TypeScript · ★ 119,087 · 🍴 6,366 · 📈 530 stars today
Collection of publicly available IPTV channels from all over the world
中文介绍 全球公开 IPTV 频道资源聚合项目，收录来自世界各地数以千计的免费直播流地址，按国家、语言、分类整理成 M3U 播放列表，供媒体播放器直接调用。适合希望在个人设备上收看国际电视直播的用户，也常用于 Home Assistant、Kodi 等智能家居媒体中心集成。
Shell · ★ 58,343 · 🍴 6,300 · 📈 1,514 stars today
Production-grade engineering skills for AI coding agents.
中文介绍 面向 AI 编码智能体的生产级工程技能集合，由 Google Chrome 团队工程师 Addy Osmani 维护。涵盖代码质量、架构决策、安全实践等方面的结构化提示与指导规范，帮助 Claude Code、Cursor 等 AI 编码工具在实际工程场景中产出更可靠、符合工业标准的代码，适合希望提升 AI 辅助开发质量的工程团队参考使用。
Ruby · ★ 30,846 · 🍴 7,553 · 📈 83 stars today
Open-source live-chat, email support, omni-channel desk. An alternative to Intercom, Zendesk, Salesforce Service Cloud etc. 🔥💬
中文介绍 开源全渠道客户支持平台，整合实时在线客服、邮件工单、社交媒体等多种沟通渠道于统一工作台，是 Intercom、Zendesk 的自托管替代方案。基于 Ruby on Rails + Vue.js 构建，支持 Docker 部署，适合中小企业和独立开发者搭建自主可控的客户服务系统，规避 SaaS 工具的高订阅费用与数据隐私风险。
Shell · ★ 226,915 · 🍴 20,179 · 📈 924 stars today
An agentic skills framework & software development methodology that works.
中文介绍 一套面向软件开发的智能体技能框架与方法论，提供结构化的 AI 编程工作流规范，帮助 AI 编码智能体更可靠地完成复杂工程任务。强调实际可用性而非理论，包含提示模板、任务分解策略等，适合希望系统化提升 AI 辅助开发效率的开发者和工程团队采用。
Swift · ★ 36,282 · 🍴 1,033 · 📈 1,487 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 官方开源的 Mac 上 Linux 容器工具，基于轻量级虚拟机运行 Linux 容器，专为 Apple Silicon 优化。用 Swift 编写，提供类似 Docker 的容器创建与运行体验，无需依赖 Docker Desktop，启动更快、资源占用更低，适合在 M 系列 Mac 上进行 Linux 应用开发、测试和本地容器化工作流的开发者使用。
Python · ★ 1,998 · 🍴 429 · 📈 270 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 是一款免费开源的个人媒体库管理服务端，可桥接 Spotify、Tidal 等流媒体平台与家中的网络音箱（如 Sonos、Cast 设备），实现统一播放控制。作为核心后端服务，通常与 Home Assistant 集成部署，适合智能家居爱好者打造本地优先、跨平台的家庭音乐串流解决方案。
Go · ★ 2,359 · 🍴 216 · 📈 190 stars today
Local-first session intelligence and analytics for coding agents, supporting Claude Code, Codex, and more than 20 other agents. Also: 100x faster replacement for ccusage!
中文介绍 本地优先的 AI 编码智能体会话分析工具，支持 Claude Code、Codex 及 20 余款主流编码智能体的使用数据统计与可视化。定位为 ccusage 的百倍速替代品，在本地解析会话记录并提供 Token 消耗、成本、效率等维度的洞察报告，帮助开发者了解 AI 工具的实际使用情况并优化工作流。
Python · ★ 8,884 · 🍴 1,305 · 📈 238 stars today
LMCache: Supercharge Your LLM with the Fastest KV Cache Layer
中文介绍 专为大语言模型推理加速设计的高性能 KV Cache 层，通过跨请求复用 KV 缓存显著降低 TTFT（首 Token 延迟）和计算开销。可与 vLLM 等推理框架集成，支持分布式部署场景下的缓存共享，适合需要处理大量重复上下文（如 RAG、多轮对话、长系统提示）的 LLM 推理服务提供商和 AI 平台工程师。
C · ★ 134,659 · 🍴 8,078 · 📈 370 stars today
Microsoft PowerToys is a collection of utilities that supercharge productivity and customization on Windows
中文介绍 微软官方出品的 Windows 效率增强工具集，提供窗口布局管理（FancyZones）、文件批量重命名、快速启动器（PowerToys Run）、颜色拾取器、键盘映射等数十种实用功能。基于 C# 和 C++ 开发，持续迭代更新，适合希望深度定制 Windows 操作体验、提升日常工作效率的开发者和高级用户。
Python · ★ 14,104 · 🍴 1,487 · 📈 127 stars today
Simple, unified interface to multiple Generative AI providers
中文介绍 吴恩达团队开源的多 AI 提供商统一调用库，提供一致的 Python API 接口，屏蔽 OpenAI、Anthropic、Google、Mistral 等不同厂商 SDK 的差异，支持一行代码切换模型。适合需要同时评测多个大模型、构建与模型无关应用，或快速迁移底层 AI 服务提供商的开发者和研究人员使用。
Python · ★ 4,422 · 🍴 334 · 📈 804 stars today
Security scanner for AI agent skills. Detect vulnerabilities, malicious patterns, and security risks.
