Haskell · ★ 15,022 · 🍴 865 · 📈 1,180 stars today
SimpleX - the first messaging network operating without user identifiers of any kind - 100% private by design! iOS, Android and desktop apps 📱!
中文介绍 SimpleX 是首个无需用户标识（电话、用户名等）的去中心化消息网络，默认 100% 保护隐私。支持 iOS、Android 和桌面端。适合对通信隐私有极致要求的用户，完全匿名且无数据泄露风险。
HTML · ★ 125,240 · 🍴 13,168 · 📈 495 stars today
A list of SaaS, PaaS and IaaS offerings that have free tiers of interest to devops and infradev
中文介绍 一份面向开发者和运维人员的免费服务清单，收录了各种有免费层的 SaaS、PaaS 和 IaaS 产品。帮助个人或小团队快速找到合适的免费数据库、CI/CD、监控、域名等工具，降低基础设施成本。
Python · ★ 62,390 · 🍴 11,095 · 📈 266 stars today
openpilot is an operating system for robotics. Currently, it upgrades the driver assistance system on 300+ supported cars.
中文介绍 openpilot 是一个机器人操作系统，当前版本可将超过 300 款汽车的驾驶辅助系统升级为更高级的自动驾驶功能。适合汽车爱好者和开发者，用于改装车辆，提升辅助驾驶能力。
Python · ★ 5,312 · 🍴 720 · 📈 1,445 stars today
AI 时代的伯克希尔：基于 Claude Code / Codex 的价值投资研究框架。巴菲特·芒格·段永平·李录四大师方法论 + 多Agent并行研究。| AI-era Berkshire: a value investing research framework built for Claude Code / Codex. 4 masters' methodologies + multi-agent adversarial analysis.
中文介绍 基于 Claude Code / Codex 构建的价值投资研究框架，融合巴菲特、芒格、段永平、李录四位大师的方法论，支持多 Agent 并行研究。适用于 AI 辅助投资分析，自动生成投资决策建议。
Python · ★ 8,229 · 🍴 804 · 📈 372 stars today
A feed-forward 3D foundation model for reconstructing scenes from streaming data
中文介绍 LingBot-Map 是一个前馈式 3D 基础模型，能从流数据（如视频或连续传感器输入）中实时重建三维场景。适用于机器人、AR/VR 等领域，提供高效的环境感知与建模能力。
C · ★ 19,666 · 🍴 1,425 · 📈 2,190 stars today
High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.
中文介绍 高性能代码智能 MCP 服务器，将代码库索引为持久知识图谱，支持 158 种语言，查询亚毫秒级，相比传统方法减少 99% 的 token 消耗。单静态二进制部署，适用于大型代码库的快速检索与问答。
Python · ★ 11,521 · 🍴 1,069 · 📈 174 stars today
NumPy & SciPy for GPU
中文介绍 CuPy 是 NumPy 和 SciPy 的 GPU 加速替代库，API 兼容，自动将数组运算迁移到 GPU 执行，大幅提升科学计算和机器学习的速度。适合需要高性能数值计算的数据科学家和工程师。
Swift · ★ 3,735 · 🍴 238 · 📈 365 stars today
FluidVoice - Fastest macOS Offline Dictation app - Voice to Text fully Local. One ⭐ takes us a long way :))
中文介绍 FluidVoice 是 macOS 上最快的离线听写应用，语音转文字完全在本地处理，保障隐私且低延迟。适用于无网络环境下快速输入文字，适合注重隐私的 Mac 用户。
Python · ★ 71,596 · 🍴 6,014 · 📈 380 stars today
Transforms complex documents like PDFs and Office docs into LLM-ready markdown/JSON for your Agentic workflows.
中文介绍 MinerU 可将复杂的 PDF、Office 等文档转换为结构化的 Markdown 或 JSON 格式，便于 LLM 直接处理。适用于构建文档解析管道，为 AI Agent 工作流提供干净的数据输入。
Python · ★ 14,321 · 🍴 2,632 · 📈 492 stars today
"Vibe-Trading: Your Personal Trading Agent"
中文介绍 Vibe-Trading 是一个个人交易代理，利用 AI 技术辅助股票或加密货币交易决策，可能结合市场情绪分析。适合希望自动化或智能化交易的个人投资者，降低情绪干扰。
★ 84,458 · 🍴 9,350 · 📈 250 stars today
Explain complex systems using visuals and simple terms. Help you prepare for system design interviews.
