paperclipai/paperclip
TypeScript · ★ 87,187 · 🍴 15,418 · 📈 2,589 stars today
The open-source app everyone uses to manage agents at work
中文介绍 面向工作场景的 Agent 管理开源应用，把日常使用的 AI Agent 集中到一处统一管理：配置、调度、查看运行状态与产出，避免在多个工具之间来回切换。适合团队或个人把 Agent 真正落进日常工作流。
2026-09-27
← 历史归档vectorize-io/hindsight
Python · ★ 32,101 · 🍴 3,565 · 📈 2,152 stars today
Hindsight: Agent Memory That Learns
中文介绍 为 AI Agent 提供会学习的记忆层，让 Agent 在交互中积累并复用历史经验，而不是每次从零开始。开发者可将其接入自有 Agent 框架，用于长期对话、个性化助手、跨会话任务连续性等场景。
NVIDIA/Model-Optimizer
Python · ★ 4,729 · 🍴 667 · 📈 354 stars today
A unified library of SOTA model optimization techniques like quantization, distillation, pruning, neural architecture search, speculative decoding, etc. It compresses deep learning models for downstream deployment frameworks like TensorRT-LLM, TensorRT, vLLM, etc. to optimize inference speed.
中文介绍 NVIDIA 推出的模型优化统一库，集合量化、蒸馏、剪枝、神经架构搜索、投机解码等 SOTA 压缩技术，帮助把深度学习模型做小做快，便于部署到算力受限环境。适合需要推理加速与模型瘦身的工程和研究团队。
dream-num/univer
TypeScript · ★ 19,182 · 🍴 1,633 · 📈 845 stars today
The Office Harness for AI Agents — Spreadsheets, Docs, Slides, Canvas, Relational Tables, and PDF in one runtime.
中文介绍 面向 AI Agent 的 Office 运行时，把表格、文档、幻灯片、画布、关系表和 PDF 统一在同一套引擎中。Agent 可直接读写结构化办公数据并协同操作，适合构建自动报表、文档生成与数据协作类智能应用。
tensorflow/tensorflow
C++ · ★ 200,438 · 🍴 77,606 · 📈 31 stars today
An Open Source Machine Learning Framework for Everyone
中文介绍 Google 开源的老牌机器学习框架，提供从数据预处理、模型构建训练到部署的完整工具链，支持 CPU/GPU/TPU 与多种语言接口。它是深度学习入门与工业落地最常用的基础设施之一，生态与文档十分成熟。
rohitg00/ai-engineering-from-scratch
Python · ★ 58,333 · 🍴 10,117 · 📈 828 stars today
Learn it. Build it. Ship it for others.
中文介绍 面向开发者的 AI 工程实战路径，主张先理解原理、再动手实现、最后交付给他人使用，覆盖从模型调用到应用上线的完整链路。适合想从零系统补齐 AI 工程能力、把 demo 打磨成可用产品的工程师。
openbao/openbao
Go · ★ 7,989 · 🍴 590 · 📈 360 stars today
OpenBao is a software solution to manage, store, and distribute sensitive data including secrets, certificates, and keys.
中文介绍 OpenBao 是管理、存储与分发敏感数据的开源方案，覆盖密钥、证书和各类 secrets 的集中管控，可视为 HashiCorp Vault 的开源分支替代。适合需要自托管密钥管理、动态凭证与加密即服务的团队。
block/buzz
Rust · ★ 34,816 · 🍴 4,594 · 📈 367 stars today
A hive mind communication platform
中文介绍 由 Block 推出的群体智能通信平台，让多个 Agent 或人员以类似蜂群的方式共享信息、协同决策。适用于需要多智能体协作、消息广播与知识汇聚的场景，具体架构与接入方式可参考仓库文档。
microsoft/vscode
TypeScript · ★ 193,062 · 🍴 43,534 · 📈 78 stars today
Visual Studio Code
中文介绍 微软开源的跨平台代码编辑器，以轻量、启动快和插件生态丰富著称，内置 Git、调试、终端与智能补全，并通过扩展支持几乎所有主流语言。它也是当前 AI 编程插件最主要的承载平台之一。
zhaoxuya520/reverse-skill
PowerShell · ★ 37,973 · 🍴 5,268 · 📈 409 stars today
Reverse Engineering / Authorized Penetration Testing / Security Research Skill Router Pack AI-powered routing + On-demand toolchain bootstrapping + Self-evolving knowledge base Supports Claude Code, Kiro, Cursor, Cline, and other AI coding clients 逆向/渗透/安全技能路由包 - AI 自动路由 + 按需自举工具链 + 自动进化经验库 | 支持 Cla
中文介绍 面向逆向工程、授权渗透测试与安全研究的技能路由包，用 AI 做能力调度，按需拉起工具链，并自带可自我演进的知识库，兼容 Claude Code 等客户端。适合安全从业者把零散工具整合成统一工作流。
llvm/llvm-project
LLVM · ★ 40,743 · 🍴 18,823 · 📈 29 stars today
The LLVM Project is a collection of modular and reusable compiler and toolchain technologies.