中文介绍 NVIDIA 开源的 AI 智能体技能安全扫描工具，专门检测 AI 编码智能体插件和技能定义中的安全漏洞、恶意代码模式及潜在风险。随着 AI 智能体生态中第三方技能快速增长，该工具为企业和开发者在集成外部智能体技能前提供自动化安全审查能力，降低供应链安全风险。
Kotlin · ★ 47,478 · 🍴 8,081 · 📈 93 stars today
翻墙-科学上网
中文介绍 综合性翻墙与科学上网资源汇总项目，收录 V2Ray、Xray、Shadowsocks、Trojan 等主流代理工具的配置教程、客户端下载及免费节点信息，同时整理各平台翻墙软件推荐。面向需要访问境外互联网资源的中文用户，提供从工具选择到实际配置的一站式参考指南。
Rust · ★ 33,630 · 🍴 1,398 · 📈 20 stars today
Rust-based platform for the Web
中文介绍 基于 Rust 构建的高性能 Web 前端工具链平台，提供 JavaScript/TypeScript 的编译转换（替代 Babel）和代码压缩（替代 Terser）功能，速度比同类 JavaScript 工具快数十倍。被 Next.js、Vite 等主流框架采用，适合追求极致构建性能的前端工程团队，也可作为 Webpack、Rollup 等打包工具的底层转换引擎。
★ 140,299 · 🍴 34,656 · 📈 109 stars today
FULL Augment Code, Claude Code, Cluely, CodeBuddy, Comet, Cursor, Devin AI, Junie, Kiro, Leap.new, Lovable, Manus, NotionAI, Orchids.app, Perplexity, Poke, Qoder, Replit, Same.dev, Trae, Traycer AI, VSCode Agent, Warp.dev, Windsurf, Xcode, Z.ai Code, Dia & v0. (And other Open Sourced) System Prompts
中文介绍 收录当前主流 AI 编程与生产力工具完整系统提示词的公开仓库，涵盖 Claude Code、Cursor、Devin、Replit、Lovable、Manus、Perplexity 等 20 余款产品的实际 system prompt 及模型配置信息。对研究 AI 产品提示工程策略、竞品分析或构建类似工具的开发者具有重要参考价值。
👍 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
👍 2
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
👍 2
Multimodal Large Language Models (MLLMs) have shown promising reasoning capabilities in general domains, yet their performance remains limited in specialized settings such as healthcare, especially in multilingual and low-resource scenarios. This gap is critical in regions like rural India, where pa
👍 25
Holistic visual tokenizers are fundamental to unified multimodal models (UMMs) as they map diverse visual inputs into a unified representation space. In this paper, we present HYDRA-X, the first UMM that unifies image and video tokenization within a single Vision Transformer (ViT). Our design is dri
👍 102
Ultra-long-context capability is becoming indispensable for frontier LLMs: agentic workflows, repository-scale code reasoning, and persistent memory all require the model to jointly attend over hundreds of thousands to millions of tokens, yet the quadratic cost of softmax attention makes this untena
👍 4
Search Agents -- large language models augmented with search tools -- have intensified the need for future-proof evaluation benchmarks. Existing benchmarks such as BrowseComp rely on static knowledge, making them vulnerable to test-set contamination and parametric memorization. Consequently, models
👍 118
Large language model (LLM) agents have achieved strong performance on a wide range of benchmarks, yet most evaluations assume static environments. In contrast, real-world deployment is inherently dynamic, requiring agents to continually align their knowledge, skills, and behavior with changing envir
👍 18
Latent chain-of-thought compresses reasoning by replacing visible reasoning traces with continuous hidden-state recurrence, but existing formulations are difficult to optimize with standard on-policy reinforcement learning (RL) and hard to interpret causally. Our key insight is that a single pair of
👍 2
The potential impacts of world models (WMs, i.e., learned simulators) on robotics are far-reaching -- policy evaluation, policy improvement, and test-time planning -- all with limited real-world interaction. To unlock these downstream capabilities, a WM needs to jointly satisfy three desiderata: (i)
👍 21
LLM-based agents have shown increasing potential in automating scientific discovery. Given an optimizable metric and an execution environment, they can propose, validate, and iterate scientific solutions, and have produced results that outperform human-designed approaches. As model capabilities cont
👍 74
Recent image generators have demonstrated impressive photorealism and instruction-following capabilities in single-image generation and editing. However, constrained by their architectures, they cannot achieve interleaved generation (text-image sequence), which has crucial applications in visual nar
👍 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
👍 12
Multimodal image fusion aims to integrate complementary information from different modalities into a fused image that preserves rich local details while maintaining globally consistent appearance. Existing approaches build shared representations on 2D feature grids, which excel at modeling local str
👍 14
Speculative decoding (SD) addresses the high inference costs of LLMs by having lightweight drafters generate candidates for large verifiers to validate in parallel. Existing draft-verify methods use binary decisions: accept or fully recompute. Yet we find that many rejected tokens can be verified co
👍 10