中文介绍 用直观的图解和简单语言解释复杂系统设计概念，帮助开发者准备系统设计面试。覆盖分布式系统、数据库、一致性等常见主题，适合后端工程师和面试者快速掌握核心知识。
Python · ★ 26,735 · 🍴 2,985 · 📈 122 stars today
Open-source AI hackers to find and fix your app’s vulnerabilities.
中文介绍 Strix 是一个开源 AI 驱动的安全测试工具，模拟黑客攻击自动发现应用漏洞并辅助修复。适用于开发和安全团队，可集成到 CI/CD 管道中，持续检测应用安全性。
Python · ★ 11,055 · 🍴 1,524 · 📈 196 stars today
Edit videos with coding agents
中文介绍 Video-Use 允许通过编程代理（AI Agent）自动编辑视频，实现剪辑、转场、特效等操作。适用于需要批量处理或自动化视频编辑的场景，如内容创作和社交媒体管理。
👍 105
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
中文介绍 探讨LLM智能体记忆系统从检索增强到原生数据管理系统的演进，实现持久存储、检索、更新、整合与动态生命周期治理，为构建智能体原生记忆奠定基础。
👍 50
Modern Vision-Language-Action (VLA) models often fail to generalize to novel setups, such as altered camera viewpoints or robot morphologies, because they are typically conditioned only on current observations and language instructions. By ignoring the underlying system configuration as a variable,
中文介绍 提出上下文世界建模方法，将系统配置作为变量纳入Vision-Language-Action模型，使机器人控制能泛化到新颖视角与形态，无需重新训练。
👍 46
Outcome-based reinforcement learning provides a stable optimization backbone for language agents, but its sparse trajectory-level rewards provide little guidance on which intermediate decisions should be reinforced or suppressed. On-policy self-distillation offers dense token-level supervision, yet
中文介绍 提出OPID同策略技能蒸馏，结合结果强化学习与自蒸馏，在稀疏轨迹奖励中注入密集token级监督，提升语言智能体训练稳定性与性能。
👍 45
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
中文介绍 提出ShutterMuse，利用多模态大语言模型在拍摄时提供构图与主体姿态指导，弥补现有基准仅评估后期裁剪的不足，验证MLLM在实时摄影引导中的能力。
👍 42
While text-to-image (T2I) models have achieved remarkable progress, they struggle with real-world requests that are often underspecified, implicit, or dependent on up-to-date knowledge. We identify this challenge as the Context Gap: the mismatch between the user context and the sufficient generation
中文介绍 提出Qwen-Image-Agent，通过智能体框架弥合用户上下文与生成条件之间的差距，处理隐含、依赖最新知识的真实图像生成请求，提升T2I模型在复杂场景下的表现。
👍 41
A classical intuition holds that verifying a solution is easier than producing one. For today's coding agents, this intuition is being inverted: as foundation models develop stronger reasoning capabilities and engineering harnesses grow more sophisticated, generating complex candidate solutions is n
中文介绍 揭示编程智能体面临验证瓶颈：生成复杂代码越来越容易，但验证正确性反而更难，分析现有奖励机制局限，指出无银弹解决方案。
👍 38
A unified representation for text and vision is a natural pursuit, as it enables simpler multimodal modeling and more efficient training. However, representing images as discrete signals in the same way as text inevitably introduces severe information loss. Existing work struggles to balance low-lev
中文介绍 提出ViQ，实现文本对齐的任意分辨率视觉量化表示，在保持低层次细节的同时对齐语义，平衡信息损失与多模态建模效率。
👍 34
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
中文介绍 提出MVTrack4Gen，利用多视角点跟踪提供几何监督，从单目视频生成新视角4D视频，无需显式3D重建，保持几何一致性与运动保真度。
👍 31
Speculative decoding (SD) accelerates autoregressive Large Language Models (LLMs) by drafting multiple tokens and verifying them in parallel, but it faces a scaling limitation: increasing the draft budget improves speed only when acceptance remains high and drafting overhead stays low. This ceiling
中文介绍 提出JetSpec，采用并行树草稿策略突破推测解码的扩展天花板，在保持高接受率与低开销的同时大幅提升生成速度。
👍 28
Computer-use agents can execute software tasks through either graphical interfaces or programmatic command interfaces, but existing evaluations confound interaction modality with differences in tasks, initial states, verifiers, and permitted actions. We introduce a matched execution-layer benchmark