中文介绍 LLVM 项目集合模块化、可复用的编译器与工具链技术，涵盖中间表示、优化器、后端代码生成以及 Clang、LLD 等子项目。它是 C/C++、Rust 等语言编译器及众多静态分析与 JIT 工具的底层基础设施。
anthropics/claude-code-action
TypeScript · ★ 9,076 · 🍴 2,162 · 📈 15 stars today
中文介绍 Anthropic 官方提供的 GitHub Action，把 Claude Code 接入 CI/CD 流程，可在 issue、PR 等事件触发下自动执行代码审查、修复与实现任务。适合团队在 GitHub 仓库中引入 AI 自动化协作，减少重复工程杂活。
actions/runner-images
PowerShell · ★ 13,290 · 🍴 3,864 · 📈 13 stars today
GitHub Actions runner images
中文介绍 GitHub Actions 官方维护的 runner 镜像仓库，包含 Ubuntu、Windows、macOS 托管运行器预装软件的配置与构建脚本。可用于查证 CI 环境中可用的工具与版本，或自行构建与官方一致的 runner 镜像。
mobile-next/mobile-mcp
TypeScript · ★ 7,318 · 🍴 640 · 📈 143 stars today
Model Context Protocol Server for Mobile Automation and Scraping (iOS, Android, Emulators, Simulators and Real Devices)
中文介绍 基于 Model Context Protocol 的移动端自动化与抓取服务器，支持 iOS、Android 以及模拟器和真机，让 AI Agent 能统一操控设备完成点击、输入、截图等操作。适合移动端测试、数据采集与自动化流程场景。
vercel/next.js
JavaScript · ★ 142,614 · 🍴 33,147 · 📈 31 stars today
The React Framework
中文介绍 Vercel 维护的 React 全栈框架，提供服务端渲染、静态生成、路由、API 路由与增量更新等能力，兼顾开发体验与首屏性能。广泛用于内容站、电商与 SaaS 前端，是 React 生态中最主流的应用框架。
Training Object Permanence in World Models
👍 197
Object permanence and solidity are hallmarks of human cognitive priors. Recent studies show that video generation models, a paradigmatic class of current world models, have begun to show emerged reasoning abilities, making them ideal candidates for building human-like physical intelligence. Do video
中文介绍 针对视频生成模型这类世界模型是否具备客体永久性与固体性这一认知先验，论文考察其作为类人物理智能候选的潜力，并通过训练手段向视频生成模型中注入客体永久性，评估物体被遮挡、消失与再现场景下的生成一致性与推理表现。
SpeakerMem-R1: Speaker-Centered Dual-Track Memory for Multi-Party Dialogue
👍 83
Long-term conversational memory in multi-party settings requires more than retrieving relevant content from long-term conversations: it must distinguish who said what, whom each statement concerns, how individuals perceive one another, what information is shared by the group, and how states change o
中文介绍 多方对话的长期记忆不仅需要检索相关内容，还须区分谁说了什么、语句涉及谁、个体间如何相互认知、哪些信息为群体共享以及状态如何演变。论文提出以说话人为中心的双轨记忆框架 SpeakerMem-R1，将说话人建模与内容检索解耦以提升长程对话一致性。
Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs
👍 64
While Large Language Models (LLMs) rely on highly non-linear components, in this work we demonstrate that they exhibit fundamental linearity: when inputs from distinct text streams are linearly combined, the model outputs a superposition of the individual next-token distributions. We term this the S
中文介绍 论文证明看似高度非线性的大语言模型具有基础线性性：当来自不同文本流的输入被线性组合时，模型输出的下一个 token 分布恰为各输入单独输出分布的叠加，作者称之为线性叠加。该性质为理解 Transformer 内部表征与多线索并行处理提供了新视角。
Spatial-Interactor: Learning Spatial Reasoning through Interaction with the Observable Physical World
👍 47
Spatial reasoning is essential for vision-language models (VLMs) to understand and act in the physical world. Reasoning in dynamic environments requires VLMs to perceive local state transitions caused by object motion and viewpoint changes and integrate them over long trajectories to maintain an upd
HappyWorld-Bench
👍 41
Evaluating world models requires assessing both the quality of the worlds they generate and their consistency and responsiveness under exploration, interaction, and modification. We introduce HappyWorld-Bench, a comprehensive benchmark that evaluates whether generated worlds remain reliable as agent
中文介绍 提出 HappyWorld-Bench，一个面向世界模型的综合基准：不仅评估生成世界的质量，还检验其在探索、交互与修改操作下能否保持一致性、可靠性与响应性，为判断生成世界能否作为智能体可信赖的可交互环境提供统一评测标准。
The Past Frames the Future: Memory for Autoregressive Video Generation
👍 39
Advances in generative models have improved video fidelity, enabling long-horizon generation, interactive world modeling, and evolving visual environments. Autoregressive (AR) video generation extends visual sequences through causal rollouts. However, a fundamental bottleneck emerges: as the generat
中文介绍 自回归视频生成通过因果 rollout 扩展视觉序列，但随生成推进会出现记忆瓶颈，导致长程一致性与误差累积问题。论文为自回归视频生成引入记忆机制，用历史帧信息约束后续帧生成，支撑长时程生成、交互式世界建模与演化视觉环境的连贯性。
Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents
👍 34
Agentic memory systems reuse past experience to improve future performance, yet most existing designs curate memory at write time: once a task is completed, its trajectory is distilled into a fixed artifact, such as a reflection, workflow, skill, or reasoning strategy, that is later retrieved by sim
中文介绍 现有 agent 记忆系统多在写入时就将轨迹固化为反思、工作流、技能或推理策略等固定产物，检索时难以适配新任务。论文提出「即时记忆」机制，让 agent 在任务执行中动态筛选与整理历史经验，实现任务自适应的记忆策展与复用。
OmniEcho: Spatial Audio Understanding for Embodied Agents
👍 20
Humans can effortlessly localize the direction of a sound source and integrate it with visual cues for reasoning, yet this remains challenging for embodied agents. In particular, it is still unclear how to effectively evaluate and model spatial audio understanding in embodied settings. To address th
中文介绍 人类能轻松定位声源方向并与视觉线索联合推理，具身智能体却难以做到，且缺少相应的评测与建模方案。论文提出 OmniEcho，构建具身场景下的空间音频理解评测与模型框架，融合音频与视觉信息以提升声源定位和跨模态空间推理能力。
Agent-Editing World Model: Rethinking World Modeling for LLM Agents
👍 17
Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon tasks across diverse environments. To further improve agent performance, existing language world models typically predict environment observations, yet reconstructing high-entropy, execution-dependent tool res
中文介绍 现有语言世界模型让 LLM agent 预测环境观测，但重建高熵、依赖具体执行的工具返回结果十分困难。论文提出 Agent-Editing World Model，把世界建模从「预测观测」转为对状态的编辑式更新，降低建模难度并提升长时程任务中的规划与决策表现。
PACT: From Credit Assignment to Critic Alignment
👍 16
Reinforcement learning has become a central component of large language model (LLM) post-training, yet token-level credit lacks a generally accepted mathematical definition, leaving its relationship to commonly used training signals unclear. We formulate three regularity conditions, namely Completen
中文介绍 针对 LLM 强化学习后训练中 token 级信用分配缺乏公认数学定义、与常用训练信号关系不清的问题，论文提出完整性等三条正则条件加以刻画，并据此提出从信用分配到 critic 对齐的方法 PACT，改善 RL 后训练的稳定性与效果。
Capable yet Parsimonious: Extracting and Characterizing Hidden Chain-of-Thought in Frontier Models
👍 15
The rapid capability gains of frontier language models are widely attributed to improved reasoning abilities, yet this cannot be verified as raw CoT traces in closed-source systems are hidden. By registering a simple custom tool through a standard API feature, we induce frontier models to externaliz
