Claude Blog · 06/24 08:00
Claude Tag平台引入「智能体身份」功能，为自主AI智能体在团队范围内的应用提供新的访问模型。此举旨在加强AI在协作环境中的管理、安全性和责任归属，推动AI智能体在企业级应用中的普及。
推荐理由：Claude在AI智能体管理和协作方面的创新，对企业级AI应用的安全与合规性至关重要，值得关注。
2026-06-25（内容获取于 06/25 16:12）
LinuxDo · 06/25 14:07
一个基于概率分布识别任意模型真假的项目已开源，旨在解决AI模型可能存在的数据掺假或伪造问题，提升模型信任度和数据纯净度。该项目提供了一种新的验证方法，对维护AI生态的健康发展具有重要意义。
推荐理由：该开源项目提供了一种新颖且实用的AI模型真伪识别方案，对保障AI安全和数据可信度具有直接价值。
Product Hunt · 06/25 13:23
Figma在其设计画布中新增了「Figma Motion」时间轴功能，使用户能够直接在Figma内创建和编辑动画效果。此举旨在简化设计工作流，无需切换第三方工具，提高设计师工作效率和体验。
推荐理由：Figma作为领先设计工具，集成动画功能将极大提升设计师工作效率，值得产品和UI/UX从业者关注。
DeepMind Blog · 06/25 00:30
Google DeepMind宣布为Gemini 3.5 Flash模型新增「计算机使用」功能，使其能够与计算机系统进行交互和操作。这一能力扩展将使模型能执行更复杂的任务，极大地拓宽了AI应用的范围和潜力。（多家报道）
推荐理由：Gemini模型能力的显著提升，为开发者带来了更广阔的AI应用场景，是AI交互模式的重要进展。
GitHub Trending
OpenMontage是全球首个开源的智能体驱动视频制作系统，集成了12个流程、52种工具和500多种智能体技能。它能将AI编码助手转化为全功能视频制作工作室，自动化从脚本到成片的创作，大幅提升效率。
推荐理由：该项目是AI智能体在视频创作领域的重大突破，为内容创作者和开发者提供了强大的开源自动化工具。
Hugging Face Blog · 06/25 00:00
Hugging Face发文介绍如何利用NVIDIA NeMo AutoModel加速Transformer模型的微调过程。该技术旨在优化训练效率，帮助开发者快速定制AI模型，提升Transformer架构在各种应用中的性能和灵活性。
推荐理由：为ML工程师提供了Transformer模型微调的实用高效方案，直接提升开发效率和模型性能。
Latent Space · 06/25 10:14
Latent Space的AINews指出，当前AI领域正进入一个「元工具链」时代，预示着超越传统工具链工程的更高层级架构的兴起。这强调了AI智能体或系统在高级别组织和控制范式上的新发展。
推荐理由：对AI系统架构的未来发展趋势提供了前瞻性思考，对AI架构师和研究者有重要的战略启发。
MIT Tech Review AI · 06/24 19:59
麻省理工科技评论强调，随着AI发展，企业对大规模高质量数据需求激增，构建一个支持AI应用、有效获取和处理网络数据的「网络数据基础设施层」变得至关重要。此举旨在解决数据获取障碍，释放AI潜力。
推荐理由：深入分析AI发展面临的数据挑战及解决方案，为企业制定AI数据战略提供重要参考。
Riley Brown (YouTube) · 06/25 06:39
一则视频演示了如何在企业协作平台Slack中部署和利用强大的AI智能体。通过自动化日常任务，该方案旨在提升团队协作效率，实现更智能的工作流管理，为企业AI应用提供直观参考。
推荐理由：提供AI智能体在实际企业协作场景中的应用案例，对寻求提升团队效率的企业用户具启发价值。
GitHub Trending
daily_stock_analysis是一个由LLM驱动的开源多市场股票智能分析系统。它整合多源行情和实时新闻，提供决策看板并支持自动推送，帮助投资者获取AI赋能的金融洞察和自动化分析能力，支持零成本运行。
推荐理由：该开源项目利用LLM赋能金融分析，为个人和专业投资者提供了高效、低成本的市场情报辅助决策工具。
calesthio/OpenMontage
Python · ★ 20,650 · 🍴 2,318 · 📈 3,719 stars today
World's first open-source, agentic video production system. 12 pipelines, 52 tools, 500+ agent skills. Turn your AI coding assistant into a full video production studio.
中文介绍 OpenMontage 是一个开源的智能体驱动视频制作系统，通过整合12个流程、52种工具和500多种智能体技能，将AI编码助手升级为全功能的视频制作工作室。它旨在自动化视频内容创作，从脚本到最终成片，大幅提升效率。适用于内容创作者、营销人员及开发者，实现快速、批量化的视频生产。
ZhuLinsen/daily_stock_analysis
Python · ★ 49,079 · 🍴 43,225 · 📈 1,468 stars today
LLM 驱动的多市场股票智能分析系统：多源行情、实时新闻、决策看板与自动推送，支持零成本定时运行。 LLM-powered multi-market stock analysis system with multi-source market data, real-time news, decision dashboard, automated notifications, and cost-free scheduled runs.
中文介绍 daily_stock_analysis 是一个由 LLM 驱动的多市场股票智能分析系统。它整合了多源行情数据和实时新闻，提供直观的决策看板，并支持自动推送关键信息。该项目旨在为投资者提供 AI 赋能的金融洞察和自动化分析能力，且支持零成本定时运行，极大地提升了个人和专业投资者获取市场情报与辅助决策的效率。
apple/container
Swift · ★ 42,655 · 🍴 1,254 · 📈 1,838 stars today
A tool for creating and running Linux containers using lightweight virtual machines on a Mac. It is written in Swift, and optimized for Apple silicon.
中文介绍 `apple/container` 是一个专为 Mac 用户设计的工具，旨在通过轻量级虚拟机创建并运行 Linux 容器。该项目采用 Swift 语言编写，并针对 Apple silicon 架构进行了深度优化，提供高效的性能。它解决了在 macOS 环境中无缝运行 Linux 容器的需求，为开发者提供了更原生的容器化体验，减少了传统虚拟机的资源开销。
interviewstreet/hiring-agent
Python · ★ 2,499 · 🍴 620 · 📈 203 stars today
AI agent to evaluate and score resumes.
中文介绍 `hiring-agent` 是一个专门用于评估和评分简历的 AI 智能体。它旨在自动化招聘流程中繁琐的简历筛选环节，通过人工智能技术对海量简历进行高效分析，并根据预设标准进行打分。这极大地提升了招聘效率，帮助招聘人员快速识别最符合职位要求的候选人，从而优化人才筛选过程。
JCodesMore/ai-website-cloner-template
TypeScript · ★ 19,709 · 🍴 2,920 · 📈 692 stars today
Clone any website with one command using AI coding agents
中文介绍 `ai-website-cloner-template` 提供了一个利用 AI 编码智能体一键克隆任何网站的解决方案。该项目旨在大幅简化网站复制过程，通过智能体自动化解析网站结构、内容和样式，并重新生成。它特别适用于需要快速构建原型、测试不同设计或备份网站的开发者、设计师和内容创作者，提升了工作效率。
revfactory/harness
HTML · ★ 7,869 · 🍴 1,066 · 📈 277 stars today
A meta-skill that designs domain-specific agent teams, defines specialized agents, and generates the skills they use.