Deep search requires agents to answer complex questions through multi-step web search, browsing, evidence comparison, and synthesis. A central challenge is deciding how to search when several directions look plausible but only some will later lead to reliable evidence. If an agent greedily follows t
👍 3
Compact language models (LMs) reduce cost, latency, and deployment risk for tool agents. Yet MCP-style tool use requires more than isolated function calling: an agent must discover tools from live catalogs, satisfy schemas, preserve dependencies across intermediate outputs, and ground final response
👍 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
👍 8
Adversarial robustness evaluations of large language models (LLMs) typically report attack success rate (ASR) under fixed query budgets, implicitly treating all attacks as equally costly. In practice, the computational expense of different attack strategies can vary by orders of magnitude. Consequen
👍 2
We present Flash-GMM, a fused Triton kernel for efficient computation of Gaussian Mixture Models (GMMs) over large-scale data in a single GPU pass. By eliminating the need to materialize the full responsibility matrix in GPU memory, Flash-GMM achieves a 20times speedup over existing implementations
👍 2
Built on pretrained vision foundation models (VFMs), representation autoencoders (RAEs) have recently emerged as a promising approach for constructing semantically rich latent spaces for image generation. However, their reconstruction quality often remains suboptimal, largely because deep VFM repres
👍 1
Advances in handwritten text recognition have enabled large-scale transcription of historical documents, but still provide limited access to interpretable visual measurements for paleography, the study of historical scripts. In this paper, our main insight is that morphological script analysis, in p
👍 7
Visual reasoning requires integrating evidence distributed across regions, attributes, and relations, making single-chain reasoning prone to early perceptual commitment and hallucination. We propose Visual Para-Thinker++, a single-policy multi-agent framework in which one shared MLLM policy is insta
👍 6
On-policy distillation (OPD) trains a student on its own trajectories with dense per-token supervision from a stronger teacher, and often outperforms off-policy distillation and standard reinforcement learning. However, we find that its effectiveness implicitly relies on two assumptions that frequen
👍 94
Computer-use agents (CUAs) increasingly operate in runtimes that combine visual desktop control, command-line execution, code editing, browsers, and external tools. Existing benchmarks, however, often evaluate these interfaces as separable capabilities, leaving long-horizon cross-interface orchestra
👍 4
Representation alignment with pretrained vision models has recently shown strong potential for accelerating diffusion transformer training. By aligning intermediate diffusion features with clean-image representations from self-supervised vision encoders, existing methods improve convergence and gene
👍 74
Multimodal Large Language Models (MLLMs) have demonstrated remarkable success in visual understanding, yet their performance degrades significantly under real-world visual corruptions. While existing robustness enhancement approaches exist, they are limited: black-box feature alignment lacks interpr
👍 3
Robotic simulators are a cornerstone of modern research in aerial robotics, serving both as a vehicle for the development of new control algorithms and as the data source for training reinforcement learning (RL) policies. Yet, existing quadcopter learning environments often face a trade-off between
👍 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
👍 2
Large language models deployed as agents over large tool catalogs face a critical tool-retrieval bottleneck. As embedding-based retrieval approaches rely on compact encoders that may under-capture specialized tool semantics, parametric tool retrieval addresses this by encoding each tool as a virtual
👍 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
@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
中文介绍 Anthropic 官方发布了 Claude 模型的 prompting 完整指南，但藏在 API 文档深处鲜有人读。博主将其提炼翻译，整理成可直接上手的 prompt 工程手册，涵盖最佳实践与技巧，适合想系统掌握 Claude 高级用法的开发者和工程师参考。
@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
中文介绍 Recursive 发布自动化 AI 研究系统早期成果，覆盖模型训练与 GPU 内核基准测试三项实验。该系统可自主开展 AI 研究任务，是迈向「AI 自动研究 AI」的早期探索，展示了自动化研究流水线在前沿硬件与训练优化领域的初步能力。
@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
中文介绍 大多数人把 Claude 4 当普通聊天工具用，从未真正跑过 agent 系统。博主分享 14 步构建自我改进 agent 的完整方法，涵盖循环控制、动态工作流与 routines 设计，核心在于充分利用 Claude 4 的 agent 推理能力而非仅用作更大上下文窗口的 Sonnet。