中文介绍 闭源前沿模型的原始思维链不可见，其能力提升是否源于推理增强无法验证。论文通过标准 API 注册一个自定义工具，诱导前沿模型外化隐藏的 CoT，从而提取并刻画其推理轨迹，发现模型能力增强的同时 CoT 使用仍相对克制。
Rufus-Air: An Open LLM Post-Training Recipe
👍 13
Rufus-Air is an open and reproducible post-training recipe on GLM-4.5-Air-Base (106B-A12B), organized as a serial pipeline of eight stages: SFT, Reasoning RL, Coding RL, Instruction-Following RL, General Agent, Coding Agent, Search Agent, and RLHF. We document the data, reward design, infrastructure
中文介绍 Rufus-Air 是在 GLM-4.5-Air-Base（106B 总参、A12B 激活）上构建的开源可复现后训练方案，串行组织 SFT、推理 RL、代码 RL、指令遵循 RL、通用 Agent、代码 Agent、搜索 Agent 与 RLHF 共八个阶段，并公开数据、奖励设计与基础设施细节。
Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms
👍 12
Language-model agents increasingly face long-horizon tasks with evolving state, interdependent decisions, and delayed outcomes. Scaling their training requires diverse agentic environments, dependable outcome signals, and low extension cost. Existing generation pipelines commonly construct an enviro
中文介绍 面向状态演化、决策相互依赖且反馈延迟的长时程 agent 任务，论文提出可验证隐藏动力学游戏（Verifiable Hidden Dynamics Play）：从已有解法的机制出发自动生成 agentic RL 环境，提供可靠的结果验证信号，并显著降低环境扩展成本。
PackLab: A Comprehensive Framework for Developing, Training, and Evaluating MLLMs in Robotic Bin Packing
👍 11
Robotic bin packing requires long-horizon sequential decision-making, as each object placement affects the available space for subsequent packing. Existing methods primarily rely on hand-crafted geometric heuristics that optimize predefined objectives or reinforcement learning policies learned throu
中文介绍 机器人装箱是长时程顺序决策问题，每次放置都会改变后续可用空间，现有方法多依赖手工几何启发式或强化学习策略。PackLab 提出集开发、训练与评测于一体的框架，用多模态大模型在装箱任务中实现更强的决策与泛化能力。
MemBodied: Recurrent Associative Memory for Vision-Language-Action Models
👍 11
Vision-Language-Action models provide a strong foundation for general-purpose robot control, yet a vast majority of policies do not preserve and leverage episode-level information beyond the current observation. This limitation is consequential in history-dependent manipulation tasks that depend on
中文介绍 多数 Vision-Language-Action 策略只依赖当前观测，不保留 episode 级信息，在依赖历史的操作任务中表现受限。MemBodied 为 VLA 模型引入循环联想记忆模块，跨时间步累积与检索历史表征，提升历史依赖型长时程操作任务的表现。
Qwen-Planner-Agent: A Closed-Loop AI-for-AI Framework for Real-World Mobile Planner Agents
👍 10
The rapid progression of large language models is extending AI from passive content generation into the active workflows of engineering and scientific discovery. This shift raises a compelling question: can AI be both the object of development and an active participant in building next-generation AI
中文介绍 论文提出 Qwen-Planner-Agent，一个面向真实移动端规划智能体的闭环 AI-for-AI 框架：让 AI 既是开发对象也是研发参与者，通过自动数据构造、训练与评测反馈的闭环持续迭代规划能力，探索以 AI 加速下一代 agent 研发的路径。
Hunyuan-A13B Technical Report
👍 10
We present Hunyuan-A13B, an open-source large language model based on a Mixture-of-Experts architecture. It contains 80 billion total parameters but activates only 13 billion during inference, balancing model capability, computational efficiency, and deployment cost. The model is pretrained on a rig
中文介绍 Hunyuan-A13B 是采用 MoE 架构的开源大语言模型，总参数 80B、推理时仅激活 13B，在模型能力、计算效率与部署成本之间取得平衡。模型在大规模语料上预训练，兼顾性能与推理开销，便于实际场景部署。
Coding Agents for Generalized Task and Motion Planning Problems
👍 9
Task and motion planning (TAMP) problems remain difficult even with full observability and object-centric states because discrete decisions are tightly coupled to geometric, kinematic, and dynamic constraints. Generalized TAMP addresses this difficulty by exploiting regularities across problem insta
中文介绍 任务与运动规划（TAMP）中离散决策与几何、运动学、动力学约束紧密耦合，即便状态完全可观测仍然困难。论文利用问题实例间的规律性，用编码智能体生成可复用的规划程序，实现泛化的 TAMP 求解与跨实例迁移。
IterSynth: Rethinking Deep Search Agents via Role-Decoupled Iterative Synthesis
👍 9
Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer from two limitations: role coupling, where one policy must handle planning, evidence use, and synthesis; and context accumulation, where growing
中文介绍 深度搜索要求 agent 分解查询、检索证据并综合有据答案，而现有 ReAct 式智能体存在角色耦合（单一策略兼顾规划、证据使用与综合）与上下文累积两大局限。IterSynth 提出角色解耦的迭代合成框架，分离不同职责并多轮收敛以提升答案质量。
WhatWorkedBench: Benchmarking Experimental Understanding in AI Agents
👍 9
AI research agents need reliable knowledge of how their experiments change outcomes. We introduce WhatWorkedBench to measure experimental understanding, the accuracy of predictions about component changes after budgeted experimentation. Agents inspect code, select measurements, and submit a response
中文介绍 提出 WhatWorkedBench，用于衡量 AI 科研智能体的「实验理解」能力：在有限实验预算内，agent 需阅读代码、选择度量并预测改动某一组件后实验结果的变化，从而量化其对实验因果关系的可靠知识，为科研 agent 提供新的评测维度。
The New Copilot
@jacobandreou · 7.8K 粉丝 · 521.0K 阅 · 500 赞 · 77 转
The bottleneck is no longer intelligence. Over the last six months, intelligence has accelerated dramatically. But smarter models are not enough for everyone to benefit from novel intelligence. New
中文介绍 作者认为 AI 的瓶颈已不是智能本身：过去半年模型智能飞速提升，但光有更强的模型不足以让所有人享受到新智能，还需要新的产品形态与分发方式。
We're open sourcing the company brain, Here's how we designed the multi-player harness
@DhravyaShah · 63.2K 粉丝 · 112.4K 阅 · 608 赞 · 45 转
It's been just 2 weeks since we discontinued the company brain harness. We had to offboard tons of people to other products, but many of our customers, and others suggested us to open source our work
中文介绍 分享为何把「company brain」这套多人协作 harness 开源：产品两周前刚停掉、大量用户被迁往其他产品，客户与社区建议公开，于是发布多玩家 harness 的设计思路。
Why AI is booming, but productivity isn't
@chamath · 2.4M 粉丝 · 97.8K 阅 · 551 赞 · 44 转
In June, I wrote that vibe coding was dead and that ROI-driven analysis of AI was about to go from a nice-to-have to a necessity. Here's what AI ROI is and why it's tricky to see in GDP numbers so
中文介绍 Chamath 重提 6 月观点：vibe coding 已死，AI 的 ROI 分析从可选变必需。他解释 AI ROI 是什么、为何难在 GDP 数据里体现——AI 火热但生产率没跟上。
Prescriptions for Prosperity in the Digital Economy
@saylor · 5.2M 粉丝 · 96.2K 阅 · 661 赞 · 98 转
Artificial intelligence will make it possible for individuals and companies to produce far more than they can today. That makes the freedom to create, finance, own, and exchange things more important.