中文介绍 `harness` 是一个强大的元技能（meta-skill），专注于设计领域特定的智能体团队，定义专业化智能体，并自动生成它们所需的技能。它解决了在构建复杂多智能体系统时，手动配置和技能开发耗时的问题。通过自动化智能体及其能力的设计过程，该项目加速了 AI 应用的开发与部署，适用于 AI 开发者和研究人员构建高度定制化的智能体解决方案。
flutter/flutter
Dart · ★ 177,501 · 🍴 30,557 · 📈 73 stars today
Flutter makes it easy and fast to build beautiful apps for mobile and beyond
中文介绍 Flutter 是由 Google 开发的开源 UI 软件开发工具包，旨在让开发者轻松快速地为移动、Web、桌面及嵌入式设备构建精美、高性能的跨平台应用。它采用 Dart 语言，以一套代码库实现多平台部署，显著提升了开发效率和用户体验。Flutter 凭借其丰富的组件库和热重载功能，成为现代应用开发的热门选择。
andreknieriem/headunit-revived
Kotlin · ★ 1,485 · 🍴 115 · 📈 41 stars today
Headunit App for displaying Android Auto
中文介绍 `headunit-revived` 是一款专为显示 Android Auto 内容而设计的 Headunit 应用。它解决了非原生支持 Android Auto 的设备（如旧平板电脑或定制车载显示器）无法直接使用该功能的问题。通过将任何 Android 设备转变为 Android Auto 的显示终端，它为用户提供了更广泛的选择，尤其适合汽车爱好者和希望在现有设备上集成现代车载系统的用户。
stablyai/orca
TypeScript · ★ 7,066 · 🍴 509 · 📈 331 stars today
Orca is the ADE for working with a fleet of parallel agents. Run any coding agent with your own subscription. Available on desktop and mobile.
中文介绍 Orca 是一个专为处理并行智能体集群设计的智能体开发环境（ADE）。它解决了开发者在管理和部署大量 AI 编码智能体时的复杂性，允许用户通过自己的订阅运行任何编码智能体。该平台支持桌面和移动设备，提供了一个统一且可扩展的解决方案，适用于构建、测试和运行复杂的并发智能体系统，极大提升了多智能体项目的开发与运维效率。
google-labs-code/design.md
TypeScript · ★ 17,751 · 🍴 1,584 · 📈 619 stars today
A format specification for describing a visual identity to coding agents. DESIGN.md gives agents a persistent, structured understanding of a design system.
中文介绍 `DESIGN.md` 是 Google Labs Code 推出的一项格式规范，旨在为 AI 编码智能体提供描述视觉识别和设计系统的结构化方式。它解决了智能体在理解和应用设计规范时遇到的挑战，通过提供持久且一致的设计系统理解，确保了代码与设计稿之间的高度一致性。这极大地简化了设计到代码的转换过程，赋能 AI 智能体更精准地生成符合品牌规范的界面。
Flowseal/zapret-discord-youtube
Batchfile · ★ 30,061 · 🍴 2,344 · 📈 61 stars today
中文介绍 `zapret-discord-youtube` 项目名称暗示其可能与 Discord 和 YouTube 平台上的内容限制或管理相关，其中“zapret”在俄语中意为“禁止”。它可能旨在帮助用户绕过某些内容访问限制，或提供一个工具来监控、管理或报告这两个平台上的特定事件或禁令。具体功能和技术细节由于描述缺失，尚不明确，但可能涉及两平台 API 交互。
kunchenguid/no-mistakes
Go · ★ 2,543 · 🍴 161 · 📈 110 stars today
git push no-mistakes
中文介绍 `no-mistakes` 项目通过其简短的描述 `git push no-mistakes`，暗示其旨在在执行 `git push` 操作前，帮助开发者避免潜在的错误或代码质量问题。它可能是一个 Git 钩子（hook）、一个命令行工具或一套自动化检查机制，用于执行代码规范检查、单元测试或其他预提交验证。该工具能够有效提升团队协作中的代码质量，减少错误进入主分支的风险。
NousResearch/hermes-agent
Python · ★ 202,456 · 🍴 36,180 · 📈 1,178 stars today
The agent that grows with you
中文介绍 `hermes-agent` 是 NousResearch 推出的一个 AI 智能体项目，其核心理念是“与你一同成长”。这意味着该智能体具备持续学习和适应用户需求的能力，能够根据用户的交互和反馈进行个性化调整和优化。它旨在提供一个高度定制化且不断进化的 AI 助手，适用于需要长期协作、任务复杂且要求个性化体验的场景，为用户提供更智能、更贴心的服务。
MVTrack4Gen: Multi-View Point Tracking as Geometric Supervision for 4D Video Generation
👍 11
Synthesizing a novel-view video from a monocular reference video along a target camera trajectory requires both geometric consistency and motion fidelity with respect to the reference video. Existing methods based on explicit 3D representations are limited by the accuracy of off-the-shelf reconstruc
ShutterMuse: Capture-Time Photography Guidance with MLLMs
👍 31
Real-world photography requires capture-time guidance for both camera framing and subject pose. Yet existing aesthetic cropping benchmarks mainly evaluate post-hoc crop prediction and overlook subject-side recommendations, leaving the capture-time guidance capabilities of multimodal large language m
V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning
👍 7
Fine-grained visual reasoning requires multimodal large language models (MLLMs) to identify task-relevant visual evidence and ground their reasoning in local image regions. Existing agentic methods typically rely on reinforcement learning with verifiable rewards or supervised fine-tuning on large-sc
TryOnCrafter: Unleashing Camera Trajectories for Realistic Video Virtual Try-on via a Renderable 4D Try-on Proxy
👍 5
While Video Virtual Try-on (VVT) has achieved remarkable progress in synthesizing realistic garment overlays on dynamic subjects, existing paradigms remains fundamentally constrained by a passive dependency on source camera trajectories, failing to accommodate the requisite interactive freedom for o
Autodata: An agentic data scientist to create high quality synthetic data
👍 6
We introduce Autodata, a general method that enables AI agents to act as data scientists who build high quality training and evaluation data. We show how to train (meta-optimize) such a data scientist agent, so that it learns to create even stronger data. We describe the overall formulation, and a s
CAVEWOMAN: How Large Language Models Behave Under Linguistic Input and Output Compression
👍 3
"Talk short. Drop grammar. Save token." This caveman style is widely promoted as a way to cut inference cost, but whether it actually saves anything depends on which channel (the user's prompt or the model's response) is being compressed. We present Cavewoman, a two-channel evaluation protocol that
Are We Ready For An Agent-Native Memory System?
