： 使用热泵并不能显著改善解热能力，热泵会产生额外的视在功率，虽然能够稍微增加热辐射，但效率提升实在有限。 所以为什么还有那么多人对此深信不疑，他们是不是觉得在太空3K的背景温度相当于把整机泡在液氮里？他们大概也计算到了数据中心巨大的损耗能够消化星舰（在PPT上展示的）运载能力，能够与人类历史上最大的太空运载工具（和地球资源浪费）完成闭环？ 伟大的马斯克用美股
Fox to Buy Roku Streaming Service in $22B Deal
15 points · 5 comments
Salesforce to Acquire Fin (formerly Intercom) for $3.6BN
63 points · 37 comments
Anthropic's Safety Superpower
123 points · 96 comments
Show HN: I wrote a C++ ray tracer from scratch without AI
42 points · 13 comments
What the Fuck Happened to Nerds
429 points · 266 comments
Openrouter Fusion API
92 points · 30 comments
Dalus (YC W25) Is Hiring a Senior Software Engineer in Germany
1 points · 0 comments
Successful Psilocybin Treatment of Alzheimer
33 points · 11 comments
Curl will not accept vulnerability reports during July 2026
545 points · 215 comments
Apple Foundation Models
261 points · 100 comments
Even more batteries included with Emacs
262 points · 61 comments
21 years and counting of 'eight fallacies of distributed computing' (2025)
108 points · 28 comments
Your ePub Is fine
727 points · 238 comments
Windows 11 users are tired of MS account requirements creeping into everything
442 points · 299 comments
Show HN: Kage – Shadow any website to a single binary for offline viewing
600 points · 119 comments
Ask HN: What are you working on? (June 2026)
What are you working on? Any new ideas that you're thinking about?
Rio de Janeiro's "homegrown" LLM appears to be a merge of an existing model
371 points · 195 comments
Formal methods and the future of programming
300 points · 97 comments
Show HN: Trace – Offline Mac meeting transcripts you can flag mid-call
I'm the developer of Trace, a non-intrusive, shortcut-driven Mac app that records and transcribes your meetings on-device. I know, another meeting transcription app. Please bear with me though, I'm confident that this is at least a little novel.I primarily built Trace for myself. I'd
Why does paper fold so well?
69 points · 33 comments
Being an old school web-based sports sim dev in the era of vibe coded games
44 points · 28 comments
Ported my C game to WASM, here's everybug that I hit
43 points · 30 comments
Exploring building a tiny FUSE filesystem
35 points · 6 comments
Asciline – real-time ASCII video rendering engine
19 points · 8 comments
Foreign business owners are scrambling to raise capital to stay in Japan
82 points · 54 comments
Bitsy
231 points · 6 comments
There Is(Ǝ) – Such That (∋)
63 points · 20 comments
Teenagers Stayed Overnight at Their School and Found Hidden Ancient Roman Ruins
23 points · 0 comments
A short history of Cerro Torre, the most controversial mountain (2012)
55 points · 31 comments
Firewood Splitting Simulator
884 points · 255 comments
今日主题
今日AI领域呈现出高速发展与深刻反思并存的复杂局面。OpenAI正通过启动「合作伙伴网络」并进行战略收购（如Ona），加速将企业级AI解决方案推向市场，旨在将AI深度融入商业运作，并推出实用AI课程赋能下一代劳动力。同时，AI Agent技术展现出惊人潜力，例如Claude Fable 5仅凭视觉就击败《宝可梦 火红》，以及Agent-Reach赋予AI Agent全面互联网访问能力。然而，伴随能力增长，对AI安全性和可控性的担忧也日益加剧。Anthropic旗下高潜力项目Fable因「过于危险」而停止发布，Google DeepMind也对数百万AI代理交互可能带来的不可预测风险表达忧虑，呼吁加强研究。此外，NVIDIA发布AI Agent安全扫描工具SkillSpector，AllenAI推出模型评估工作台olmo-eval，共同构筑更稳健的AI开发生态。整体来看，今天的AI世界正一边大步向前探索应用边界，一边谨慎地审视其潜在影响，力求在创新与安全之间取得平衡。