中文介绍 Saylor 谈数字经济下的繁荣处方：AI 将让个人与企业产出远超当下，因此创造、融资、拥有和交换的自由变得更关键。属宏观叙事型观点，非教程。
Jev is a state-of-the-art search reranker
@ErikKaum · 2.0K 粉丝 · 75.0K 阅 · 501 赞 · 44 转
We evaluated Jev as a reranker for search results in turbopuffer. The results may surprise you. Jev isn’t just a good reranker, it performed better than existing state-of-the-art rerankers in our
中文介绍 turbopuffer 团队把 Jev 用作搜索结果 reranker 做评测，结论出乎意料：它不只是好用的 reranker，还超过了现有 SOTA 方案，并公开了评测过程与数据。
Using Claude Code: Spending your effort
@trq212 · 354.7K 粉丝 · 62.7K 阅 · 965 赞 · 52 转
One of the best parts of our newest Claude models is how they respond to effort without breaking the prompt cache in Claude Code, but I’ve received a lot of questions on this from users. What is
中文介绍 讲 Claude Code 里如何「花」effort：新 Claude 模型对 effort 响应更好且不会破坏 prompt cache。作者被问得太多，于是解释 effort 参数的用法与注意点。
We built a 20x faster Grok bot
@cerebras · 76.6K 粉丝 · 37.6K 阅 · 554 赞 · 28 转
Written by @milksandmatcha and @0xSero A personal assistant should save you time and effort. Over the past few weeks, we’ve been obsessively testing AI personal assistants on everyday tasks, from
中文介绍 Cerebras 展示自家 20 倍速的 Grok 个人助理：团队花数周在日常任务上反复测试 AI 助理，对比延迟与体验，说明快在哪里、实际是否值得用。
Ghast AI Season 3 Starts September 25
@Ghast_AI · 10.0K 粉丝 · 6.3K 阅 · 542 赞 · 31 转
We just introduced the Capability Economy, a market where private AI work can be bought and verified without exposing the method behind it. Every Capability in that market starts as a workflow someone
中文介绍 Ghast AI 第三季 9 月 25 日开跑，同步推出「Capability Economy」：一个可买卖并验证私有 AI 工作成果、却不必暴露方法的市场，每个 capability 都源自某人的工作流。
OpenRouter: from Seed to Stripe — with OpenRouter’s Alex Atallah & AMP’s Anjney Midha
In 2023 most people doubted that there could be more than 1 or 2 frontier model labs. Now there are dozens.... and Stripe just bought the best known one for $7B.
中文介绍 OpenRouter 的 Alex Atallah 与 AMP 的 Anjney Midha 对谈，回顾公司从种子轮走向被 Stripe 以 70 亿美元收购的历程；并提到 2023 年多数人质疑前沿模型实验室只能有一两家，如今已多达数十家。
Proaction boosts sales 60% and saves 75+ hours with Codex
With Codex, GPT-Live-1, and GPT-6 Astra, Proaction builds, operates, and sells modern fleet management faster.
中文介绍 OpenAI 发布客户案例：Proaction 借助 Codex、GPT-Live-1 与 GPT-6 Astra 构建并运营现代车队管理业务，销售额提升 60%，节省 75 小时以上工时。
The Pentagon wants $30 million to build an AI-powered lie detector
The US government wants to spend $30.3 million over the next five years on an improved form of lie detector, according to a Department of Defense budget request. The program, called Polygraph+ or Polygraph Next, will focus on scoring algorithms that use artificial intelligence and machine learning a
中文介绍 美国国防部预算申请显示，计划未来五年投入 3030 万美元开发改进型测谎系统，项目名为 Polygraph+（又称 Polygraph Next），重点研发基于人工智能的评分算法。
[AINews] The Future of Latent Space
A quiet day lets us discuss the work behind the scenes - now open for business!
中文介绍 Latent Space 发布每日 AI 新闻栏目 AINews，本期以「Latent Space 的未来」为题，介绍团队幕后的工作，并宣布正式对外开放业务。
Runway’s WorldPrompt and the Engineering of Real-Time Worlds
GWM Worlds 2 uses persistent context and timed actions to steer a world model generating video and audio in real time.
中文介绍 Runway 推出 WorldPrompt 与世界模型 GWM Worlds 2，通过持续上下文与定时动作引导模型实时生成视频与音频，文章讨论实时世界生成背后的工程实现。
ChatGPT Pro Max 🤖, Muse realtime avatar 🎭, DeepSeek $1B ARR 💰
中文介绍 TLDR AI 当日要闻：ChatGPT Pro Max 发布、Muse 实时虚拟形象、DeepSeek 年化经常性收入达 10 亿美元。
Introducing Gemini 3.8 Live with Live Avatar
中文介绍 DeepMind 发布 Gemini 3.8 Live，并新增实时虚拟形象（Live Avatar）功能。
Foundries vs Navigators: Lowering the Cost of Science
Guest Post: In science, thinking has gotten cheap but doing has not. This asymmetry is reshaping how research companies operate, largely inconspicuously.