@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
中文介绍 四个月前 Anthropic 还是 AI 圈的宠儿，如今形势逆转。博主梳理 Anthropic 近期舆论变化，分析其从「安全标杆」走向争议中心的过程，涉及 OpenAI 竞争格局演变及多位关键人物离职事件，探讨 Anthropic 是否正在失去行业信任与话语权。
@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
中文介绍 如何构建在垂直领域真正有竞争力的 agent？博主指出基础架构（模型外层套 while 循环）已被标准化，真正的差异化在于领域专业知识的深度融入与任务质量把控。文章探讨垂直 agent 超越通用方案的关键设计思路，适合构建行业 AI 产品的工程师参考。
@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
中文介绍 博主质疑 Anthropic「安全实验室」形象背后的矛盾：其对外塑造负责任 AI 开发者的人设，但围绕开源 AI 的一系列行为模式显示出截然不同的利益导向。文章梳理 Anthropic 在开源领域的具体动作，提出其公关叙事与实际策略之间存在明显落差。
@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
中文介绍 Coinbase 正式推出面向 AI agent 的服务「Coinbase for Agents」，允许 AI agent 直接连接用户 Coinbase 账户，在用户设定的权限范围内自主完成交易、支付和工作流执行。今日上线，以 MCP 协议提供接入，是加密原生 agent 自主操作的重要基础设施。
@ORACLEAIFND · 31.9K 粉丝 · 63.6K 阅 · 1.5K 赞 · 563 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
中文介绍 ORACLE 项目宣称旗下自主 AI agent 已在预测市场 Polymarket 上进行交易，并援引「2026 年 Polymarket 逾 30% 活动来自算法与 AI 钱包」的数据。内容疑似加密项目营销，真实性存疑，需谨慎甄别。
@OracleTrdading · 39.7K 粉丝 · 61.5K 阅 · 1.5K 赞 · 577 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
中文介绍 ORACLE 项目宣称旗下自主 AI agent 已在预测市场 Polymarket 上进行交易，并援引「2026 年 Polymarket 逾 30% 活动来自算法与 AI 钱包」的数据。内容与多个账号完全重复，疑似加密项目协同营销，需谨慎甄别。
@OracleTrdade · 41.7K 粉丝 · 60.6K 阅 · 1.5K 赞 · 560 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
中文介绍 ORACLE 项目宣称旗下自主 AI agent 已在预测市场 Polymarket 上进行交易，并援引「2026 年 Polymarket 逾 30% 活动来自算法与 AI 钱包」的数据。内容与多个账号完全重复，疑似加密项目协同营销，需谨慎甄别。
@OracleMarkett · 45.0K 粉丝 · 60.1K 阅 · 1.5K 赞 · 568 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
中文介绍 ORACLE 项目宣称旗下自主 AI agent 已在预测市场 Polymarket 上进行交易，并援引「2026 年 Polymarket 逾 30% 活动来自算法与 AI 钱包」的数据。内容与多个账号完全重复，疑似加密项目协同营销，需谨慎甄别。
@elvissun · 45.0K 粉丝 · 42.0K 阅 · 502 赞 · 45 转
99% people are using /goal and loops wrong. The hype they hear is "long-running loops prompting autonomous agent": point it at a task, walk away, come back to working code. But top agentic engineers
中文介绍 博主指出大多数人对 /goal 指令和 agent 循环的用法存在根本误区，光靠「甩任务走人」并不可靠。真正的 agentic 工程师会引入类似损失函数的目标约束机制，在 30 小时内用单一 prompt 完成产品提炼。文章提供完整 playbook，聚焦如何精确定义目标与约束以驱动稳定自主循环。
@leerob · 258.6K 粉丝 · 36.8K 阅 · 586 赞 · 40 转
At Cursor, we run thousands of agents to help us train the next version of Composer. We give them research tasks, and if they aren't succeeding or run into issues, they DM us on Slack or page us via
中文介绍 Cursor 工程师分享内部实践：每天运行数千个 agent 来训练下一版 Composer 模型，这些 agent 执行研究任务，遇到问题时会主动通过 Slack 私信或 PagerDuty 告警联系团队。这是递归式 agent 系统在生产环境落地的真实案例，展示了人机协作中「agent 主动上报」的工程模式。
We are in the strangest timeline.
中文介绍 Latent Space AI新闻简报报道，Fable与Mythos两款AI产品被官方认定风险过高而不予发布。报道以「我们正处于最奇异的时间线」作为开篇，暗示当前AI发展局势的不寻常性。具体风险内容及相关机构决策细节未在摘要中披露。
中文介绍 Allen AI在Hugging Face发布olmo-eval评估工作台，专为模型开发循环设计，旨在为研究人员提供系统化的模型评估流程支持。正文内容暂未提供更多细节。
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推出三门Academy新课程，面向职场应用场景，帮助用户构建实用AI技能、建立可复用工作流程，并将AI智能体应用于日常工作任务。课程定位于推动下一阶段工作模式变革。
a quiet day lets us highlight a great concept from Peter Steinberger, Boris Cherny, and Andrej Karpathy
中文介绍 Latent Space AI新闻简报介绍「Loopcraft：循环叠加的艺术」概念，重点引用了Peter Steinberger、Boris Cherny和Andrej Karpathy提出的相关理念，在相对平静的新闻日中深度解读这一AI开发方法论。
Preply uses OpenAI to launch AI-generated lesson summaries, providing personalised feedback and language learning exercises.
中文介绍 语言学习平台Preply借助OpenAI技术推出AI生成课程摘要功能，为学习者提供个性化反馈与语言练习内容，将AI辅助与真人教师相结合，实现个性化语言学习体验的规模化落地。
中文介绍 TLDR AI 2026年6月12日简报涵盖：OpenAI宣布收购Ona、Anthropic就此前立场作出调整，以及小米发布MiMo代码模型等多项AI行业动态。
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正资助研究，探讨数百万AI智能体在网络上大规模相互交互时可能引发的潜在风险。负责AGI安全与对齐研究的Rohin Shah表示，智能体大众化普及后的群体交互安全问题已引起该公司高度关注。
a quiet day lets us reflect on a great essay
中文介绍 Latent Space AI新闻简报以Sarah Guo的一篇文章为核心，探讨开放模型、模型实验室与智能体实验室之间的竞争关系，以及哪些能力目前仍无法通过训练获得等议题。
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《AI内容透明度行为准则》，致力于推进AI生成内容的溯源标准与工具建设，帮助公众识别和理解AI生成内容，以支持欧洲构建可信赖的AI生态系统。
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 西班牙国际银行BBVA已将ChatGPT Enterprise部署至全行10万名员工，并与OpenAI达成深度合作，将AI置于银行业务核心，加速推进全球范围内AI驱动的银行业务转型。
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，旨在为Codex扩展安全、持久的云端运行环境，支持企业工作流程中长时间运行的AI智能体任务，进一步强化OpenAI在企业级AI智能体领域的能力布局。
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借助OpenAI Codex构建黑洞模拟程序，用于研究极端物理现象并验证爱因斯坦广义相对论。Codex在科研场景中的应用展示了AI辅助工具在基础科学研究领域的实际价值。
中文介绍 Hugging Face发布PyTorch性能分析系列第二篇，介绍如何从标准nn.Linear层出发，逐步演进至融合MLP（多层感知机）实现，重点探讨模型计算效率优化与算子融合技术在实际开发中的应用。
中文介绍 TLDR AI 2026年6月11日简报涵盖：Anthropic创始人Dario Amodei的政策动态、谷歌DiffusionGemma模型新进展，以及WhatsApp计划开放机器人接入等多项AI行业要闻。
Access OpenAI models and Codex through Oracle Cloud, using existing commitments to build and deploy AI with enterprise security and governance.