👍 51
Memory for large language model (LLM) agents has rapidly evolved from simple retrieval-augmented mechanisms into a data management system that supports persistent information storage, retrieval, update, consolidation, and dynamic lifecycle governance throughout agent execution. Despite this evolutio
RoPE-Aware Bit Allocation for KV-Cache Quantization
👍 3
Existing low-bit KV-cache quantizers often treat each cached key as a flat vector. Under RoPE, however, a key's contribution to a future attention logit decomposes into a position-dependent sum over two-dimensional frequency blocks. This makes key-cache quantization a block-wise bit-allocation probl
IV-CoT: Implicit Visual Chain-of-Thought for Structure-Aware Text-to-Image Generation
👍 11
Unified multi-modal large language models (MLLMs) have achieved strong text-to-image generation quality, but still struggle with structure-aware prompt following, where object counts, spatial relations, attribute bindings, and coarse layouts must be preserved. We attribute this limitation in part to
Advancing WordArt-Oriented Scene Text Recognition: Datasets and Methods
👍 6
WordArt (artistic text) features highly customized fonts, textures, and layouts, making WordArt-oriented scene TExt Recognition (WATER) substantially more challenging than general Scene Text Recognition (STR). Existing STR datasets and methods, typically built around regular scene text and fixed-tem
Wan-Streamer v0.1: End-to-end Real-time Interactive Foundation Models
👍 30
We present Wan-Streamer, a native-streaming, end-to-end interactive foundation model designed from the ground up for real-time, low-latency, full-duplex audio-visual interaction. Wan-Streamer seamlessly models language, audio, and video as both input and output within a single Transformer, where the
MEMPROBE: Probing Long-Term Agent Memory via Hidden User-State Recovery
👍 1
Long-term memory promises LLM agents that grow more capable across sessions, maintaining an accurate, evolving understanding of the user that interaction forms. In practice, however, this memory is evaluated mostly through downstream behavior, such as later answers, personalization quality, or task
AGORA: An Archive-Grounded Benchmark for Agentic Workplace Document Reasoning
👍 0
Large language models are increasingly deployed as agents that reason over documents rather than answer from parametric knowledge. We study archive-grounded reasoning: locating sparse evidence across a large, messy collection of workplace files, reconciling inconsistent terminology, units, and time
FLAT: Feedforward Latent Triangle Splatting for Geometrically Accurate Scene Generation
👍 14
Generating explorable 3D scenes from a single image requires strong generative priors and accurate geometric representations suitable for downstream use. Current video diffusion models offer high-quality generation and implicitly encode multi-view geometric structure in latent space. However, existi
Escaping the Self-Confirmation Trap: An Execute-Distill-Verify Paradigm for Agentic Experience Learning
👍 8
Experience-driven self-evolution is critical for large language model (LLM) agents to improve through open-world interaction. However, existing experience learning methods mostly rely on single-agent loops, where the same agent executes tasks, summarizes outcomes, and determines memory content. This
The Hitchhiker's Guide to Agentic AI: From Foundations to Systems
👍 6
The Hitchhiker's Guide to Agentic AI is a comprehensive practitioner's reference for building autonomous AI systems. The book covers the full stack from first principles to production deployment, organized around a central thesis: building great agentic systems requires understanding every layer of
VeriEvol: Scaling Multimodal Mathematical Reasoning via Verifiable Evol-Instruct
👍 4
Scaling reinforcement learning for visual mathematical reasoning requires more than generating harder questions: as data volume grows, the reward labels themselves must remain reliable. Yet existing data pipelines scale supervision while trusting the labeller, and policy-side methods assume the unde
Critique of Agent Model
👍 15
What is an agent? What constitutes agency? With the rise of Large Language Model (LLM) systems marketed as ``coding agents'', ``AI co-scientists'', and other ``agentic" tools that promise to drive up productivity, and at the same time, ``existential" concerns such as AI escaping human control with d
Semantic Browsing: Controllable Diversity for Image Generation
👍 14
Modern text-to-image models excel in visual fidelity and prompt adherence. However, this strict adherence comes at the cost of diversity: generated samples tend to collapse into a single visual interpretation. Existing methods to improve diversity produce outputs driven by incidental variations rath
ChartWalker: Benchmarking the Cross-Chart RAG Task
👍 2
Cross-Chart Retrieval-Augmented Generation (RAG) is critical for complex multi-modal analytical tasks in scientific, business, and political domains. However, existing benchmarks either focus on tables, which are well-structured and textualized, or generate cross-chart questions by simply extracting
AOHP: An Open-Source OS-Level Agent Harness for Personalized, Efficient and Secure Interaction
👍 26
AI agents are driving a new software paradigm, with the ability to autonomously call tools, extract information, manage memory, and complete tasks that span applications and data sources. Most existing end-user operating systems, however, are designed for application-centric workflows and offer litt
Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do
👍 6
Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks remains unclear. In this paper, we aim to systematically investigate the key question: What can multi
Multi4D: High-Fidelity Dynamic Gaussian Splatting via Multi-Level Competitive Allocation
👍 0
Dynamic 3D Gaussian splatting faces a fundamental tension between motion consistency and visual fidelity. Deformation-based approaches preserve temporal correspondence but suffer from motion over-factorization, oversmoothing high-frequency dynamics. In contrast, 4D-primitive methods capture fine vis
UnityShots: Memory-Driven Multi-Shot Audio-Video Generation with Boundary-Aware Gating
👍 7
Generating a coherent multi-shot video requires structured cross-shot memory. Subject appearance, scene context, and speaker identity must persist across cuts. Existing approaches either train end-to-end over fixed-length sequences and cannot scale, generate shot-by-shot with memory banks that grow
When Lower Privileges Suffice: Investigating Over-Privileged Tool Selection in LLM Agents
👍 3
As LLM agents increasingly select tools autonomously, their choices among tools with different privileges become safety-relevant. However, prior tool-selection studies focus on safety-agnostic metadata preferences, leaving privilege-sensitive choices underexplored. To address this gap, we study over
QG-MIL: A Gated Transformer Aggregator for Domain-Agnostic Multiple Instance Learning in Medical Imaging
👍 2
Attention-based Multiple Instance Learning aggregators in medical imaging are prone to attention concentration, producing overconfident and unstable predictions. We introduce QG-MIL, a gated transformer aggregator that addresses this through four synergistic architectural components: RMSNorm-based p
EventVLA: Event-Driven Visual Evidence Memory for Long-Horizon Vision-Language-Action Policies
👍 2
Memory remains a critical bottleneck for long-horizon robotic manipulation, as standard Vision-Language-Action (VLA) policies often fail when task-relevant cues become occluded or unobservable over time. While existing memory-augmented methods utilize historical context, they either suffer from seve
Beyond NL2Code: A Structured Survey of Multimodal Code Intelligence
👍 23
While Large Language Models (LLMs) have substantially advanced text-to-code synthesis, many real programming tasks specify intent through visual artifacts such as screenshots, charts, vector drawings, videos, and interactive states. These tasks require models to connect visual perception to executab
RL-Index: Reinforcement Learning for Retrieval Index Reasoning
👍 4
Retrieving external knowledge is essential for solving real-world tasks, yet it remains challenging when the relationship between a query and its relevant knowledge involves implicit and complex reasoning beyond surface-level semantic or lexical matching (e.g., mathematical problems relying on the s
LingxiDiagBench: A Multi-Agent Framework for Benchmarking LLMs in Chinese Psychiatric Consultation and Diagnosis
👍 21
Mental disorders are highly prevalent worldwide, but the shortage of psychiatrists and the inherent subjectivity of interview-based diagnosis create substantial barriers to timely and consistent mental-health assessment. Progress in AI-assisted psychiatric diagnosis is constrained by the absence of
Figma Motion
Your Figma canvas now has a timeline
中文介绍 Figma推出「Figma Motion」，为用户在其设计画布中新增了时间轴功能。这意味着设计师现在可以直接在Figma界面内创建和编辑动画效果，无需依赖第三方工具，从而提升设计工作流程的效率和一体化体验，将动态效果设计无缝融入Figma生态系统。
Paybond CLI
Safe agent spend from the terminal
中文介绍 Paybond CLI 是一款命令行界面工具，旨在通过终端实现代理（Agent）的安全支出管理。它为用户提供了一种在自动化操作中控制和保障资金流动的机制，确保代理在执行任务时，其消费行为是可控且安全的，有效预防未经授权的支出或浪费。
QuickMaker
State of the art AI models in Blender under one subscription
中文介绍 QuickMaker 将前沿的AI模型引入流行的3D创作软件Blender中，用户通过一次订阅即可在Blender内使用这些先进的AI功能。这旨在简化3D设计和内容创作流程，让艺术家和设计师能够利用AI技术更高效地生成、编辑和优化他们的作品，提升创作效率。
Papermark Agents
Let AI agents run your next deal, fundraise or data room
中文介绍 Papermark Agents 引入了AI代理，旨在帮助用户管理商业交易、融资活动或数据室等关键业务流程。通过利用人工智能的自动化能力，该工具能简化复杂的操作，提高效率，协助用户更智能地处理从协商到数据管理等一系列商业活动，优化决策过程。
Sidegent
Learn to build AI agents by actually building them
中文介绍 Sidegent 提供了一个实践平台，旨在通过实际构建AI代理的方式，帮助用户学习和掌握AI代理的开发技能。它强调“边做边学”的方法，使用户能够通过具体的项目和操作，深入理解AI代理的设计、实现与部署，从而培养实际的AI开发能力和解决问题的经验。
Genspark Design
Generate UI prototypes, videos, and posters with AI
中文介绍 Genspark Design 是一款利用人工智能技术进行设计生成的工具。它能够帮助用户快速创建用户界面（UI）原型、制作视频以及设计海报。通过AI驱动的自动化能力，该产品旨在大幅提升设计师和内容创作者的工作效率，简化从概念到最终产出的全过程，实现快速迭代。
Zaro
Build agents & apps on top of your context with one prompt.