中文介绍 客座文章指出，科学研究中「思考」的成本已大幅下降，但「执行」的成本并未同步降低；这一不对称正在重塑研究型公司的运作方式，且多为外界不易察觉。
Accelerating vision-language models with LFM2.5-VL-DSpark
中文介绍 Liquid AI 在 Hugging Face 博客发布 LFM2.5-VL-DSpark，介绍加速视觉语言模型的方法。
[AINews] Meta Connect 2026: Muse glasses, voice, video, and Charm
Team Zuck is absolutely on fire.
中文介绍 Latent Space 的 AINews 汇总 Meta Connect 2026 大会内容，包括 Muse 智能眼镜、语音与视频功能，以及 Charm。
Gemini TTS 🗣️, Claude’s novel enzyme 🧬, Google private memory 🔒
中文介绍 TLDR AI 当日要闻：谷歌发布 Gemini 文本转语音功能、Claude 参与发现新型酶、谷歌推出私密记忆功能。
How to Use NVIDIA Warp and MjWarp to Accelerate Robotics Simulation and Learning Workflows
中文介绍 英伟达在 Hugging Face 发布教程，介绍如何使用 Warp 与 MjWarp 加速机器人仿真与学习工作流。
Advancing Private AI Compute with secure, server-side memory
Introducing private, server-side memory to Private AI Compute for personal AI.
中文介绍 DeepMind 为旗下的 Private AI Compute 引入私密的服务端记忆功能，面向个人 AI 使用场景。
Two years of OpenAI Academy
Marking two years of OpenAI Academy and bringing AI skills to even more communities.
中文介绍 OpenAI 发文纪念 OpenAI Academy 成立两周年，并称将继续向更多社区推广 AI 技能。
Gemini 3.8 text-to-speech says hello
中文介绍 DeepMind 发布 Gemini 3.8 文本转语音功能。
美股方面，SPY 报 771.35、距 52 周高 -1.03%，QQQ 报 744.50、RSI14 64.7 且 MACD 7.4683 高于信号线 4.0132，两者与 AAPL(RSI 65.7)、MSFT、NVDA 均为多头排列，指数与权重股技术面整体偏强；但 META RSI 71.3 进入超买且单日回落 3.33%，GOOGL RSI 49.7、价格卡在 SMA20 与 SMA50 之间，大型科技股内部出现分化。加密方面，恐慌贪婪指数 74 处于贪婪区间，总市值 2.90T 美元、24 小时 -2.21%，BTC 主导率 58.3%、ETH 11.3%；BTC 84307.45、ETH 2691.95、SOL 121.44 仍保持多头排列，RSI 依次 64.7、63.7、68.6，但 5 日涨跌幅分别为 -2.65%、-3.04%、+2.27%，且距 52 周高仍有 33% 至 49% 空间，形态为趋势向上、动能放缓。中概整体偏弱：BABA、PDD、JD、0700.HK 均为空头排列，RSI 依次 42.1、37.0、33.5、48.2，虽四者皆出现 MACD 金叉但读数仍在零轴下方，属超跌后的弱修复。商品与外汇方面，黄金期货 MACD 转负且空头排列；WTI 原油触发 MACD 死叉；美元指数 101.04 多头排列、RSI 65.2 并接近 52 周高；10Y 美债收益率 5.18%、RSI 76.3 超买；VIX 14.87、单日 -5.11% 且空头排列，避险指标同步走低。
今日关注
^TNX
10Y 美债收益率 (%)
偏上行
现价 5.18，站上 SMA20(4.92)、SMA50(4.77) 与 SMA200(4.42)，呈多头排列；RSI14 76.3 已进入超买区，MACD 0.1082 高于信号线 0.0874，近 5 日 +3.72%，距 52 周高仅 -0.88%，属趋势向上、短线动量过热的偏上行读数。
QQQ
Nasdaq 100 ETF
偏上行
现价 744.50，位于 SMA20(720.87)、SMA50(712.65)、SMA200(665.59) 之上，多头排列；RSI14 64.7 尚未触及超买，MACD 7.4683 明显高于信号线 4.0132，红柱扩张，近 5 日 +3.19%，距 52 周高只有 -0.55%，价格与动量同向上行。
BABA
阿里巴巴 (BABA)
偏下行
现价 109.74，低于 SMA20(111.67)、SMA50(117.81) 与 SMA200(132.14)，为空头排列；RSI14 42.1 处于弱势区，MACD -1.8395 虽上穿信号线 -2.1282 形成金叉，但两者均在零轴下方，近 5 日 -3.09%、距 52 周高 -43.04%，属下行趋势中的弱修复。
GC=F
黄金期货
偏下行
现价 4321.2，低于 SMA20(4407.47)、SMA50(4354.55) 与 SMA200(4556.06)，呈空头排列；MACD -23.8271 低于信号线 -4.5374，柱状体转负，RSI14 42.8 偏弱，近 5 日 -2.34%，距 52 周高 -22.65%，下行动能相对占优。
GOOGL
Alphabet (GOOGL)
中性
现价 343.92 与 SMA20(342.17)、SMA50(344.29) 几近重合，SMA200 为 338.21，均线相互纠缠；RSI14 49.7 贴近中性 50，MACD 0.0833 仅微幅高于信号线 -0.3674，近 5 日 -1.61%，无明确趋势或交叉信号，读数方向不明。
全部资产
^VIX
VIX 恐慌指数
$14.87
-5.11%
- 5 日
- +0.41%
- 距 52w 高
- -57.9%
- RSI(14)
- 46.3
- 趋势
- 空头
- SMA 20 / 50 / 200
- 15.68 / 15.97 / 18.06
- MACD / 信号
- -0.174 / -0.063
MACD 死叉 (4 天前)空头排列
^TNX
10Y 美债收益率 (%)
$5.18
+0.43%
- 5 日
- +3.72%
- 距 52w 高
- -0.9%
- RSI(14)
- 76.3
- 趋势
- 多头
- SMA 20 / 50 / 200
- 4.92 / 4.77 / 4.42
- MACD / 信号
- 0.108 / 0.087
RSI 超买接近 52 周高多头排列
DX-Y.NYB
美元指数 DXY
$101.03
-0.25%
- 5 日
- +0.81%
- 距 52w 高
- -0.8%
- RSI(14)
- 65.2
- 趋势
- 多头
- SMA 20 / 50 / 200
- 99.83 / 99.95 / 99.20
- MACD / 信号
- 0.361 / 0.144
接近 52 周高多头排列
SPY
S&P 500 ETF
$771.35
+0.54%
- 5 日
- +1.27%
- 距 52w 高
- -1.0%
- RSI(14)
- 55.9
- 趋势
- 多头
- SMA 20 / 50 / 200
- 765.35 / 761.57 / 718.45
- MACD / 信号
- 1.583 / 0.988
MACD 金叉 (3 天前)接近 52 周高多头排列
QQQ
Nasdaq 100 ETF
$744.50
+0.46%
- 5 日
- +3.19%
- 距 52w 高
- -0.6%
- RSI(14)
- 64.7
- 趋势
- 多头
- SMA 20 / 50 / 200
- 720.87 / 712.65 / 665.59
- MACD / 信号
- 7.468 / 4.013
接近 52 周高多头排列
AAPL
Apple
$341.07
+1.53%
- 5 日
- +1.47%