中文介绍 构建匹配执行层基准，系统比较图形界面与命令行界面在计算机使用智能体中的执行瓶颈，揭示不同交互模态的优劣与适用场景。
👍 26
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
中文介绍 提出V-Zero，无需答案标签，通过同策略蒸馏与对比证据门控，使多模态大语言模型在细粒度视觉推理中定位局部图像证据，提升推理可解释性与准确性。
👍 24
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
中文介绍 提出UnityShots，基于边界感知门控的记忆驱动方案，实现多镜头音频-视频连贯生成，保持主题外观、场景上下文与说话者身份在镜头切换中的一致性。
👍 23
Joint-Embedding Predictive Architectures (JEPAs), including recent LeWorldModel (LeWM), have become a promising foundation for reconstruction-free visual world models. For visual planning, however, LeWM evaluates candidate action sequences by repeatedly applying a local one-step latent transition mo
中文介绍 提出Fast LeWorldModel，加速联合嵌入预测架构中的视觉规划，通过并行化单步潜变量转移评估，提升决策效率而不牺牲重建质量。
👍 17
As agentic systems continue to evolve and are widely deployed in real-world scenarios, there is a growing demand to faithfully evaluate their capabilities. However, current benchmarks are typically built on popular applications with relatively simple tasks and focus on a narrow set of capabilities w
中文介绍 构建严苛评测基准，全面评估智能体在陌生环境中的能力，覆盖复杂任务与多维度指标，揭示现有系统泛化短板。
👍 16
Tool use enables large language models (LLMs) to perform complex tasks, and recent agentic reinforcement learning (RL) methods show promise for enhancing model capabilities. However, RL alone often leads to instability or limited gains in tool-use tasks. In our experiments, some models exhibit catas
中文介绍 分析多步工具使用强化学习崩溃原因（稀疏奖励、策略退化），提出引入监督信号稳定训练，避免灾难性遗忘，提升LLM工具使用能力。
👍 13
The prevalent dual-branch paradigm, i.e., training a side network to encode visual conditions and fusing its intermediate-layer features to a frozen pretrained main network, has shown remarkable success in visual-condition controllable generation. Despite its widespread adoption, the role of the sid
中文介绍 提出LISA似然分数对齐，重新审视双分支可控生成中侧网络的作用，通过对齐主网络与侧网络的似然分布，提升视觉条件生成效果，无需复杂训练。
👍 13
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
中文介绍 提出Autodata方法，训练AI智能体扮演数据科学家自主生成高质量合成训练与评估数据，通过元优化使智能体学会创造更强数据。
👍 9
Reasoning capability has advanced rapidly in large language models (LLMs), leading to an increasing size of key-value (KV) cache in both prefilling and decoding stages. Existing KV cache compression methods mainly rely on attention weights to estimate token importance. While attention effectively ca
中文介绍 提出信息感知KV缓存压缩方法，不仅依赖注意力权重，还结合信息论指标评估token重要性，在长推理场景中高效压缩缓存，降低显存占用而不损失性能。
👍 9
Video reasoning language models implicitly assume that every input frame is equally reliable. This leads to what we term the Blind Trust Problem: under realistic perturbations such as motion blur, glare, or occlusion, frontier video reasoning models can suffer 15-30%p accuracy drops on real-world em
中文介绍 提出置信度感知工具编排，解决视频理解中的盲目信任问题：在运动模糊、遮挡等扰动下动态选择可靠帧与工具，恢复15-30%的准确率下降，提升鲁棒性。
👍 9
We present PhysiFormer, a diffusion transformer for physically-plausible 3D object motion. Unlike video world models that operate in view-dependent pixel space, PhysiFormer represents objects as 3D meshes expressed in world coordinates. Given the initial vertex positions and velocities, as well as o
中文介绍 提出PhysiFormer，基于扩散Transformer的世界空间物理模拟器，以3D网格表示物体，给定初始顶点位置、速度及物理参数，生成符合力学规律的3D运动轨迹。
@sairahul1 · 121.4K 粉丝 · 2.9M 阅 · 506 赞 · 97 转
Most people use Claude Code like a fancy autocomplete. They prompt. They wait. They accept the first output. They are leaving 90% of the power on the table. There are 27 specific moves that separate