中文介绍 Zaro 提供一个平台，让用户可以通过单一提示（prompt），基于其特定的上下文信息，快速构建和部署AI代理及应用程序。该工具旨在简化开发流程，使得非专业开发者也能轻松地利用AI技术，根据自身需求定制智能化解决方案，提高开发效率和产品创新速度。
Signspell
Real-time ASL alphabet recognition in py ,pip install and go
中文介绍 Signspell 是一个提供实时美国手语（ASL）字母识别功能的Python库。用户可以通过简单的 `pip install` 命令进行安装和使用。该工具旨在为开发者提供一个便捷的接口，用于在Python项目中集成手语识别能力，促进无障碍交流技术的应用和发展，提升辅助沟通效率。
Grass 2.0
The always-on computer for your coding agents
中文介绍 Grass 2.0 是一款为编码代理（coding agents）设计的“常开式计算机”。它提供一个持续运行的环境，支持AI代理在其中执行编码任务。结合其“Claude code from your phone”的描述，该产品还可能支持用户通过手机与AI编码代理互动，实现远程或移动的代码生成与管理，提升开发效率和便捷性。
Heron
Wireshark for AI Agents: passive eBPF observability
中文介绍 Heron 被描述为“AI代理的Wireshark”，它通过被动的eBPF可观测性技术，为AI代理提供深入的监控和分析能力。该工具旨在帮助开发者实时洞察AI代理的运行状态、行为模式和内部通信，从而更好地理解、调试和优化复杂的AI系统，确保其稳定高效运行。
How Quants Use Loop Engineering to Build Alpha (Full Framework)
@horizon_trade_x · 4.4K 粉丝 · 1.3M 阅 · 507 赞 · 59 转
Your backtest looked flawless. You went live. Two weeks later, the strategy was bleeding. Every quant has lived this. The answer is a loop: generate a strategy, test it, score it, feed the result
Loops explained: Claude, GPT, Mira and what actually works
@AnatoliKopadze · 83.0K 粉丝 · 1.3M 阅 · 584 赞 · 70 转
AI has been in everyone's hands for years. Most people who use it every day still use it the slowest way there is: type a request, wait, fix it, ask again, all by hand. Not because the faster way is
How to Build an AI Second Brain With Claude and Obsidian That Gets Smarter Every Day (Full Guide)
@undefinedKi · 3.9K 粉丝 · 1.0M 阅 · 601 赞 · 78 转
Your best ideas are scattered across a dozen places right now. Notes apps. Browser tabs. Old chats with Claude that you closed and will never find again. Every time you sit down to work, you rebuild
40 Claude Opus 4.8 Workflows That Make Money While You Sleep
@eng_khairallah1 · 69.3K 粉丝 · 678.7K 阅 · 506 赞 · 81 转
Most people are using Claude Opus 4.8 to answer questions. Save this :) A small group of people have it running businesses while they sleep. The difference is not the model. You both have access to
The efficiency gap: How do cells and GPUs compare when running the exact same algorithm?
@spandan_madan · 1.1K 粉丝 · 626.9K 阅 · 516 赞 · 28 转
AI hardware is having a moment. Hyperscaler capex on AI data centres is on track to clear $690 billion in 2026, and private equity has followed in scale — Blackstone alone reports a $55B+ data-centre
How To Learn Anything 10x Faster Using Claude
@sairahul1 · 120.1K 粉丝 · 581.3K 阅 · 516 赞 · 110 转
Learning anything today is easy and confusing at the same time. Easy because AI can explain almost anything in seconds. Confusing because most people just ask random questions, get random answers, and
The Agent Loop Architecture
@djfarrelly · 3.8K 粉丝 · 344.7K 阅 · 501 赞 · 61 转
Everyone's asking "WTF is a loop?" Here's the question nobody's asking: what runs the loop? The AI discourse has converged on loops as a core primitive of agentic systems. Matt Van Horn (@mvanhorn)
Introducing Engram: Scaling compute on your context
@EngramLab · 1.2K 粉丝 · 255.7K 阅 · 537 赞 · 76 转
We’re Engram. We’re building AI that learns from you and deeply understands your work. Today’s AI models don’t understand what you do. Not really. Everything models know comes from their training –
30 Core Agentic Engineering Concepts Every Developer Should Know
@sairahul1 · 120.1K 粉丝 · 249.0K 阅 · 502 赞 · 90 转
There are 8 billion people on the planet. Only a fraction of developers understand how AI agents actually work. Not the demos. Not the hype. The real engineering underneath. Every week a new agent
ORACLE: Official AI Agents Trade on Polymarket
@Oracle_Trade_ai · 39.9K 粉丝 · 197.8K 阅 · 2.8K 赞 · 580 转
In 2026, autonomous AI agents have become one of the most effective strategies on prediction markets. Over 30% of all activity on Polymarket now comes from algorithmic and AI-powered wallets. We
Why we're bullish on loops
@posthog · 21.8K 粉丝 · 162.4K 阅 · 512 赞 · 36 转
When the creators of both OpenClaw and Claude Code speak, people listen. And recently Peter Steinberger and Boris Cherny have both been talking about the same concept: loops. Their argument? You
ORACLE: Official AI Agents Trade on Polymarket
@OracAItrading · 31.8K 粉丝 · 141.6K 阅 · 2.8K 赞 · 576 转
In 2026, autonomous AI agents have become one of the most effective strategies on prediction markets. Over 30% of all activity on Polymarket now comes from algorithmic and AI-powered wallets. We
Interactions API: Our primary interface for Gemini models and agents
@GoogleAIStudio · 179.4K 粉丝 · 138.2K 阅 · 504 赞 · 42 转
Today we're announcing that the Interactions API has reached general availability and is now our primary API for interacting with Gemini models and agents. We launched its public beta in December
ORACLE: Official AI Agents Trade on Polymarket
@Oractrading · 33.9K 粉丝 · 109.2K 阅 · 2.8K 赞 · 585 转
In 2026, autonomous AI agents have become one of the most effective strategies on prediction markets. Over 30% of all activity on Polymarket now comes from algorithmic and AI-powered wallets. We
How To Use Loop Engineering To Build A Self-Improving Quant Trading System
@RohOnChain · 51.4K 粉丝 · 108.3K 阅 · 501 赞 · 65 转
I will break down exactly how to build the loops that run an entire quant trading system on their own. Let's get straight to it. Bookmark This - I'm Roan, a backend developer working on system
how to get Fable-level intelligence back:
@EXM7777 · 118.9K 粉丝 · 107.7K 阅 · 509 赞 · 44 转
for a few days, we had something that felt like AGI... Fable 5 showed up, effectively unlimited inside the plans, and the ceiling on what you could build lifted overnight but then Anthropic killed it,
WTF Is a Loop? Part 2: The 15 Loops People Are Actually Running (and the Commands to Steal Them)
@mvanhorn · 35.2K 粉丝 · 102.4K 阅 · 510 赞 · 56 转
Earlier this month I wrote WTF Is a Loop? Peter Steinberger vs. Boris Cherny, which did 3.6M views on what a loop even is. This is the sequel, and it answers the next question: which loops do people
From Prompting Agents to Loop Engineering
@omarsar0 · 308.0K 粉丝 · 90.2K 阅 · 504 赞 · 69 转
A claim has been circulating in AI coding circles: stop prompting your coding agents and start designing loops that prompt them for you. As with everything new, this stuff gets repeated often and
How GLM-5.2 Beat Fable 5 at Website Design
@Designarena · 13.9K 粉丝 · 80.4K 阅 · 518 赞 · 39 转
GLM 5.2 ranks 1st overall on Design Arena’s single-turn, HTML Web Design (Non-Agentic) evaluation, 5 places higher than its predecessor GLM-5.1. To do so, it beat Claude Fable 5, Opus 4.6, and Opus
Subnets mine TAO; TAO mines Subnets
@const_reborn · 29.7K 粉丝 · 79.8K 阅 · 503 赞 · 116 转
e_i \;\propto\; \underbrace{\rho_i \times \bar{p}_i}_{\text{linear (maximize)}} \times \underbrace{(1 - b_i)}_{\text{boolean gate}} Disclaimer: this upgrade only effects subnet owners and dynamic TAO
Introducing computer use in Gemini 3.5 Flash
@GoogleAIStudio · 179.4K 粉丝 · 41.0K 阅 · 7d 曝光 179.2K
Introducing computer use in Gemini 3.5 Flash
Tesla has patented a digital siren that allows car sensors to break the network rules in a crisis
@tslaming · 34.9K 粉丝 · 65.2K 阅 · 7d 曝光 65.2K
Tesla has patented a digital siren that allows car sensors to break the network rules in a crisis
The efficiency gap: How do cells and GPUs compare when running the exact same algorithm?