- 距 52w 高
- -1.2%
- RSI(14)
- 65.7
- 趋势
- 多头
- SMA 20 / 50 / 200
- 329.40 / 321.83 / 287.77
- MACD / 信号
- 6.211 / 5.408
接近 52 周高多头排列
MSFT
Microsoft
$516.17
+3.66%
- 5 日
- +4.53%
- 距 52w 高
- -6.8%
- RSI(14)
- 63.2
- 趋势
- 多头
- SMA 20 / 50 / 200
- 500.04 / 475.81 / 432.06
- MACD / 信号
- 6.781 / 7.551
多头排列
NVDA
Nvidia
$225.07
+0.22%
- 5 日
- +1.26%
- 距 52w 高
- -4.8%
- RSI(14)
- 55.2
- 趋势
- 多头
- SMA 20 / 50 / 200
- 221.75 / 215.98 / 199.47
- MACD / 信号
- 2.281 / 1.915
MACD 金叉 (3 天前)多头排列
GOOGL
Alphabet
$343.92
+0.46%
- 5 日
- -1.61%
- 距 52w 高
- -15.8%
- RSI(14)
- 49.7
- 趋势
- 中性
- SMA 20 / 50 / 200
- 342.17 / 344.29 / 338.21
- MACD / 信号
- 0.083 / -0.367
TSLA
Tesla
$372.11
-1.54%
- 5 日
- +2.15%
- 距 52w 高
- -25.4%
- RSI(14)
- 55.3
- 趋势
- 中性
- SMA 20 / 50 / 200
- 365.68 / 348.25 / 395.62
- MACD / 信号
- 6.364 / 5.286
META
Meta
$751.66
-3.33%
- 5 日
- +12.90%
- 距 52w 高
- -3.6%
- RSI(14)
- 71.3
- 趋势
- 中性
- SMA 20 / 50 / 200
- 660.86 / 615.75 / 626.47
- MACD / 信号
- 42.818 / 31.149
RSI 超买
BTC-USD
Bitcoin
$84,307.45
+0.32%
- 5 日
- -2.65%
- 距 52w 高
- -33.2%
- RSI(14)
- 64.7
- 趋势
- 多头
- SMA 20 / 50 / 200
- 80,157.59 / 75,711.51 / 71,000.97
- MACD / 信号
- 2,458.735 / 2,159.914
多头排列
ETH-USD
Ethereum
$2,691.95
+0.05%
- 5 日
- -3.04%
- 距 52w 高
- -43.4%
- RSI(14)
- 63.7
- 趋势
- 多头
- SMA 20 / 50 / 200
- 2,567.21 / 2,372.48 / 2,097.62
- MACD / 信号
- 93.023 / 89.208
多头排列
SOL-USD
Solana
$121.44
-0.47%
- 5 日
- +2.27%
- 距 52w 高
- -48.8%
- RSI(14)
- 68.6
- 趋势
- 多头
- SMA 20 / 50 / 200
- 107.90 / 97.71 / 84.78
- MACD / 信号
- 6.246 / 5.095
多头排列
BABA
阿里巴巴 (BABA)
$109.74
-0.80%
- 5 日
- -3.09%
- 距 52w 高
- -43.0%
- RSI(14)
- 42.1
- 趋势
- 空头
- SMA 20 / 50 / 200
- 111.67 / 117.81 / 132.14
- MACD / 信号
- -1.839 / -2.128
MACD 金叉 (4 天前)空头排列
PDD
拼多多 (PDD)
$77.57
-0.81%
- 5 日
- -1.69%
- 距 52w 高
- -44.4%
- RSI(14)
- 37.0
- 趋势
- 空头
- SMA 20 / 50 / 200
- 80.02 / 84.49 / 95.41
- MACD / 信号
- -1.861 / -1.939
MACD 金叉 (3 天前)空头排列
JD
京东 (JD)
$26.51
-0.90%
- 5 日
- -1.52%
- 距 52w 高
- -28.1%
- RSI(14)
- 33.5
- 趋势
- 中性
- SMA 20 / 50 / 200
- 27.37 / 29.39 / 28.96
- MACD / 信号
- -0.652 / -0.692
MACD 金叉 (4 天前)
0700.HK
腾讯控股 (0700.HK)
HK$436.60
-0.41%
- 5 日
- +4.20%
- 距 52w 高
- -36.1%
- RSI(14)
- 48.2
- 趋势
- 空头
- SMA 20 / 50 / 200
- 435.78 / 449.78 / 502.45
- MACD / 信号
- -3.242 / -4.902
MACD 金叉 (3 天前)空头排列
GC=F
黄金期货
$4,321.20
+0.54%
- 5 日
- -2.34%
- 距 52w 高
- -22.6%
- RSI(14)
- 42.8
- 趋势
- 空头
- SMA 20 / 50 / 200
- 4,407.47 / 4,354.55 / 4,556.06
- MACD / 信号
- -23.827 / -4.537
空头排列
CL=F
WTI 原油期货
$92.41
-2.33%
- 5 日
- -7.87%
- 距 52w 高
- -22.7%
- RSI(14)
- 48.9
- 趋势
- 多头
- SMA 20 / 50 / 200
- 95.31 / 88.04 / 81.55
- MACD / 信号
- 2.362 / 3.553
MACD 死叉 (3 天前)多头排列
USDCNY=X
美元 / 人民币
¥6.70
-0.11%
- 5 日
- +0.09%
- 距 52w 高
- -6.8%
- RSI(14)
- 38.1
- 趋势
- 空头
- SMA 20 / 50 / 200
- 6.71 / 6.73 / 6.83
- MACD / 信号
- -0.008 / -0.009
MACD 金叉 (3 天前)接近 52 周低空头排列
以上内容均基于公开行情数据与公开技术指标的计算读数，仅供技术指标解读参考，不构成任何投资建议或买卖依据；指标状态随时可能变化，过去走势不代表未来表现，请读者独立判断并自行承担风险。
Australia news live: Albanese says announcement of Medicare hack ‘purely in national interest’
Follow live updates Get our breaking news email, free app or daily news podcast Good morning Welcome to our Sunday live blog. Flight delays and cancellations caused frustration and heartache for some footy fans racing to make it to Saturday’s AFL grand final before the first bounce. And the PM says
中文摘要 澳大利亚总理阿尔巴尼斯表示，公布医保系统遭黑客攻击一事「纯粹出于国家利益」。该消息出自其周日新闻直播博客，内容还包括部分球迷因航班延误和取消而错过澳式橄榄球联盟总决赛。
Evicted 87-Year-Old Woman Urges Spaniards to ‘Fight’ to Avoid Same Fate
María del Carmen Abascal, whose case has become a symbol of Spain’s housing crisis, appeared in a video from a hospital bed after being removed from her apartment.