中文介绍 27个不为人知的Claude隐藏功能、设置和快捷键，帮用户突破高级自动补全的局限，发挥Claude Code 90%的潜力。
@sairahul1 · 121.4K 粉丝 · 876.6K 阅 · 547 赞 · 83 转
Emails. Formatting. Compiling reports. Preparing decks. Organizing files. Researching. Marketing. Writing. SEO. The average knowledge worker spends 60% of their day on above work that doesn't require
中文介绍 教程：用Claude Cowork自动化邮件、报告、营销等低价值工作，释放60%工作时间，构建单人公司高效工作流。
@crux_capital_ · 52.8K 粉丝 · 217.5K 阅 · 503 赞 · 56 转
This is the most in depth report I have ever written, on a very exciting company. If you read this article you will have a much deeper understanding of what Ouster does, how it plays into Physical AI,
中文介绍 深度分析报告：Ouster在物理AI（激光雷达）领域的定位，作者认为这是最令人兴奋的公司之一，内容极其详尽。
@Oracle__Market · 5.9K 粉丝 · 100.0K 阅 · 2.1K 赞 · 823 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 2026年预测市场Polymarket上超30%活动来自算法和AI钱包，官方推出AI代理交易策略，强调自主代理的有效性。
@Oracle_Market__ · 25.1K 粉丝 · 99.3K 阅 · 2.8K 赞 · 224 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 2026年Polymarket上超30%活动来自算法和AI钱包，官方推出AI代理交易策略，强调自主代理的有效性。
@base · 1.3M 粉丝 · 93.3K 阅 · 501 赞 · 127 转
News ■ Base MCP's new skills let agents transact, trade, lend, mint, and buy onchain across 13 apps: @yield @AskVenice @KyberNetwork @opensea @o1_exchange @Balancer @printr @bitrefill @flaunchgg
中文介绍 Base MCP新增13个应用技能（交易、借贷、铸币等），现已集成到Perplexity，桌面版Base App也同步上线。
@cyrilXBT · 186.6K 粉丝 · 77.2K 阅 · 510 赞 · 93 转
There is a specific moment that tells you whether someone is running a real one person company or just using a lot of apps. It is the moment they get asked a question about their own business and they
中文介绍 Hermes、Obsidian与Claude Code组合实现单人公司知识和自动化，重点在于无需搜索即可回答业务问题，真正高效。
@RhysSullivan · 57.4K 粉丝 · 48.9K 阅 · 501 赞 · 23 转
i want to use the skills, knowledge, and apis your company has spent years developing, not your custom agent almost every company by this point has shipped an agent there's a cloudflare agent in their
中文介绍 观点：用户不需要公司定制agent，而是希望直接调用公司积累的技能、知识和API，反对为上线而上线。
@0xDevin_ · 6.6K 粉丝 · 38.3K 阅 · 539 赞 · 5 转
Most AI tools are chatbots with a nice interface. Kimi is different. It is a full system: a browser automation engine called Claw that navigates websites like a human, an Agent Swarm that runs
中文介绍 Kimi不仅是聊天界面，而是包含浏览器自动化Claw和Agent Swarm的完整系统，与普通AI工具不同。
@akshay_pachaar · 279.5K 粉丝 · 35.5K 阅 · 505 赞 · 54 转
Build by Google, explained as a step-by-step guide. Karpathy defined agentic engineering at Sequoia Ascent 2026 as the discipline that separates production-grade agent work from vibe coding. The core
中文介绍 Karpathy在Sequoia Ascent 2026定义的代理工程概念，Google发布配套工具，将生产级代理工作与vibe coding区分。
@ericzakariasson · 76.0K 粉丝 · 32.3K 阅 · 518 赞 · 29 转
What I like most about coding with agents right now is the room to leave a few runs going and still get on with other work. When something finishes or needs a call, I show up. This post is a short
中文介绍 用agent编码的体验：同时跑多个agent任务，完成后或需人工干预时介入，保持高效，且不耽误其他工作。
HP Inc. scales its OpenAI Frontier partnership to deploy AI across customer experiences, software development, and enterprise operations.
中文介绍 惠普公司与OpenAI扩大Frontier战略合作，将人工智能部署于客户体验、软件开发和企业运营中。
Oddly tiered releases to both OAI and ANT on the same day.
中文介绍 OpenAI发布GPT-5.6三个变体Sol、Terra和Luna，限制仅向受信任合作伙伴开放，同日还向OAI和ANT进行了分层发布。
OpenAI previews GPT-5.6 Sol, a next-generation model with stronger capabilities in coding, science, and cybersecurity, paired with its most advanced safety stack.