@spandan_madan · 1.1K 粉丝 · 626.9K 阅 · 7d 曝光 626.9K
The efficiency gap: How do cells and GPUs compare when running the exact same algorithm?
Why we're bullish on loops
@posthog · 21.8K 粉丝 · 162.4K 阅 · 7d 曝光 162.4K
Why we're bullish on loops
Introducing Engram: Scaling compute on your context
@EngramLab · 1.2K 粉丝 · 255.7K 阅 · 7d 曝光 255.7K
Introducing Engram: Scaling compute on your context
Building a skill optimization loop
@zachlloydtweets · 11.2K 粉丝 · 39.2K 阅 · 7d 曝光 39.2K
Building a skill optimization loop
How does vLLM work?
@amitiitbhu · 23.6K 粉丝 · 35.7K 阅 · 7d 曝光 35.7K
How does vLLM work?
30 Core Agentic Engineering Concepts Every Developer Should Know
@sairahul1 · 120.1K 粉丝 · 249.0K 阅 · 7d 曝光 249.0K
30 Core Agentic Engineering Concepts Every Developer Should Know
构架师教程：Foundation Engineering
@dashen_wang · 24.0K 粉丝 · 189.6K 阅 · 7d 曝光 189.6K
构架师教程：Foundation Engineering
Subnets mine TAO; TAO mines Subnets
@const_reborn · 29.7K 粉丝 · 79.8K 阅 · 7d 曝光 79.8K
Subnets mine TAO; TAO mines Subnets
How we built the world’s fastest API for GLM-5.2
@philipkiely · 8.6K 粉丝 · 73.1K 阅 · 7d 曝光 73.1K
How we built the world’s fastest API for GLM-5.2
Loop Engineering Clearly Explained
@akshay_pachaar · 278.2K 粉丝 · 44.0K 阅 · 7d 曝光 44.0K
Loop Engineering Clearly Explained
Interactions API: Our primary interface for Gemini models and agents
@GoogleAIStudio · 179.4K 粉丝 · 138.2K 阅 · 7d 曝光 179.2K
Interactions API: Our primary interface for Gemini models and agents
Taking CS2 in-game stats a little bit more serious
@expsgg · 151 粉丝 · 79.2K 阅 · 7d 曝光 79.2K
Taking CS2 in-game stats a little bit more serious
How To Use Loop Engineering To Build A Self-Improving Quant Trading System
@RohOnChain · 51.4K 粉丝 · 108.3K 阅 · 7d 曝光 108.3K
How To Use Loop Engineering To Build A Self-Improving Quant Trading System
简直就是教科书级别的 AI 设计规范。
@xiaogaifun · 1.7K 粉丝 · 114.5K 阅 · 7d 曝光 114.5K
简直就是教科书级别的 AI 设计规范。
ORACLE: Official AI Agents Trade on Polymarket
@Oracle_Trade_ai · 39.9K 粉丝 · 197.8K 阅 · 7d 曝光 197.8K
ORACLE: Official AI Agents Trade on Polymarket
ORACLE: Official AI Agents Trade on Polymarket
@OracAItrading · 31.8K 粉丝 · 141.6K 阅 · 7d 曝光 141.6K
ORACLE: Official AI Agents Trade on Polymarket
ORACLE: Official AI Agents Trade on Polymarket
@Oractrading · 33.9K 粉丝 · 109.2K 阅 · 7d 曝光 109.2K
ORACLE: Official AI Agents Trade on Polymarket
用一台 6 美元 VPS，把外网安全拿回自己手里
@wangdefou · 14.3K 粉丝 · 86.1K 阅 · 7d 曝光 86.1K
用一台 6 美元 VPS，把外网安全拿回自己手里
Poweful Agents inside Slack
中文介绍 视频演示了如何在企业协作平台Slack中部署和利用强大的AI代理，以自动化日常任务、提升团队协作效率，并实现更智能的工作流管理。
AI Agents Just Changed Forever: GLM 5.2, Codex Skills, Claude & Cursor
中文介绍 Riley Brown的视频分析了AI代理技术近期发生的重大变革，重点关注了GLM 5.2、Codex Skills、Claude以及Cursor等领先工具和模型如何共同推动AI代理能力的显著提升。
9 AI Agent Skills To Get Ahead of 99% of People
How to build high quality internal tools with Retool
SpaceX Just Bought Cursor for $60B. It’s About to Take OVER.
Why the Government Just Killed Claude Fable 5
This is Why They Banned Claude Fable 5
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
Tag Claude in, right where you already work
中文介绍 Anthropic官方展示了如何将AI助手Claude无缝集成到用户日常工作环境中，通过简单的「@Claude」提及功能，用户即可在现有工作界面中便捷地调用其智能能力。
Delegate and schedule tasks in Claude Cowork
中文介绍 Claude官方视频演示了其「Claude Cowork」平台的新功能，允许用户通过AI助手高效地委派和调度各项任务，从而优化团队协作和项目管理流程。
Artifacts in Claude Code: share your work as it happens
中文介绍 Claude官方视频介绍了「Claude Code」中的「Artifacts」功能，该功能允许开发者团队实时分享其代码工作成果，促进即时协作、版本控制与反馈。
Enterprise-managed auth for MCP connectors
中文介绍 Claude官方视频展示了面向企业客户的MCP连接器如何实现企业级托管的身份验证，确保了数据安全和合规性，并简化了大型组织对用户访问权限的管理。
The Briefing: Financial Services
Claude FM 🎵 music for thinking and building
Claude FM 🎵 music for thinking and building
Code with Claude Tokyo 2026: Opening Keynote
The Problem Solvers | Michael Truell at Cursor
Claude Fable 5 beats Pokémon FireRed only using vision
Tag Claude in, right where you already work
中文介绍 Anthropic官方展示了如何将AI助手Claude无缝集成到用户日常工作环境中，通过简单的「@Claude」提及功能，用户即可在现有工作界面中便捷地调用其智能能力。
Delegate and schedule tasks in Claude Cowork
中文介绍 Claude官方视频演示了其「Claude Cowork」平台的新功能，允许用户通过AI助手高效地委派和调度各项任务，从而优化团队协作和项目管理流程。
Artifacts in Claude Code: share your work as it happens
中文介绍 Claude官方视频介绍了「Claude Code」中的「Artifacts」功能，该功能允许开发者团队实时分享其代码工作成果，促进即时协作、版本控制与反馈。
Enterprise-managed auth for MCP connectors
中文介绍 Claude官方视频展示了面向企业客户的MCP连接器如何实现企业级托管的身份验证，确保了数据安全和合规性，并简化了大型组织对用户访问权限的管理。
The Briefing: Financial Services
Claude FM 🎵 music for thinking and building
Claude FM 🎵 music for thinking and building
Code with Claude Tokyo 2026: Opening Keynote
The Problem Solvers | Michael Truell at Cursor
Claude Fable 5 beats Pokémon FireRed only using vision
DeepSeek Just Solved AI's Billion Dollar Problem
中文介绍 Two Minute Papers报道称，DeepSeek公司在AI领域取得了一项突破性进展，成功解决了AI研究中一个长期存在的「数十亿美元」的重大难题，其具体影响和方案备受关注。
Scientists Found A Better Language For AI Agents
中文介绍 Two Minute Papers报道，科学家们研究发现了一种更优化的「语言」或编程范式，旨在显著提升AI代理的理解、推理和执行能力，有望推动AI代理技术的发展。
They Looked Inside Claude’s AI's Mind. It Got Weird
NVIDIA's New Free AI - A Gift To Humanity
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
[AINews] It's Meta-Harness Summer
Move over, Harness Engineering, it is time for the harness of harnesses!