中文摘要 西班牙87岁老妇阿巴斯卡尔被逐出住所后，在医院病床上录制视频，呼吁民众「抗争」以免重蹈其覆辙。她的遭遇已成为西班牙住房危机的象征。
Trump rejects Iran deal to reopen Strait of Hormuz in seven days
Iran's foreign minister acknowledges Trump's comments, but adds that Tehran is waiting for the "definitive views" of mediators.
中文摘要 美国总统特朗普拒绝伊朗提出的七日内重开霍尔木兹海峡的方案。伊朗外长确认了特朗普的表态，并称德黑兰正等待调解方的最终意见。
Student protesters disrupt NVIDIA AI climate panel
Student protesters have interrupted an NVIDIA panel during NYC Climate Week.
中文摘要 在纽约气候周期间，学生抗议者打断美国芯片企业英伟达举办的一场关于人工智能与气候议题的座谈会。
Nor'easter brings flooding as New York and New Jersey declare emergency
Tens of millions of people from Maine to Virginia are in the path of the powerful storm.
中文摘要 美国东北风暴带来洪水，纽约州和新泽西州已宣布进入紧急状态。从缅因州到弗吉尼亚州，数千万人处于这场强风暴的影响路径之上。
Australian backlash to datacentres is faux import from US, officials say: ‘We are not the United States’
Pro-datacentre officials argue scale of Australia’s buildout is smaller, more regulated and less damaging than in US Australian politicians and business leaders are arguing that their populace’s growing resentment of AI datacentres is an inauthentic import from the United States. Like the US, Austra
中文摘要 澳大利亚政界和商界人士称，本国民众对人工智能数据中心的反对情绪是从美国输入的，并非本土问题。他们主张澳方建设规模更小、监管更严、破坏性更低。
Trump says he is rolling back Biden-era US fuel economy rules for cars
The US president said he would end a so-called 'EV mandate' that steered consumers to electric vehicles.
中文摘要 特朗普表示将撤销拜登时期美国汽车的燃油经济性规定，并终止其所说的引导消费者购买电动汽车的「电动车强制令」。
At Summit, Xi Sought to Tilt Trump’s Stance on America’s Place in Asia
In his talks with President Trump, the Chinese leader used praise, warnings and wartime history to challenge American support for Taiwan and Japan.
中文摘要 在与特朗普的会谈中，中国领导人运用赞美、警告和战时历史，试图动摇美国在台湾和日本问题上的立场，以影响美国在亚洲的角色定位。
Trump gave Xi Jinping a warm welcome. What did the talks accomplish?
From the National Archives to the state dinner, President Trump and Xi Jinping looked comfortable together in Washington. Two former NPR China correspondents on what the visit did and didn't deliver.
中文摘要 特朗普在华盛顿以参观国家档案馆、国宴等安排热情接待习近平，两人显得相处融洽。两位前美国国家公共电台驻华记者分析此次访问取得与未取得的成果。
Is Congo's Ebola outbreak coming under control?
Four months in, Congo's Ebola outbreak is one of the largest ever recorded. Officials and outside experts can't agree on whether it's finally slowing down.
Mother of woman found hanging in tree shocked as police say body was staged
Christy Spivey tells the BBC she only has more questions now police think Tasia Fortune's body was put in the tree after her death.
中文摘要 警方认为，被发现吊在树上的女子福琼的遗体是在死后被人放置于树上的。其母亲斯派维对警方这一说法感到震惊，称自己如今疑问更多。
Yamal nets in Spain’s 3-2 comeback win against England in Nations League
Lamine Yamal goal opens the scoring, but Spain made to comeback in Nations League win as Harry Kane misses penalty.
中文摘要 在欧洲国家联赛中，西班牙队以3比2逆转战胜英格兰队，亚马尔率先破门，英格兰队长凯恩罚失点球。
Trump Rejects Iran’s Cease-Fire Proposal to Reopen Strait of Hormuz
The president said the deal was not “acceptable” and demurred on whether he would restart military strikes after the midterm elections.
中文摘要 特朗普拒绝伊朗提出的重开霍尔木兹海峡的停火方案，称该协议「不可接受」，并回避了是否在中期选举后重启军事打击的问题。
Russia scales up strikes on Ukraine as largest steelmaker halts operations
A wave of Russian strikes has hit various regions across Ukraine.
中文摘要 俄罗斯对乌克兰多地发动新一轮打击，乌克兰最大钢铁企业已暂停运营。
Why is violence between Pakistan and Afghanistan recurring?
Pakistan has carried out new strikes inside Afghanistan.