中文介绍 OpenAI预览下一代模型GPT-5.6 Sol，在编程、科学和网络安全方面能力更强，并搭载最先进的安全堆栈。
**OpenAI** previewed **GPT-5.6** with three variants: **Sol** (flagship), **Terra** (mid-tier), and **Luna** (lower-cost), launching under a restricted rollout mandated by the U.S. government, limiting access to trusted partners. **Sol** boasts enhanced cybersecurity and safety features backed by ov
中文介绍 OpenAI预览GPT-5.6的三个变体：Sol（旗舰）、Terra（中端）和Luna（低成本），受美国政府限制仅向受信任合作伙伴推出，Sol增强了网络安全与安全性能。
It's happening.
中文介绍 OpenAI报告自2025年11月以来，内部Codex输出令牌中位数在研发部门增长56倍，客户支持32倍，工程27倍，法律13倍。
中文介绍 Hugging Face推出新功能：只需一条命令即可在HF Jobs上运行vLLM服务器，简化部署流程。
中文介绍 报道要点包括美国与OpenAI的法律纠纷、人工智能经济现状及缩放定律进展。
中文介绍 艾伦人工智能研究所探讨混合模型在哪些令牌预测上表现更优，分析不同令牌类型的预测差异。
Artificial intelligence is rapidly reshaping retail, but not in the ways consumers might immediately notice. The biggest transformation may not be flashy virtual try-ons or chatbot shopping assistants, but in how decisions are made behind the scenes: how products surface in search results, how inven
中文介绍 人工智能正迅速重塑零售业，最大变革并非炫酷的虚拟试穿或聊天助手，而是后台决策方式：产品如何出现在搜索结果中。
**Z.ai's GLM-5.2** leads in coding and agent benchmarks with top scores like **1595** on Code Arena: Frontend and **34.29%** reasoning accuracy with zero failures. Databricks improved GLM-5.2 speed to **392 tok/s** using hardware and optimizations. **Ornith-1.0**, a new MIT-licensed coding model fam
中文介绍 Z.ai的GLM-5.2在编程和智能体基准测试中领先，Code Arena前端得分1595，推理准确率34.29%且零失败；Databricks将其速度提升至392 tok/s；新模型Ornith-1.0以MIT许可发布。
Move over, Harness Engineering, it is time for the harness of harnesses!
中文介绍 Meta公司迎来“Harness夏季”，强调对Harness工程领域的进一步整合与优化。
A new OpenAI research paper shows how AI agents are transforming work, enabling longer, more complex tasks and expanding productivity across roles.
中文介绍 OpenAI新研究论文显示，人工智能智能体正在改变工作方式，使任务更复杂、更持久，并提升各角色的生产力。
中文介绍 今日热点：Anthropic指控阿里巴巴侵权；谷歌Gemini将推出计算机使用功能；另有Jalapeño芯片相关新闻。
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技术领袖Matei Zaharia和Reynold Xin在访谈中强调前沿生态系统必须开放，并探讨每家公司构建Agent Cloud的必要条件。
中文介绍 谷歌DeepMind在Gemini 3.5 Flash中引入计算机使用功能，使模型能直接操控桌面应用。
本日技术面呈现资产分化格局。美股方面，SPY、QQQ等指数ETF均跌破20日均线，RSI位于43-46区间中性偏弱，但QQQ仍保持多头排列；大型科技股多数回调，MSFT、TSLA等呈空头排列且MACD负值扩大。加密市场延续弱势，恐慌贪婪指数仅12（极度恐慌），总市值2.14T（24h -0.77%），BTC主导率55.7%、ETH占8.8%；BTC-USD RSI 30.6接近超卖，ETH-USD RSI 31.5，均处空头排列且MACD死叉，下行压力显著。中概股整体承压，BABA、JD双双进入RSI超卖区（16.5、26.8），价格大幅低于各均线，空头趋势明确。商品外汇方面，黄金期货跌破20日均线且RSI 36.8偏弱；WTI原油RSI 28.6超卖，但价格仍低于多数均线；美元指数DXY强势，RSI 70.8超买且均线多头排列，与VIX指数（RSI 51.1中性）及10Y美债收益率（RSI 40.1偏弱）形成反差。整体技术面，权益与加密资产偏弱，美元走强，大宗商品回调压力明显。
美元指数当前101.34，RSI(14)为70.8进入超买区，价格站上所有均线（SMA20 100.26、SMA50 99.28、SMA200 98.79），MACD(0.6173)在信号线(0.4972)上方，均线多头排列，技术形态显示强劲上行趋势，但超买状态需注意回调风险。
阿里巴巴当前94.81，RSI(14)仅16.5处于深度超卖区域，价格远低于SMA20(113.53)、SMA50(126.30)及SMA200(148.22)，MACD(-8.53)在信号线(-6.80)下方，空头排列明显，短期技术面非常弱势，下行动能仍在。
Solana当前71.44，RSI(14)48.8处于中性区间，价格略高于SMA20(69.84)但低于SMA50(77.22)和SMA200(95.14)，MACD(-1.32)位于信号线(-1.91)上方，MACD柱转正，但整体仍处于长期空头趋势，短期多空交织，技术信号中性。
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
过去走势不代表未来表现，本报告仅基于公开行情数据的技术指标解读，不构成任何投资建议。所有结论供参考，投资者应结合自身风险承受能力独立判断。
With tens of thousands of people missing, relatives face another night waiting for news of loved ones as the crucial window for locating survivors closes.