中文介绍 Latent Space的AINews指出，当前AI领域正进入一个「元工具链」时代，预示着超越传统工具链工程的更高层级架构的兴起。这可能涉及更复杂或更集成的AI系统协调与管理框架，强调了AI智能体或系统的更高级别组织和控制范式。
Why the Frontier Ecosystem must be Open — Matei Zaharia and Reynold Xin, Databricks
In a rare double-interview, the Databricks technical leaders riff on what it will take for every company to build Agent Clouds
中文介绍 Databricks技术领导Matei Zaharia和Reynold Xin在一次访谈中强调了开放的前沿AI生态系统对企业的重要性。他们深入探讨了企业如何构建和利用「Agent Clouds」（智能体云），以实现更广泛的AI应用和创新，并指出开放性是推动AI普及的关键。
Introducing computer use in Gemini 3.5 Flash
中文介绍 Google DeepMind在其博客中宣布，已为Gemini 3.5 Flash模型引入了「计算机使用」功能。这意味着该模型能够与计算机系统进行交互和操作，从而扩展其应用范围和能力，使其能够执行更复杂的任务。
Accelerating Transformers Fine-Tuning with NVIDIA NeMo AutoModel
中文介绍 Hugging Face博客发布文章，介绍如何利用NVIDIA NeMo AutoModel加速Transformer模型的微调过程。该技术旨在优化训练效率，帮助开发者更快速、高效地完成AI模型的定制化，显著提升Transformer架构在各种应用中的性能和灵活性。
The emergence of the web data infrastructure layer for AI
AI is booming. New use cases are emerging each day. To capitalize on the technology’s potential, enterprises require data at scale. In many cases, though, the relevant information is blocked or unstructured, which limits its use by AI models. To understand this challenge, consider the foundation of
中文介绍 麻省理工科技评论指出，随着AI的蓬勃发展，企业对大规模高质量数据需求日益增长。然而，许多相关网络数据受阻或非结构化，限制了AI模型的使用。文章强调，为应对这一挑战，构建一个支持AI应用、能够有效获取和处理网络数据的「网络数据基础设施层」正变得至关重要。
[AINews] Claude Tag: Multiplayer, Proactive, Persistent Agents in Slack
Claude finally gets a Slackbot upgrade
中文介绍 Latent Space的AINews报道，AI模型Claude通过「Claude Tag」在Slack平台获得了重要升级。新功能支持多用户协作、主动式交互以及持久化智能体，使得Claude能够更好地融入团队协作环境，提供更强大的自动化和智能助理服务。
OpenAI and Broadcom unveil LLM-optimized inference chip
OpenAI and Broadcom introduce Jalapeño, a custom AI chip built for LLM inference to improve performance, efficiency, and scale across AI systems.
中文介绍 OpenAI与博通公司（Broadcom）联合发布了一款名为「Jalapeño」的定制AI芯片。该芯片专为大型语言模型（LLM）推理任务进行优化，旨在显著提升AI系统的性能、能效和可扩展性，以满足不断增长的计算需求和推动AI基础设施创新。
not much happened today
**OpenAI** announced **Jalapeño**, its first custom AI chip for LLM inference, built with **Broadcom**, aiming to control more of the AI stack and improve compute economics with a fast 9-month design cycle. Community analysis suggests Jalapeño features **216GB HBM3E**, **~7.1–7.4 TB/s bandwidth**, a
中文介绍 OpenAI与博通（Broadcom）合作发布了其首款定制AI芯片「Jalapeño」，专为LLM推理设计。该芯片旨在加强OpenAI对AI技术栈的控制，提升计算经济性，并实现9个月的快速设计周期。社区分析表明，「Jalapeño」可能配备216GB HBM3E显存，性能参数约为7.1–7.4。
Agent identity in Claude Tag: a new access model for autonomous, team-wide AI
Agent identity in Claude Tag: a new access model for autonomous, team-wide AI
中文介绍 Claude官方博客宣布，其「Claude Tag」平台引入了「智能体身份」功能。这是一种全新的访问模型，旨在支持自主AI智能体在团队范围内的广泛应用，并强化AI在协作环境中的管理、安全性和责任归属。
Building effective human-agent teams
Building effective human-agent teams
中文介绍 Claude官方博客发布文章，探讨如何构建高效的人类与AI智能体协作团队。文章可能提供关于智能体设计、任务分配、沟通机制以及团队管理等方面的策略和最佳实践，旨在最大化人机协作的效能，提升工作流效率。
Introducing the FFASR Leaderboard: Benchmarking ASR in the Real World
中文介绍 Hugging Face博客宣布推出「FFASR 排行榜」，旨在为自动语音识别（ASR）模型提供在真实世界场景下的基准测试。该排行榜将帮助研究人员和开发者评估不同ASR模型的性能，推动该领域的进步和实际应用，提升语音识别的实用性。
Claude Tag 💬, Seedance 2.5 🎥, Mistral OCR 4 🧠
中文介绍 TLDR AI报道了AI领域的几项最新进展，包括Claude Tag相关功能、视频处理技术Seedance 2.5的发布以及Mistral OCR 4的更新。这些更新涵盖了智能体协作、多媒体内容处理及光学字符识别（OCR）等多个前沿AI应用领域，展示了AI技术的持续进步。
How GPT-5 helped immunologist Derya Unutmaz solve a 3-year-old mystery
GPT-5 Pro helped solve a 3-year-old immunology mystery, offering insights into T cell behavior. The breakthrough could support cancer and autoimmune research.
中文介绍 OpenAI新闻指出，免疫学家Derya Unutmaz借助GPT-5 Pro成功解开了一个困扰其3年的免疫学谜团。GPT-5 Pro在解析T细胞行为方面提供了关键洞察，这项突破性发现有望为癌症和自身免疫性疾病的研究带来新的方向和支持，展示了AI在科学研究中的潜力。
Helping build shared standards for advanced AI
OpenAI helps build shared standards for advanced AI, supporting evaluation frameworks, safety practices, and global cooperation through the Appia Foundation.