中文介绍 OpenAI宣布与Oracle Cloud合作，企业用户可通过现有Oracle云计算承诺额度直接调用OpenAI模型及Codex服务，在企业级安全与治理框架下构建和部署AI应用，进一步拓展OpenAI企业服务渠道。
本期技术面呈现「美股结构偏强、加密持续承压、中概全面下行、商品外汇分化」的格局。美股方面，SPY与QQQ均保持多头排列，SPY现价741.75美元高于SMA50（722.80）与SMA200（686.30），QQQ的RSI14达55.2，5日涨幅2.31%，短期动量相对健康。然而个股内部分化明显：MSFT（RSI14=37.1）与META（RSI14=34.7）均跌破SMA20/50/200全部三条均线，空头排列信号确立，AAPL近5日下跌5.27%，RSI14降至44，中短期均线压力增大。宏观指标方面，VIX（17.68）1日内急跌9.05%、5日跌幅17.81%，显示近期恐慌情绪明显消退，但SMA50开始下穿SMA200形成死叉，需持续观察。10年期美债收益率（4.49%）RSI14位于50.8中性区域，多头排列结构维持，对权益资产的压制暂趋缓和。加密市场整体技术面偏弱：加密恐慌贪婪指数仅为18（极度恐慌），加密总市值2.28万亿美元，BTC主导率56.6%。BTC（64385美元）与ETH（1678美元）价格均大幅低于SMA20/50/200，ETH的RSI14已跌至32.5，SOL的RSI14为39，三大主流币均呈空头排列，市场情绪极度悲观。中概股方面，BABA（RSI14=29.6，超卖）、PDD（RSI14=32.4）和JD均处于空头排列，整体下行压力明显。黄金期货当日大涨3.63%但RSI14仅35.4，短期反弹与中期空头排列形成背离，走势存在不确定性；WTI原油1日跌3.23%、5日跌6.25%，技术面偏弱。美元指数DXY（99.81）RSI14为58.7，多头排列，技术面相对稳健。
当前价112.82美元，RSI14跌至29.6进入超卖区域，SMA20/50/200分别为125.81/130.33/149.57，三均线呈空头排列且价格全面低于三线。MACD为-4.9544，信号线为-3.3542，柱状图持续负值扩大。距52周高点下跌41.44%，系统信号同时触发「RSI超卖」与「空头排列」，下行动量显著。
当前价64385美元，SMA20/50/200分别为67511/74157/77769，价格深埋三条均线之下，空头排列结构完整。RSI14为36.4，接近超卖边界。MACD为-3544，信号线-3565，二者虽触发金叉但均处于大幅负值区域，反弹动能有限。加密恐慌贪婪指数仅18（极度恐慌），加密总市值24小时仅微涨1.01%，宏观情绪压制明显。
当前价390.74美元，RSI14为37.1，逼近超卖区间。SMA20/50/200分别为419.75/411.80/453.73，价格低于全部三条均线且三线呈空头排列。MACD为-4.1351，信号线为+1.2713，MACD柱负向且MACD与信号线之间存在明显负差距，趋势系统标注「bearish」。近5日跌幅达6.22%，距52周高点已回撤29.65%，下行趋势结构清晰。
当前价4238.80美元，1日内大涨3.63%但5日仍下跌2.27%，走势分歧。RSI14为35.4，接近超卖但尚未触发信号。价格低于SMA20（4421.88）、SMA50（4587.76）及SMA200（4419.51），整体空头排列，但单日反弹力度较强，短期动能与中期趋势出现背离，暂无明确方向信号，趋势系统亦标注「neutral」。
当前价741.75美元，SMA50/200分别为722.80/686.30，价格高于两条均线，趋势系统标注「bullish」且多头排列信号触发。RSI14为52.9，处于中性偏多区域。MACD为3.77，信号线为7.39，虽然MACD尚低于信号线存在一定弱势，但价格距52周低点上涨25.32%，结构性上行趋势较为完整，近1日及5日涨幅分别为0.54%和0.57%，短期动量稳定。
VIX 恐慌指数
10Y 美债收益率 (%)
美元指数 DXY
S&P 500 ETF
Nasdaq 100 ETF
Apple
Microsoft
Nvidia
Alphabet
Tesla
Meta
Bitcoin
Ethereum
Solana
阿里巴巴 (BABA)
拼多多 (PDD)
京东 (JD)
腾讯控股 (0700.HK)
黄金期货
WTI 原油期货
美元 / 人民币
本报告仅基于公开行情数据计算的技术指标进行客观描述，仅供技术指标解读参考，不构成任何投资建议或操作指引。过去走势不代表未来表现，市场存在不可预测风险，投资者应结合自身风险承受能力独立决策。
The Tasmanian Liberal senator has announced he will leave parliament before the end of the year. Follow the day’s news live Get our breaking news email, free app or daily news podcast The Senate will deliver its report from the NDIS inquiry on Tuesday. Butler doesn’t directly answer a question about
中文摘要 澳大利亚塔斯马尼亚州自由党参议员乔纳森·杜尼亚姆宣布将于年底前退出政坛。与此同时，悉尼河流中发现一对父女陈尸。澳大利亚参议院定于周二发布NDIS（国家残障保险计划）调查报告。
Brazil and Morocco supporters sang, danced and chanted their way to one of the most anticipated group-stage matches.