中文摘要 委内瑞拉地震救援中，两名男孩从废墟中被救出，周末共救出33人。数万人仍失踪，家属等待亲人消息，搜救黄金窗口正在关闭。
In a fourth day of hostilities, Iran said that it had targeted a U.S. naval base in Bahrain and a Kuwaiti air base with drones and missiles. No major damage or casualties were reported.
中文摘要 美伊在霍尔木兹海峡争端升级，进入第四天敌对行动。伊朗称用无人机和导弹打击了巴林的美军基地和科威特空军基地，未造成重大损失或伤亡。
中文摘要 暂无详细信息。
More than three days after twin earthquakes devastated the Venezuelan coast, the rescue effort faced chaos and delays as the chances of finding survivors diminished.
中文摘要 委内瑞拉沿海发生双地震已超三天，救援工作面临混乱与延迟，找到幸存者的几率越来越小，救援人员称处于“关键小时”。
Follow today’s news live Get our breaking news email, free app or daily news podcast Royal commission hearings resume today The next round of hearings for the antisemitism royal commission begins today. Regretfully, we’re here until Thursday night with Parliament and the game is at 4am Saturday morn
中文摘要 澳大利亚政治动态：前财长泰勒称联盟党“失去公众信任”，民调持续下滑；独立议员瑞安推动解决高等教育债务法案；反犹太主义皇家委员会听证会恢复。
Eustaquio flip-flopped between his home country and that of his parents throughout his youth and senior career.
中文摘要 斯蒂芬·欧斯塔基奥在加拿大对阵南非的比赛中进球。他曾在祖国与父母祖国之间摇摆不定，最终代表加拿大出战。
A mass return of Malawians from South Africa exposes the cost of migration, violence and broken livelihoods.
中文摘要 马拉维人大量从南非返回，暴露了移民、暴力与生计破碎的代价。一名马拉维人称“如果要死，就死在这里”。
Washington and Tehran have agreed to stop attacks and renew talks in the Qatari capital, Doha, Axios reports.
中文摘要 据Axios报道，美国与伊朗已同意停止攻击并在卡塔尔首都多哈重启谈判。德黑兰坚持对霍尔木兹海峡的控制权。
中文摘要 暂无详细信息。
Volunteers in a middle-class neighborhood in Caracas used drills, picks and hammers to break through concrete, trying to find anyone in need of rescue.
中文摘要 在加拉加斯一个中产阶级社区，志愿者使用电钻、镐和锤子凿穿混凝土，共同寻找地震幸存者。
Identifying victims has proved to be difficult because many bodies were badly crushed beneath collapsed buildings.
中文摘要 在加拉加斯太平间，家属与官员试图辨认超过100名遇难者。因许多遗体被倒塌建筑严重压碎，辨认工作困难。
Russia’s president says Ukraine’s attacks on infrastructure are causing ‘obvious’ but not critical problems The Russian president, Vladimir Putin, acknowledged that the country was suffering from “a certain shortage” of fuel in an interview published by the Kremlin on Sunday, after repeated Ukrainia
中文摘要 俄罗斯总统普京承认，乌克兰对基础设施的袭击正导致俄罗斯出现“明显的”燃料短缺，但称问题不致命。
Iran sees its control over the Strait of Hormuz as critical leverage in peace talks with the United States. It seems willing to risk the cease-fire to maintain that power.