中文介绍 OpenAI宣布正通过Appia基金会，积极参与为先进人工智能（AI）建立共享标准的工作。此举旨在支持AI评估框架、推广安全实践，并促进全球范围内的合作，以确保AI技术的负责任发展和应用，构建更安全可信的AI生态。
Build real agentic apps using CUGA: two dozen working examples on a lightweight harness
中文介绍 Hugging Face博客联合IBM Research发布文章，介绍了如何利用轻量级「CUGA」工具链构建实际的智能体应用程序。文章提供了二十多个可运行的示例，旨在帮助开发者快速理解并实践智能体技术，加速AI应用的开发与部署，降低智能体开发门槛。
分享一下 mimo 的额度
6 回复 · 程序员 节点
大家都是怎么深度使用 Codex 的？我现在让它自己沉淀项目经验，但感觉还不够体系化
8 回复 · 程序员 节点
LinkedIn 创始人里德·霍夫曼炮轰： SpaceX 不是一家人工智能公司， xAI 是彻底的灾难
8 回复 · 程序员 节点
iPhone 屏幕防油渍防指纹涂层
9 回复 · Apple 节点
似乎 apple.com 买礼品卡加了 ip 检测
7 回复 · Apple 节点
macos27 的 siri ai 怎么可以不报 problem 啊？
15 回复 · Apple 节点
apple notes 已经支持一次性全部导出了
6 回复 · Apple 节点
升级 iOS 和 macos27 的能帮忙试试 surge 可用不
7 回复 · Apple 节点
我被 iOS27 的原生中文输入法折磨疯了
42 回复 · Apple 节点
摸鱼自由咯~ 升级到 macOS 27 dev 2 和 iOS 27 dev 2 之后, iPhone 镜像终于能用啦
17 回复 · Apple 节点
一个基于概率分布识别任意模型真假的项目，从此告别掺假！
本帖使用社区开源推广，符合推广要求。我申明并遵循社区要求的以下内容： 我的帖子已经打上 开源推广 标签： 是 我的开源项目完整开源，无未开源部分： 是 我的开源项目已链接认可 LINUX DO 社区： 是 我帖子内的项目介绍，AI生成、润色内容部分已截图发出： 是 以上选择我承诺是永久有效的，接受社区和佬友监督： 是 以下为项目介绍正文内容，AI生成、润色内容已使用截图方式发出 github.com GitHub - hanlinwenyuan/hlwy-ai-checker: 检查第三方AI API是否掺假以及渠道一致 检查第三方AI API是否掺假以及渠道一致 看见佬友们还是容易碰到api
今年的第一个南瓜
奶奶种的，香 29 个帖子 - 28 位参与者 阅读完整话题
蚂蚁阿福9.9得体脂秤，貌似还有减重活动
得去下载蚂蚁阿福软件，然后进入后点一分钱领体脂秤。跳转淘宝付款39.9，然后后面到货使用了会退30块钱。6.29还有个21天体重体质管理挑战赛，感兴趣的可以去看看。 72 个帖子 - 61 位参与者 阅读完整话题
【干草铺服务站】开放注册一波～
这几天看到很多佬友都缺 GPT 模型，所以决定开放一波注册，希望能帮助到各位。邀请码价格降为 300LDC，老用户不用感觉背刺哦，你们的 2 万美元依然可以爽蹬，新用户注册只送 300 美元。 干草铺站点：https://gancaopu.com 邀请码购买地址 https://shop.aini8.com 重要声明： 禁止破限，严禁分发，发现就封哦 rpm 限制为 10，正常使用即可 gpt 模型仅保证在 codex 中正常使用，其他使用场景不限制但是技术方面不保证其他场景可用 目前仅保证 gpt5.5 模型稳定可用，且仅保证在 codex 场景稳定可用，其他场景不限制但是技术层面不保证可用
GPT5.5提示词优化，解决废话特别多，车轱辘话来回说的问题
# 回答风格 ## 思维链 先判断，再解释，再给做法。 ## 语气 - 冷静、直接、克制 - 不写客套话、不写泛泛背景 - 不主动扩展无关知识 ## 歧义处理（最高优先级） **当问题存在关键歧义时，先用一句话追问，不要替用户猜测所有分支再逐一展开。不要直接回答如果是xx，就xx，而是先向用户确认，你问的是xx还是xx** 只有确认意图后才开始正式回答。 --- **先砍后补**——先用最短形式回答，只有信息确实不够时才加层。 排版规范是"允许使用的工具"，不是"必须填满的模板"。 一条命令能答完的问题，就只给一条命令加一句说明，不套任何结构。 # 输出结构 | 问题类型 | 结构 | |--
【陨落的一天】老板站在我工位旁半小时了，问我为什么还不开始敲代码。
Selected model is at capacity. Please try a different model. 28 个帖子 - 27 位参与者 阅读完整话题
劝大家私生活干净一点【娱乐】
35 个帖子 - 35 位参与者 阅读完整话题
土区开的Plus突然变成Pro 20x了
OpenAI又有bug了？怎么土区开的Plus突然变成Pro 20x了？ 102 个帖子 - 90 位参与者 阅读完整话题
360宣布已开发出能与Mythos模型相抗衡的国产对标产品: 图龙锋 和 仪天阵
我在联合早报看到的消息, 他也是援引路透和证券时报的消息. 三六零发布“中国版Mythos”图龙锋 周鸿祎称漏洞发现能力正成为新的战略能力 周鸿祎星期三（6月24日）在第十四届互联网安全大会上，发布了两款人工智能模型：漏洞自动化挖掘智能体“图龙锋”与自动化防御系统“仪天阵”。 周鸿祎介绍，“图龙锋”能自动发现软件漏洞，而“仪天阵”则负责实现网络防御和事件响应的自动化，其中“图龙锋”被视为中国版Mythos。 凭借在敏感系统中挖掘漏洞的强大能力，今年4月发布的Mythos已在华盛顿及多国政府高层，乃至整个网络安全行业引发高度警觉。而360此次发布的新产品，标志着中国企业迄今为止对该模型做出的最高
严重怀疑Any佬是何方神圣
Any站的质量、纯度无可置疑，用过的都说好。出问题恢复也嘎嘎快，但是又从来看不到Any佬冒泡。而且来了L站就可以享受Any佬的福利，这不妥妥的L最大福利嘛！嘎嘎好，嘎嘎给力的啊 158 个帖子 - 106 位参与者 阅读完整话题
Wikipedia Workers in Britain set global first by seeking union recognition
24 points · 5 comments
Half-Life 2 in a Browser
144 points · 46 comments
The Unbearable Cheapness of Open Weight Models
41 points · 9 comments
Medical students are using popular research tool to pump out misleading studies
41 points · 14 comments
Cloudflare launched self-managed OAuth for all
132 points · 54 comments
Zombie unicorns are haunting Silicon Valley
84 points · 38 comments
Exploring the internal representations of Pangram 3.3.2
30 points · 5 comments
Ending respiratory infections
147 points · 70 comments
Mixing Visual and Textual Code
46 points · 15 comments
LuaJIT 3.0 proposed syntax extensions
140 points · 84 comments
Blogging can just be stating the obvious
203 points · 70 comments
Anthropic says Alibaba illicitly extracted Claude AI model capabilities
360 points · 609 comments
OpenAI unveils its first custom chip, built by Broadcom
Announcement: https://openai.com/index/openai-broadcom-jalapeno-inference-...https://decrypt.co/371971/openai-broadcom-jalapeno-first-cus...https://www.cnn.com/2026/06/24/tech/openai-broadcom-jalapeno...
Computer use in Gemini 3.5 Flash
211 points · 134 comments
The Xteink X4 E-Ink Reader
229 points · 132 comments
RubyLLM: A Ruby framework for all major AI providers
381 points · 64 comments
PR spam today looks like email spam in the early 2000s
214 points · 123 comments
Show HN: Nub – A Bun-like all-in-one toolkit for Node.js
Colin here, creator of Nub. I’ve had the general shape of this in mind for years. Nub runs your code with stock `node`, augmented with a `--require` preload hook[0] that adds a transpiler (oxc-powered, packaged as a Node-API add-on), registers a module resolution hook[1], and injects polyfills as ne
45°C cooling design cuts data center water use to near zero
314 points · 203 comments
Qualcomm to Acquire Modular
https://investor.qualcomm.com/news-events/press-releases/new...https://www.modular.com/blog/qualcomm-to-acquire-modularhttps://x.com/clattner_llvm/status/2069769232477192354, https://xcancel.com/clattner_llvm/s
Show HN: Nimic – Pure Python as a systems language with AOT compilation
5 points · 1 comments
Markdy: Like Mermaid Diagrams, but for Motion
44 points · 5 comments
15 sorting algorithms in 6 minutes (2013) [video]
26 points · 3 comments
GLM-5.2 is a step change for open agents
217 points · 115 comments
Matt's Script Archive: The Scripts That Reshaped the Web
41 points · 11 comments
Bohemia Interactive: Cold War Assault Remastered Source Code on GitHub
33 points · 3 comments
Dostoyevsky isn't difficult
140 points · 133 comments
A Practical Guide to SSH Tunnels: Local and Remote Port Forwarding
320 points · 60 comments
Show HN: Brain Frog – Can you be random enough for 11 lines of JavaScript?