中文摘要 巴西队与摩洛哥队在热烈氛围中展开小组赛对决，两队球迷赛前载歌载舞、高呼助威，该场比赛被视为本届小组赛阶段最受瞩目的焦点赛事之一。
Charlie Verco managed to grab hold of the woman and bring her back to shore after the Sydney shark attack on Saturday Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Elite paddleboarder Charlie Verco has only seen one shark bigger th
中文摘要 澳大利亚悉尼库吉海滩周六发生鲨鱼袭击事件，一名女性遭鲨鱼咬伤，水中血迹弥漫。精英桨板运动员查理·韦尔科随即将其拖救上岸，该事件引发当地广泛关注。
A security expert who had recently become chief of staff to the new defense minister was abducted, the latest example of violence gripping the country.
中文摘要 海地一名高级安全官员遭绑架，此人近期刚刚出任新任国防部长幕僚长。该事件再次凸显海地国内持续恶化的暴力与安全形势。
An note on the Ticketmaster website for buying tickets to the game caused confusion and backlash among New Yorkers.
中文摘要 Ticketmaster网站上一则关于购票的提示信息引发纽约尼克斯球迷恐慌，担忧无法入场观看比赛。Ticketmaster随后回应称，球迷不会因此被拒之门外，并对造成的混乱表示歉意。
A niche group of content creators called "virtual parents" have become hugely popular among young people.
中文摘要 一类被称为「虚拟父母」的网络内容创作者在年轻人群体中迅速走红。部分年轻人因感受不到来自原生家庭的肯定与认可，转而在网络上寻求陌生中年夫妇的情感支持与鼓励。
The children — four girls and one boy — were among 10 people in the van at the time of the crash. An infant was seriously injured, the police said.
中文摘要 加拿大安大略省农村地区发生一起面包车与SUV相撞的严重交通事故，造成5名儿童死亡，死者包括4名女孩和1名男孩，均乘坐于载有10人的面包车内。警方表示，另有一名婴儿伤势危重。
Thousands mourn Thai princess as body taken to lie in state in Bangkok
中文摘要 泰国公主遗体被移送曼谷供民众瞻仰，数千名民众聚集沿途悼念，场面庄严肃穆，社会各界对这位王室成员的离世表达了深切哀思。
Dashcam shows US cop accidentally shooting colleague during ‘horseplay’.
中文摘要 美国一段行车记录仪视频显示，一名警察在与同僚嬉闹玩耍时意外走火，导致同事中弹受伤。该视频曝光后引发公众对警方执勤规范与枪支管理问题的广泛关注。
More than 3,700 people in Lebanon have died in the war between Israel and Hezbollah. In a village in southern Lebanon, one airstrike last month killed 14 people, including 10 women and children.
中文摘要 以色列与黎巴嫩真主党冲突已造成黎巴嫩逾3700人死亡。黎巴嫩南部一村庄上月遭以军空袭，单次袭击即造成14人罹难，其中包括10名妇女和儿童，当地居民仍沉浸在巨大悲痛之中。
Exiled Russian journalist Vladimir Raevsky says laughter can be a weapon against authoritarianism. NPR's Michele Kelemen reports on his history-inspired comedy act.
中文摘要 俄罗斯流亡记者弗拉基米尔·拉耶夫斯基转型为喜剧演员，以历史为题材创作脱口秀节目，他认为笑声是对抗威权主义的有力武器。美国全国公共广播电台对其创作经历进行了专题报道。
Thousands of police deployed to keep the rallies apart as far-right 'remigration' initiative gains traction.
中文摘要 意大利罗马爆发大规模移民议题对立示威，支持移民与反对移民的两批人群分别上街游行。当局部署数千名警察以维持秩序、隔离双方。极右翼推动的「反移民遣返」倡议近期在意大利持续升温。
An Iranian foreign ministry official sought to temper expectations, saying there were no plans for a Sunday signing and an agreement could be inked in the coming days.
中文摘要 特朗普表示伊核协议将于周日正式签署，但伊朗外交部官员随即否认，称目前没有周日签署的计划，协议可能在未来数日内完成签署。双方在协议签署时间表上存在明显分歧。
Explosions shook southern Lebanon on Saturday as Israel continued its attacks.
中文摘要 美国宣称伊朗核协议将于周日签署之际，以色列周六持续对黎巴嫩南部发动军事打击，多处地点传出爆炸声，局势仍处于紧张状态。
The European Union's new migration policy has come into force.
中文摘要 欧盟新移民政策已正式生效，各界对其实际执行效果存有疑虑。新规旨在加强欧盟边境管控并统一成员国移民处理机制，但能否有效应对移民压力仍是舆论争议焦点。
For several global central banks, the question of whether the Iran war poses more of an immediate danger to inflation or to growth is likely to remain open in the coming week.