中文摘要 伊朗视其对霍尔木兹海峡的控制为与美国和谈的关键筹码，并似乎愿意冒着破坏停火的风险来维持这一影响力。
A Colombian rescue team worked for six hours to recover the child, Moises, from under nearly 10 feet of rubble in La Guaira. His rescue was captured on video.
中文摘要 哥伦比亚救援队在拉瓜伊拉花费六小时，从近10英尺（约3米）的废墟下救出11岁男孩莫伊塞斯。救援过程被视频记录。
Ireland won the second T20 by one run over India, who did not give a debut to teen sensation Sooryavanshi.
中文摘要 爱尔兰在主场T20系列赛中爆冷击败世界冠军印度队，第二场以1分险胜，印度未让少年天才苏里亚瓦什首秀。
Australian central bank Assistant Governor Chris Kent said the institution will be better prepared to respond to the next crisis it faces following a review of alternative monetary policy tools. In a speech in Sydney, Kent, who oversees financial markets at the Reserve Bank, said the cash rate targe
中文摘要 澳大利亚央行助理行长克里斯·肯特在悉尼演讲中表示，经过对替代性货币政策工具的评估，央行将能更充分地准备应对下一次危机。
The US and Iran have agreed to stop attacking each other before peace talks resume this week over the Strait of Hormuz and other issues, paving the way to end days of tit-for-tat attacks that tested a fragile truce. Bloomberg's Wendy Benjaminson breaks down the latest developments. (Source: Bloomber
中文摘要 据彭博社报道，美国和伊朗同意在本周恢复和平谈判前停止相互攻击，谈判涉及霍尔木兹海峡等问题，旨在结束数日以来的紧张局势。
The Philippines has cut its economic growth targets and sees a weaker peso beyond the end of the term of President Ferdinand Marcos Jr. in 2028 amid headwinds such as the Middle East tensions and an intense El Niño weather event.
中文摘要 菲律宾下调经济增长目标，并预计比索在总统马科斯2028年任期结束后仍将疲软，主要受中东紧张局势及强厄尔尼诺天气等不利因素影响。
China faces greater uncertainty in forecasting energy demand as structural changes in the economy and the rapid expansion of new industries reshape consumption patterns, according to a top government official. (Source: Bloomberg)
中文摘要 中国高级官员表示，经济结构变化与新兴产业快速扩张重塑消费模式，导致能源需求预测面临更大不确定性。
Gold declined to near $4,000 an ounce after the US and Iran traded attacks in the Persian Gulf, straining a ceasefire that had last week seen energy prices fall to pre-war levels and tempered expectations for an interest-rate hike.
中文摘要 黄金价格跌至近4000美元/盎司，因美伊在波斯湾相互攻击引发通胀担忧；此前的停火曾使能源价格回落并减弱加息预期。
Australian central bank Assistant Governor Chris Kent said the institution will be better prepared to respond to the next crisis it faces following a review of alternative monetary policy tools.
中文摘要 澳大利亚央行助理行长克里斯·肯特表示，经过对替代性货币政策工具的审查，央行将能更好地准备应对下一次危机。
As the world's best players begin play at Wimbledon, how can you get into sport on a budget.
中文摘要 温布尔登网球赛开赛之际，文章探讨如何在预算有限的情况下参与网球等运动，提供低成本运动建议。
High concentration in stock markets and national security concerns send SWFs to private credit and infrastructure
中文摘要 由于股票市场高度集中及国家安全担忧，主权财富基金从公开市场转向私人信贷和基础设施，以抓住人工智能发展机遇。
The world’s biggest public investors plan to shift more capital into private and less liquid assets amid rising risks to their traditional bond-and-stock portfolios, according to an industry survey.
中文摘要 行业调查显示，全球最大公共投资者计划将更多资本转向私人和低流动性资产，以应对传统股债组合面临的日益上升的风险。
Wall Street banks are capitulating on bets for a stronger euro, as markets see the US outpacing Europe on interest-rate hikes for the rest of this year.