01
模型发布/更新
Model Releases 33 篇Agent-Reach开源工具赋能AI Agent全网实时信息获取
开源项目GitHub Trending
Agent-Reach是一款开源命令行工具，赋予AI Agent直接访问整个互联网的能力，使其能够实时阅读和搜索Twitter、Reddit、YouTube等平台内容。该项目通过避免高昂的API费用，解决了传统AI Agent在实时信息获取方面的痛点，尤其适用于希望构建具备实时互联网感知能力的AI Agent开发者，以提升其在复杂任务中的决策和信息处理效率。
Kronos：专注于金融市场语言的开源基础模型发布
开源项目GitHub Trending
Kronos是一个专为金融市场设计的基础语言模型，旨在深度理解、分析和生成与金融行业相关的文本信息，如财报、新闻及市场报告。该模型弥补了传统AI在金融领域专业知识的不足，为金融分析师、量化研究员和金融科技公司提供了强大的自然语言处理能力，支持更精准的市场预测和风险评估，推动金融领域的智能化转型。
Claude Fable 5仅凭视觉成功通关《宝可梦 火红》
官方Claude (YouTube)
Claude发布视频展示了其「Claude Fable 5」系统，在仅接收视觉输入的情况下，成功击败了经典游戏《宝可梦 火红》。这一演示突显了Claude Fable 5在高级视觉理解、复杂策略规划和游戏操作方面的卓越能力，预示着AI Agent在处理无结构化实时环境、执行长期任务方面取得显著进展，为其在更广泛领域实现自主决策和交互应用奠定了基础。
02
产品发布/更新
Product 33 篇Chatwoot开源客户支持平台提供多渠道服务台
开源项目GitHub Trending
Chatwoot是一款免费开源的客户支持平台，集成了在线聊天、邮件及多渠道服务台功能。它为企业提供了一个成本效益高且功能全面的客户沟通与管理解决方案，是Intercom和Zendesk等商业产品的优秀替代品。Chatwoot旨在帮助中小企业和初创公司统一管理客户咨询，显著提升客户服务效率和体验。
英伟达推出SkillSpector工具确保AI Agent技能安全
开源项目GitHub Trending
NVIDIA发布了SkillSpector，一款专为AI Agent技能设计的安全扫描工具。它能够检测AI Agent技能中的潜在漏洞、恶意模式及安全风险，确保其运行的可靠性与安全性。这款工具旨在解决AI应用在部署前进行安全审计的难题，主要面向AI Agent开发者和企业安全团队，以降低AI系统在实际应用中可能面临的威胁。
AllenAI发布olmo-eval模型评估工作台提升开发效率
官方Hugging Face Blog
AllenAI推出名为「olmo-eval」的评估工作台，旨在为模型开发循环提供全面的评估工具。该平台将帮助开发者更有效地测试、分析和改进其AI模型，从而提升开发效率和模型性能。olmo-eval的发布，为AI模型从概念验证到实际部署的全生命周期管理提供了标准化、系统化的评估解决方案，有助于研究者和工程师更快地迭代优化模型。
03
行业动态
Industry 44 篇OpenAI推出合作伙伴网络加速企业AI应用落地
官方OpenAI News
OpenAI正式启动「合作伙伴网络」，并计划投资1.5亿美元，旨在支持全球合作伙伴加速企业级AI解决方案的采纳、部署和转型。此举是OpenAI推动AI技术在商业领域深化应用的关键战略，通过构建强大的生态系统，帮助各行各业的客户更高效地利用AI赋能业务创新和增长，从而扩大其在企业AI市场的领导力。
OpenAI宣布计划收购Ona公司以扩展Codex能力
官方OpenAI News
OpenAI正式宣布计划收购Ona公司，此举旨在通过安全、持久的云环境扩展其Codex能力。此次收购将赋能AI代理在企业工作流中实现长期运行和更复杂的任务自动化，进一步提升AI在商业应用中的效率和稳定性。Ona的技术预计将为OpenAI的代理产品提供更强大的底层支持，加速其在企业级解决方案中的部署和落地。
Anthropic旗下AI项目Fable因安全顾虑停止发布
大咖博客Latent Space
据Latent Space报道，Anthropic旗下的高潜力AI项目Fable和Mythos因被认为「过于危险」而正式停止发布。此消息反映了业界对高级AI技术潜在风险的深刻担忧，特别是模型在安全性和可控性方面面临的严峻挑战。这一事件也促使人们再次审视AI技术发展与伦理安全之间的平衡，并可能对未来AI模型的发布策略产生影响。
Google DeepMind关注数百万AI代理交互的潜在风险
综合资讯MIT Tech Review AI
Google DeepMind正资助一项研究，深入探究数百万AI代理在网络上相互作用时可能产生的潜在危险。公司AGI安全与对齐研究负责人Rohin Shah表示，随着AI代理大规模普及，其互动可能带来不可预测的风险，引发业界对AI安全问题的普遍担忧。此举强调了在推动AI技术发展的同时，必须同步加强对其复杂性和潜在社会影响的审慎评估和管理。
04
技巧与观点
Tips & Takes 33 篇《Hello 算法》：动画图解数据结构与算法学习资源
开源项目GitHub Trending
《Hello 算法》是一本以动画图解方式呈现的数据结构与算法教程，支持多语言和多种编程语言的代码实现。该项目通过可视化和“一键运行”的代码示例，极大降低了学习门槛，旨在帮助初学者和准备面试的工程师轻松理解并掌握核心算法概念，从而有效提升编程基础和解决问题的能力。
AI工程学习指南：从零构建AI解决方案的实践之路
开源项目GitHub Trending
「AI Engineering From Scratch」是一份从零开始的AI工程学习指南，专注于AI项目从概念、构建到最终交付的整个生命周期。该指南强调实践和产品化能力，为希望系统学习AI解决方案开发与部署的工程师、开发者和技术爱好者提供了一套全面的学习路径，助力他们掌握AI技术在真实世界中的应用与落地。
OpenAI学院发布新课程赋能AI工作技能转型
官方OpenAI News
OpenAI学院推出了三门新课程，旨在帮助个人掌握实用的AI技能，创建可重复的AI工作流程，并将AI代理应用于日常工作中。这些课程面向下一代工作场景，以加速AI技术的普及和应用，确保劳动力能够适应并驾驭AI时代带来的变革，提升个人和组织在日益智能化的工作环境中的竞争力。