中文摘要 伊朗战争爆发100天后，美联储与英国央行等多家全球主要央行在利率决策上保持审慎观望。各央行面临两难困境：伊朗战争究竟对通胀还是对经济增长构成更直接威胁，这一问题在未来一周内仍难以定论，预计两家央行将按兵不动，维持现有利率水平。
Wall Street has an answer for retail investors seeking to profit from elevated interest rates and dodge defaults in private credit: funds that buy collateralized loan obligations.
中文摘要 在高利率环境持续及私募信贷市场违约风险上升的背景下，华尔街推出以担保贷款凭证（CLO）为底层资产的ETF产品，为散户投资者提供参与高息市场的渠道，同时规避私募信贷的违约风险。CLO ETF市场近期呈现快速扩张态势。
The news doesn’t stop when markets close. Hosts David Gura, Christina Ruffini and Lisa Mateo bring clarity, context and a bit of humor to the weekend’s biggest headlines, LIVE from New York. Joined by Swarovski CEO Alexis Nasard, Johns Hopkins Center for Health Security Senior Scholar Dr. Amesh Adal
中文摘要 彭博《本周末》节目于2026年6月13日在纽约直播，由主持人David Gura、Christina Ruffini及Lisa Mateo主持，嘉宾包括施华洛世奇首席执行官Alexis Nasard以及约翰斯·霍普金斯大学健康安全中心资深学者，围绕当周重大财经与社会热点话题展开讨论。
While plenty of Americans are emphatically unready to give up caffeine, many are experimenting with a new range of options beyond the traditional cup of hot java, paying heed to caffeine’s impact on their sleep, mood and energy level. Consumers are becoming more cognizant of “energy management” in t
中文摘要 美国消费者对咖啡因的态度正在转变，越来越多的人开始关注咖啡因对睡眠、情绪及精力水平的影响，并探索传统热咖啡以外的多元化低咖啡因或替代饮品选项。"咖啡因极简主义"趋势兴起，消费者对能量摄入的意识明显提升。
Swarovski CEO Alexis Nasard joined Bloomberg This Weekend's Christina Ruffini for a tour and discussion about the significant brand transformation the company is undergoing as it aims to modernize its luxury image by partnering with celebrities like Ariana Grande and Venus Williams, as well as incor
中文摘要 施华洛世奇首席执行官Alexis Nasard接受彭博专访，介绍这家拥有131年历史的品牌正进行重大转型，以「现代奢华」为核心战略目标，通过与爱莉安娜·格兰德、维纳斯·威廉姆斯等明星合作，重塑品牌奢侈品形象，推动公司重返盈利轨道。
Pakistan said an interim US-Iran deal to reopen the Strait of Hormuz could be finalized within 24 hours, raising expectations that the sides may be nearing a broader agreement after recent skirmishes near the strategic waterway. Bloomberg News State Department Reporter Eric Martin and Jerusalem Repo
中文摘要 巴基斯坦方面表示，美伊关于重开霍尔木兹海峡的临时协议最快可在24小时内敲定。尽管双方近期在该战略水道附近发生小规模冲突，但谈判进展令外界对两国达成更广泛协议的预期明显升温。彭博国务院记者参与了相关报道。
The FIFA World Cup is expected to create a "hospitable environment" for pathogens, with millions of people gathering in packed stadiums, and there are concerns about the spread of diseases such as Ebola, dengue, and measles. Johns Hopkins Center for Health Security Senior Scholar Dr. Amesh Adalja jo
中文摘要 世界杯开幕之际，公共卫生官员发出警告，数百万球迷聚集于密集场馆，将为病原体传播提供「有利环境」。当前主要关注麻疹疫情扩散风险，此外埃博拉病毒及登革热同样令人担忧。约翰斯·霍普金斯大学健康安全中心资深学者就此发表评估意见。
SpaceX’s first day on the stock market transformed the startup into one of the world’s most-valuable public companies, handed buyers of the IPO a 19% return and turned its founder Elon Musk into the world’s first trillionaire. Bloomberg Tech Co-Host Ed Ludlow joined David Gura and Christina Ruffini
中文摘要 SpaceX完成750亿美元规模的历史性IPO，首日股价收涨19%，公司由此跻身全球市值最高上市公司之列。创始人埃隆·马斯克个人财富随之突破万亿美元大关，成为全球首位万亿富翁。彭博科技联合主持人Ed Ludlow就此进行了详细报道。
The firm’s co-founder talks about the celebrity buzz that sparked her interest in prediction markets, the legal battles her firm faces and Jay-Z’s advice.
中文摘要 预测市场平台Kalshi联合创始人Lopes Lara在专访中讲述，名人八卦话题如何激发其对预测市场的兴趣，进而将公司发展为估值数十亿美元的机构。她还就公司面临的法律纠纷以及说唱歌手Jay-Z给予的创业建议进行了分享。
The US’s best weapon against a deadly cattle parasite threatening the beef industry is more than a year away from showing meaningful results, raising concerns over how far the outbreak could spread before then.
中文摘要 美国对抗致命牛只寄生虫——螺旋蝇的最有效防治手段距离取得显著成效仍需逾一年时间。在此期间，疫情扩散风险持续上升，牛肉产业面临较大威胁，相关防控措施能否及时遏制蔓延态势引发广泛担忧。
Soaring prices for the world’s largest beef producer, Brazil, mean households will be buying less red meat when they gather to watch Brazil’s first game of the FIFA World Cup this Saturday.