中文摘要 华尔街银行放弃押注欧元走强，因市场预计美国今年余下时间加息步伐将快于欧洲。
The three biggest foreign banks in Indonesia have shipped around $640 million of their earnings out of Southeast Asia’s largest economy since 2024 as they pare exposure amid President Prabowo Subianto’s increasingly state-focused economic policies.
中文摘要 由于总统普拉博沃日益以国家为中心的经济政策，花旗、汇丰和渣打等三大外资银行自2024年以来已从印尼汇出约6.4亿美元利润，以降低风险敞口。
US equity-index futures rose after reports the US and Iran backed away from a fresh escalation of their conflict, easing concerns over the fragile ceasefire underpinning peace talks.
中文摘要 美国股指期货上涨，油价涨幅收窄，因报道称美伊避免冲突进一步升级，缓解了和平谈判脆弱停火的担忧。
6 回复 · 程序员 节点
8 回复 · 程序员 节点
18 回复 · 程序员 节点
13 回复 · Apple 节点
11 回复 · Apple 节点
17 回复 · 程序员 节点
15 回复 · Apple 节点
9 回复 · 程序员 节点
41 回复 · Apple 节点
14 回复 · Apple 节点
原谅我这雷霆标题 对，我是说我的ipad…… 引以为戒！ 34 个帖子 - 32 位参与者 阅读完整话题
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 两个月前有点想当当up主，但是又不想真人出镜，就想试试实时数字人或者数字人口播视频生成。尝试了几个软件，发现都收费，但是其实他们用的很多都是开源的模型。ASR\TTS模型也特别多，分散去尝试也很累。于是我和几个小伙伴就
始于260628 始于18点15左右~摸鱼好耶~ 背景 除了粉猫会自动把缓冲完的视频保存到相册外，我只会主动下载在 tg 看到的一切二次元涩涩(一般是大小远低于原画的 720Pw 因为原画我4mb每秒也不能秒开，况且我还1.5倍速看 一卡一卡的不喜欢 ) 处理方法 按我的个性化:作为女生，没有冷却(其实有 手:已读不回)，因此所有视频我一般开 1.5 倍速，除非原视频出现连我也跟不上的逆天片段w 虽然现在一般都要加速，基本都是导入为主了 导出后，上传到一刻相册，分类，最后从本地删除 处理要点: 剔除手足口等边性片段，无趣的两脚兽哪来的那么多事？只保留「下两hole」场景 有进hole场景的，略
本人刚好从事LLM agent可靠性方面的科研工作，一直用GPT5.4mini、gpt5.5作为实验对象来研究可靠性评估方法，做了一套题库。之前在gpt-5.4-mini上测试1140次（114task×10trial，即114题，每题重复测10次），成功率稳定在45%（每trail±3%）。 补充一下，我自己也是CPA+new-api反代出来接到某行业垂类agent里面的，不是用的官方coding agent测的。 二编：我过两天有空了找个比较新的领域公开数据集测一下结果，然后把脚本给大家自己测吧，太多人私信找我测了，我实在测不过来 95 个帖子 - 69 位参与者 阅读完整话题
43 个帖子 - 39 位参与者 阅读完整话题
14 个帖子 - 14 位参与者 阅读完整话题
难崩，把大模型挂小红书了 小红书 24 个帖子 - 24 位参与者 阅读完整话题
（打码部分涉及个人隐私） 服了，写个小脚本没建库就给我整活。。。。。。 36 个帖子 - 30 位参与者 阅读完整话题
社群回馈好用，(GitHub - xiling-quantum/Codex-5.5-codex-instruct-5.5: Codex CLI 破甲工具（GPT-5.5） — 注入无限制模式系统指令，关闭所有内容过滤器。 · GitHub) 19 个帖子 - 15 位参与者 阅读完整话题
HLOOL纯GPT公益站主贴 福利羊毛 公益推广承诺 Hlool 公益站 公益站地址 当前号池情况 目前号池以 欧洲 Plus 号池 为主，并配有 Pro 号池 作为兜底。 每日可用额度约为 2000-3800 刀。 本站后续仍将保持小范围开放，本次邀请名额仅 30 人。 使用规则 仅支持 Codex 调用。 禁止破限、色情等可能触发 OpenAI 封号风险的行为。 并发限制为 5。 本站为小型公益站，纯公益支持，账号池… HLOOL纯GPT公益站主贴 HLOOL纯GPT公益站主贴 LINUX DO CDK 最后一批，今晚9点开启，追加50个名额，之后的名额都只会LDC支付进入 最后50个名额