中文摘要 巴基斯坦在阿富汗境内发动新一轮打击，两国之间的暴力冲突再次发生。半岛电视台时事节目探讨这一冲突反复出现的原因。
Bond Market Volatility Signals Risks for Corporate Debt Investors
Corporate bonds have been relatively resilient amid a global government bond selloff, but the strength may not last.
中文摘要 全球政府债券遭抛售之际，企业债表现相对抗跌，但彭博指出这一韧性恐难持续，债券市场波动率指标正对企业债投资者发出风险警示。
Bangladesh Targets Up to $1 Billion in Debut Sovereign Bond Sale
Bangladesh is preparing to raise between $500 million and $1 billion through its maiden foreign-currency sovereign bond sale within the next three months, a senior government official said.
中文摘要 孟加拉国计划在未来三个月内发行首只外币主权债券，拟募资5亿至10亿美元。一名政府高级官员表示，这将是该国首次发行外币主权债。
Burnham to revive Help to Buy scheme in bid to aid first-time buyers
Concern mounts among ministers that low housing construction volumes could hit growth forecasts
中文摘要 英国大曼彻斯特市长伯纳姆拟重启「购房援助」计划，以帮助首次购房者；与此同时，部长们愈发担忧住房建设量偏低可能拖累经济增长预测。
Bloomberg This Weekend 09/26/2026
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 The New York Times Kyiv Bureau Chief Andrew Kramer, The Netherlands Prime Minister Rob Jetten, Bull
中文摘要 彭博周末节目由David Gura、Christina Ruffini与Lisa Mateo在纽约直播，梳理周末重大新闻，并邀请《纽约时报》基辅分社社长Andrew Kramer及荷兰首相等嘉宾参与。
Manchester City rule breaches — potential punishments, appeals and next steps
The case could run for several years, legal experts warn
中文摘要 英超曼城俱乐部违规案的潜在处罚、上诉与后续步骤受到关注；法律专家警告，此案可能要耗时数年才能了结。
Pointed! Bloomberg's Weekly News Quiz For Risk-Takers
Pointed offers a strategic twist to the news quiz format, testing not just players’ knowledge of the news but also their confidence in their answers. Join Bloomberg's Alexis Christoforous, Christina Ruffini and David Gura as they play and check out the quiz for yourself at Bloomberg.com (Source: Blo
中文摘要 彭博推出面向风险偏好者的每周新闻测验「Pointed」，不仅考查对新闻的了解，也测试答题者的信心，由Alexis Christoforous、Christina Ruffini和David Gura主持。
US and China Seek Common Ground on AI
Bloomberg News Chief North Asia Correspondent Stephen Engle and ABC News State Department reporter Shannon Kingston tell Bloomberg This Weekend that President Donald Trump and Chinese President Xi Jinping’s summit highlighted efforts to establish dialogue on AI while building trust between the two c
中文摘要 彭博新闻驻北亚首席记者Stephen Engle与ABC新闻国务院记者Shannon Kingston表示，特朗普与习近平的峰会凸显美中在建立人工智能领域对话方面的努力。
How Syracuse's Semiconductor Gamble May Help Save the Rust Belt
What does it actually take to bring a Rust Belt city back to life? Micron is building one of the largest semiconductor factories in US history in Syracuse, a city that experienced major job loss after companies including GE, Carrier and General Motors left the area. Rob Simpson of CenterState CEO ex
中文摘要 美光公司在锡拉丘兹兴建美国史上最大半导体工厂之一。该市在通用电气、开利和通用汽车等企业撤离后流失大量就业，文章探讨这一投资能否重振「锈带」城市。
Energy costs and AI power demand reshape climate talks in New York
Disruption of global energy markets dominates discussions that previously concentrated more on cutting emissions
中文摘要 在纽约举行的气候谈判议程发生变化：全球能源市场动荡主导讨论，而此前焦点更多集中在减排；能源成本上升与人工智能用电需求成为新变量。
Birding Apps Bring More People Into the Wild
On Bloomberg This Weekend, writer and photographer Alexandra Marvar explains how apps such as Merlin are making birding easier and more accessible, helping draw new people to a hobby that offers an escape into nature. Speaking with hosts David Gura and Christina Ruffini, Marvar also explores whether
中文摘要 作家兼摄影师Alexandra Marvar在彭博周末节目中表示，Merlin等观鸟应用让观鸟更简单、更易参与，吸引更多人走进自然、接触这一爱好。
Mexico Stresses Sovereignty in US Cooperation
Mexico’s foreign minister Roberto Velasco Álvarez tells Bloomberg This Weekend that the country maintains a strong relationship with the US while insisting cooperation on security must respect Mexico’s sovereignty and territorial integrity. Speaking with hosts David Gura and Christina Ruffini, he sa
中文摘要 墨西哥外长Roberto Velasco Álvarez表示，墨西哥与美国关系牢固，但双方在安全领域的合作必须尊重墨西哥的主权与领土完整。
Russia Ukraine War Becomes Economic Battle
New York Times Kyiv bureau chief Andrew Kramer tells Bloomberg This Weekend that Russia and Ukraine are increasingly targeting each other’s economies as the war shifts toward attacks on energy, transportation and export infrastructure. Speaking with hosts David Gura and Christina Ruffini, Kramer say
中文摘要 《纽约时报》基辅分社社长Andrew Kramer表示，随着战争转向打击能源、交通和出口基础设施，俄罗斯与乌克兰正日益将对方经济作为攻击目标。
开始弃用 vscode
6 回复 · 程序员 节点
26 号还能注册 muse.ai 的方法
36 回复 · 程序员 节点
总结 muse 用途，两个字：顺滑
22 回复 · 程序员 节点
苹果真正在丢失的是品味！
11 回复 · Apple 节点
iPhone air 现在多多的 5199 价格还有下降空间吗？
7 回复 · Apple 节点
Meta Muse 国内注册避坑记录
16 回复 · 程序员 节点
Claude 250 的云端额度，领了会被封号 ？
13 回复 · 程序员 节点
Apple Pay 中秋假期出现的漏洞？求救怎么可以避免呢？
10 回复 · Apple 节点
脱离了 Windows 的保护，发现外边根本没有雨
41 回复 · Linux 节点
直营店 Mac 屏幕保修被拒了
44 回复 · Apple 节点
该源今日无内容。