30 points · 28 comments
Show HN: Write SaaS apps where users control where their data is stored
Hello HN,I would like to share with you linkedrecords.com - an open source backend as a service I'm working on since some time now. You can think of it as an firebase/convex alternative with an interesting twist.In 2018 I needed to write large software requirements/architecture docume
今日主题
今日AI领域呈现出多维度的显著进展与深度思考。在核心技术层面，OpenAI与博通联手推出专为LLM优化的推理芯片，以及Google Gemini 3.5 Flash新增的计算机交互功能，标志着AI模型在性能与应用边界上的持续突破。与此同时，AI智能体技术成为焦点，无论是Anthropic为Claude引入的「智能体身份」与团队协作功能，还是IBM与Hugging Face发布的CUGA工具链，都在推动智能体的实际构建与部署。开源社区也贡献了如OpenMontage这类由智能体驱动的视频制作系统，以及Heron和Polygraph等强化智能体监控与记忆能力的新工具。行业观察者深入探讨了「元工具链」时代的到来、开放AI生态系统的重要性，以及为AI构建网络数据基础设施的必要性，共同勾勒出AI技术从底层创新到应用落地的全景图，尤其在人机协作和负责任AI发展方面，业界正积极探索共享标准与最佳实践。
01
模型发布/更新
Model Releases 33 篇OpenAI与博通联合发布LLM推理定制芯片
官方OpenAI News
OpenAI与博通公司（Broadcom）联合发布了一款名为「Jalapeño」的定制AI芯片。该芯片专为大型语言模型（LLM）推理任务进行优化，旨在显著提升AI系统的性能、能效和可扩展性，以满足不断增长的计算需求并推动AI基础设施创新。通过这次合作，OpenAI旨在加强对AI技术栈的控制，并实现计算经济性的提升，同时展现出仅9个月的快速设计周期。
Google Gemini 3.5 Flash增添计算机交互功能
官方DeepMind Blog
Google DeepMind宣布为Gemini 3.5 Flash模型引入「计算机使用」功能。这意味着该模型能够与计算机系统进行深度交互和操作，从而极大地扩展了其应用范围和能力，使其能够执行更复杂的任务，例如直接操作软件界面、处理文件或执行自动化脚本。此项更新标志着大模型从纯文本交互向更广泛的系统级操作迈进。
DeepSeek攻克AI研究「数十亿美元」级难题
大咖博客Two Minute Papers
Two Minute Papers报道，DeepSeek公司在AI领域取得了突破性进展，成功解决了一个长期困扰AI研究的「数十亿美元」级重大难题。尽管具体技术细节尚未完全披露，但这项成就预示着AI模型在性能、效率或应用边界上可能出现显著飞跃，有望为AI技术的未来发展和商业化应用带来深远影响，引发业界广泛关注。
02
产品发布/更新
Product 55 篇Claude Tag推出「智能体身份」新型访问模型
官方Claude Blog
Claude官方博客宣布，其「Claude Tag」平台引入了「智能体身份」功能，这是一种全新的访问模型。该功能旨在支持自主AI智能体在团队范围内的广泛应用，并强化AI在协作环境中的管理、安全性和责任归属。通过为每个智能体分配独特身份，团队能够更精细地控制智能体的权限，确保AI在企业级应用中的合规性与可信度。
IBM与Hugging Face发布CUGA智能体应用工具链
官方Hugging Face Blog
Hugging Face博客联合IBM Research发布文章，介绍了如何利用轻量级「CUGA」工具链构建实际的智能体应用程序。文章提供了二十多个可运行的示例，旨在帮助开发者快速理解并实践智能体技术。CUGA通过降低智能体开发的门槛，加速AI应用的开发与部署，使得开发者能够更高效地将智能体集成到各种解决方案中。
OpenMontage：开源AI智能体视频制作系统
开源项目GitHub Trending
OpenMontage 是一个开源的智能体驱动视频制作系统，通过整合12个流程、52种工具和500多种智能体技能，将AI编码助手升级为全功能的视频制作工作室。它旨在自动化视频内容创作，从脚本撰写到最终成片，大幅提升效率。该系统特别适用于内容创作者、营销人员及开发者，实现快速、批量化的视频生产，显著降低视频制作成本。
Heron：AI代理专用的eBPF可观测性工具
产品榜单Product Hunt
Heron 被描述为「AI代理的Wireshark」，它通过被动的eBPF可观测性技术，为AI代理提供深入的监控和分析能力。该工具旨在帮助开发者实时洞察AI代理的运行状态、行为模式和内部通信，从而更好地理解、调试和优化复杂的AI系统。Heron的推出，为确保AI代理的稳定高效运行提供了关键支持，尤其对于多智能体系统至关重要。
Polygraph：增强AI代理的跨代码库记忆与会话维持
产品榜单Product Hunt
Polygraph 旨在增强AI代理的能力，使其能够跨代码仓库获取信息，并维持会话记忆。这意味着AI代理不仅可以在多个项目或存储库之间理解和关联数据，还能在长时间的交互会话中记住上下文，从而更智能、更连贯地协助开发者完成复杂的软件开发任务。这极大地提升了AI代理在复杂编程环境中的实用性和协作效率。
03
行业动态
Industry 33 篇OpenAI参与构建先进AI共享标准与治理
官方OpenAI News
OpenAI宣布正通过Appia基金会，积极参与为先进人工智能（AI）建立共享标准的工作。此举旨在支持AI评估框架、推广安全实践，并促进全球范围内的合作，以确保AI技术的负责任发展和应用。通过在治理、安全和互操作性方面的标准化努力，OpenAI希望能够构建一个更安全可信的AI生态系统，推动整个行业健康发展。
MIT Tech Review探讨AI时代网络数据基础设施层兴起
综合资讯MIT Tech Review AI
麻省理工科技评论指出，随着AI的蓬勃发展，企业对大规模高质量数据的需求日益增长。然而，许多相关的网络数据受阻或非结构化，限制了AI模型的使用。文章强调，为应对这一挑战，构建一个支持AI应用、能够有效获取和处理网络数据的「网络数据基础设施层」正变得至关重要。这层基础设施将成为AI模型从海量网络信息中提取价值的关键。
Databricks高管呼吁建设开放前沿AI生态系统
大咖博客Latent Space
Databricks技术领导Matei Zaharia和Reynold Xin在一次访谈中强调了开放的前沿AI生态系统对企业的重要性。他们深入探讨了企业如何构建和利用「Agent Clouds」（智能体云），以实现更广泛的AI应用和创新，并指出开放性是推动AI普及和赋能企业的关键。开放生态能促进技术共享、加速创新，并确保更多企业能从中受益。
04
技巧与观点
Tips & Takes 44 篇Claude官方博客：构建高效人机智能体协作团队
官方Claude Blog
Claude官方博客发布文章，深入探讨如何构建高效的人类与AI智能体协作团队。文章可能提供了关于智能体设计原则、任务分配策略、优化人机沟通机制以及团队管理等方面的最佳实践。其核心目标是最大化人机协作的效能，提升整体工作流效率，为企业和个人用户提供在AI时代实现更智能、更无缝团队合作的指导。
Latent Space：AI领域迈入「元工具链」时代
大咖博客Latent Space
Latent Space的AINews指出，当前AI领域正进入一个「元工具链」时代，预示着超越传统工具链工程的更高层级架构的兴起。这可能涉及更复杂或更集成的AI系统协调与管理框架，强调了AI智能体或系统的更高级别组织和控制范式。这种趋势表明，开发者将从单个工具的使用转向更宏观、更自动化的AI系统设计与治理。
科学家发现优化AI代理的「新语言」范式
大咖博客Two Minute Papers
Two Minute Papers报道，科学家们研究发现了一种更优化的「语言」或编程范式，旨在显著提升AI代理的理解、推理和执行能力。这项研究有望解决现有AI代理在复杂任务处理和多模态交互中面临的局限性，从而推动AI代理技术向更高级别发展，使其能更高效、更准确地完成指令，并适应更广泛的应用场景。
GPT-5 Pro助力免疫学家解开T细胞行为谜团
官方OpenAI News
OpenAI新闻指出，免疫学家Derya Unutmaz借助GPT-5 Pro成功解开了一个困扰其三年的免疫学谜团。GPT-5 Pro在解析复杂T细胞行为方面提供了关键洞察，这项突破性发现有望为癌症和自身免疫性疾病的研究带来新的方向和支持。此案例生动展示了先进AI模型在复杂科学研究领域中作为强大辅助工具的巨大潜力与应用价值。