中文摘要 作为全球最大牛肉生产国，巴西国内牛肉价格大幅攀升，导致民众在世界杯首场比赛观赛聚会中减少红肉消费。传统烤肉文化受到冲击，鸡肉和猪肉正逐渐替代牛肉，成为球迷烧烤桌上的主角。
The long-awaited SpaceX IPO tested the limits of modern capital markets, and is expected to be followed by OpenAI and Anthropic. Investor Steve Rattner says the scale of the valuations and fundraising involved is unprecedented, simultaneously reflecting optimism about artificial intelligence and a b
中文摘要 SpaceX完成备受瞩目的IPO后，OpenAI及Anthropic上市预期随之升温。投资人Steve Rattner指出，科技公司万亿美元级估值及融资规模史无前例，既折射出市场对人工智能前景的高度乐观，同时也令资本市场承压，其合理性受到广泛审视。
5 回复 · 程序员 节点
8 回复 · 程序员 节点
15 回复 · 程序员 节点
11 回复 · Apple 节点
12 回复 · 程序员 节点
7 回复 · 程序员 节点
61 回复 · Apple 节点
6 回复 · Apple 节点
23 回复 · Apple 节点
8 回复 · Apple 节点
10 个帖子 - 10 位参与者 阅读完整话题
我的ai图生视频公益站l0veyou昨天数据库被删了，源码也被删了，我真服了，不知道谁跟我这么大仇，我一直在修复，从现在开始免费24小时，如果积分还没补偿到位，那就延期免费。 49 个帖子 - 47 位参与者 阅读完整话题
6 月 13 日，深度求索研究员陈德礼在社交平台 X 上发帖，强烈投诉北京字节跳动位于蓝景丽家的施工项目。他称，周末凌晨 2 点，挖掘机仍在工地轰鸣作业，距其住所仅 200 至 300 米，严重干扰居民休息。 他表示，此前已通过 12345 热线反映问题，官方回复称该项目为“市级重点基础设施工程”，法定施工时间为早 6 点至晚 10 点，含节假日。但如今工地凌晨违规施工，令他无法接受。他拍摄视频上传至国内社交平台小红书，随后账号被禁言，视频对他人不可见。他怀疑是北京方面看到后选择“封口”而非解决问题。目前其 12345 投诉仍未得到答复。 https://x.com/victor2077558
由于测试的模型越积越多了，表格会删除一些同厂商的旧模型，你可以在之前的评测帖子里找到它们的成绩。 项目 这是一个 Unity C# 项目，我进行测试的是一份皮肤系统需求案，我已经做了好预制体，而模型需要编写代码。 本轮与上两轮评测的项目和环境都完全一致： 第一轮 … 上一轮 模型来源 GLM 5.2: 官方 Coding Plan 速度 排名 模型 时间（分钟） 备注 1 Composer 2.5 3 2 Grok 4.20 0309 Reasoning 3 3 Step-3.5-Flash 6 4 Mimo V2 Omni 7 5 Doubao-Seed-2.0-Lite 7 6 Douba
官网 https://ai.centos.hk 官方风控不大变动的话，这个价格会维持，如果风控加紧，会略微涨价 愿我们所有人 烟火向星辰 所愿皆成真～ 用前须知 https://doc.centos.hk 加入星辰AI售后群：QQ群 14 个帖子 - 12 位参与者 阅读完整话题
本帖使用社区公益推广，符合推广要求。我申明并遵循社区要求的以下内容： 我的项目是免费使用的，无收费（变相收费、赞助）部分： 是 我的帖子已经打上 公益推广 标签： 是 我的项目属于个人项目，与公司或商业机构无关： 是 我的项目不存在QQ、TG等群组引流： 是 我的项目不存在非运营必要的网站引流： 是 我的项目不存在为他人推广、AFF： 是 我的项目无关联的商业项目： 是 我的站点存在登录，并已接入 LINUX DO Connect： 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
老规矩私有bench 案例都很不错 第三个在这个案例中做热力学图的模型，前两个是Mythos和3.5Flash 正如知乎nao榜所言，日后通过中转贩子使用opus的人，都需要面对一个问题，你用的opus如果是glm5.2冒充的，而且难以分辨。 在实际bot agent体验上，如果不是对opus4.6特别熟悉的人基本无法分辨出两者。并且其追随上文的能力很强。如果上文用的opus，继续用glm5.2根本无法分别。 其缺点目前来看，上下文注意力可能不如4.6强（说实在的比4.6强的也几乎没有）。这次上1M上下文盲猜的DS V4的技术落地，DSA的注意力只能说目前来看中规中矩而已。 不过真的恭喜智谱啊
GLM5.2(思考：max)天气卡片测试： 以 iOS 18 的设计风格做一个带有动画效果的天气卡片，要求是使用 HTML、CSS 和基础 JavaScript，使用横板天气页面（拥有 4 个天气卡片 (晴天，大风，暴雨，暴雪))。应足够美观，实现一定的交互效果。 25 个帖子 - 22 位参与者 阅读完整话题
就剩2个PLUS独苗了 ） 2个PRO号的周用量 25 个帖子 - 23 位参与者 阅读完整话题
中美扯头发这么多年，对我们高科技的封锁早就不是新鲜事了，这些封锁无一不在帮助我们国产替代的发展，而且这一次有些不一样。 1.国模的coding能力比半年前已经好很多的，现在最多算不够好用 44 个帖子 - 37 位参与者 阅读完整话题