Top Project
fmtlib/fmt
C++ · ★ 25,052 · 🍴 3,014 · 📈 955 stars today
A modern formatting library
中文介绍 现代化格式化库，提供高效、类型安全的字符串格式化功能。它的重要性在于提升了代码的可维护性和可读性，降低了格式化错误的风险，适合需要处理复杂字符串的项目。
2026-09-03
Morning Brief
2026-09-03
今天值得关注的变化是AI技术的迅速发展，以及其在多个行业应用的扩展。
今日最重要的 5 件事
-
1
OpenAI承诺10亿美元支持前线防御者
为什么重要 这一资金承诺展示了OpenAI在推动网络安全方面的决心，并可能影响全球网络防御策略。
-
2
Nvidia收购AI平台Hugging Face
为什么重要 此举将强化Nvidia在AI开发领域的领导地位，推动开源社区的创新和发展。
-
3
香港首次公开募股市场迎热潮
为什么重要 香港的IPO市场活跃程度提升，可能为投资者和企业的融资模式带来新的机遇与挑战。
-
4
大众汽车宣布裁员10万人
为什么重要 这一重组显示出企业在面对经济压力时的策略调整，反映了行业内更广泛的劳动力市场变化。
-
5
DeepMind推出最新全球天气AI模型
为什么重要 这一新技术在气候变化应对中的应用将对未来的天气预测产生深远影响。
市场状态
Risk-On
- 9 个跟踪资产处于多头趋势，5 个处于空头趋势。
- SPY 当前为 bullish，5 日 +0.27%，RSI 58.8。
- 加密恐惧贪婪指数为 65（贪婪）。
- 加密总市值 24h 变化为 +2.15%。
观察 观察 BTC 在 5 日 +3.76% 后能否企稳。
招聘信号
- 升温 AI安全 OpenAI和DeepMind推出的网络安全相关项目表明对AI安全专业人才的需求正在增加。
- 升温 AI工程师 Nvidia收购Hugging Face后，对AI工程师岗位的需求将大幅增加。
- 稳定 数据分析 香港IPO市场的活跃可能需要数据分析师以支持投资决策。
论文趋势
X 讨论主题
Most Practical
mattpocock/skills
Shell · ★ 247,301 · 🍴 20,965 · 📈 1,576 stars today
Skills for Real Engineers. Straight from my .agents directory.
中文介绍 提供真实工程师的技能，直接来自个人的 .agents 目录，旨在提升工程师的工作效率和技能树。为什么重要：为工程师提供实践导向的资源，提升职业发展和问题解决能力。
Most Surprising
NousResearch/hermes-agent
Python · ★ 240,817 · 🍴 49,340 · 📈 778 stars today
The agent that grows with you
中文介绍 一个与用户共同成长的智能代理，旨在优化用户的交互体验与效率。为什么重要：反映了个性化、学习型的智能代理发展趋势，适合需要持续学习的应用场景。
查看更多 (16)
DietrichGebert/ponytail
JavaScript · ★ 123,356 · 🍴 6,667 · 📈 2,138 stars today
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
中文介绍 使AI代理能像懒惰的高级开发者一样思考，从而减少不必要的代码生成，节省开发时间。为什么重要：强调了代码简约性的价值，推动了更高效的开发文化和理念。
anthropics/skills
Python · ★ 173,634 · 🍴 20,601 · 📈 277 stars today
Public repository for Agent Skills
中文介绍 公开的代理技能库，帮助开发者为AI构建适应性技能，提升其能力。为什么重要：推动了AI技能管理和调用的标准化，适合希望扩展AI功能的团队。
affaan-m/ECC
JavaScript · ★ 247,141 · 🍴 37,241 · 📈 749 stars today
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
中文介绍 一个代理性能优化系统，支持多项能力如记忆、安全性和开发。它的重要性在于整合多元素，以提升智能代理的整体性能与应用范围。
JuliusBrussee/caveman
Go · ★ 103,075 · 🍴 5,994 · 📈 545 stars today
🪨 why use many token when few token do trick — Claude Code skill that cuts 65% of tokens by talking like caveman
中文介绍 Claude Code技能，通过简化语言与表达，实现高达65%的令牌使用量优化。为什么重要：表明了在AI生成内容中，简洁表达带来的效率提升趋势。
blader/humanizer
Python · ★ 41,437 · 🍴 3,546 · 📈 1,214 stars today
Agent skill that removes signs of AI-generated writing from text
中文介绍 去除文本中AI生成特征的代理技能，确保输出内容更加自然。为什么重要：帮助开发者提升AI生成内容的可读性，提升用户体验和信任度。
google-research/timesfm
Python · ★ 30,692 · 🍴 2,930 · 📈 1,626 stars today
TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
中文介绍 Google Research开发的预训练时间序列基础模型，用于时间序列预测。为什么重要：增强了对复杂时间序列分析的能力，为数据驱动决策提供支持。
averygan/reclip
HTML · ★ 8,341 · 🍴 1,362 · 📈 123 stars today
Download videos from almost any website. Lightweight, self-hosted media downloader with a clean web UI.
中文介绍 轻量级自托管媒体下载器，支持从几乎任何网站下载视频，提供简洁的网页UI。为什么重要：反映了自我托管和开源趋势，适合希望控制媒体内容访问的用户。
bannedbook/fanqiang
Kotlin · ★ 52,125 · 🍴 8,468 · 📈 539 stars today
翻墙-科学上网
中文介绍 提供翻墙和科学上网的工具，帮助用户绕过网络限制。为什么重要：满足了用户对信息自由获取的需求，是互联网开放性的重要体现。
addyosmani/agent-skills
JavaScript · ★ 92,018 · 🍴 9,810 · 📈 260 stars today
Production-grade engineering skills for AI coding agents.
中文介绍 为AI编码代理提供的生产级工程技能，帮助提高开发效率。为什么重要：促进了AI在软件开发领域的实用性和应用，助力智能化开发流程的推进。
ByteByteGoHq/system-design-101
★ 88,312 · 🍴 9,820 · 📈 158 stars today
Explain complex systems using visuals and simple terms. Help you prepare for system design interviews.
中文介绍 通过可视化和简单术语解释复杂系统，帮助用户准备系统设计面试。为什么重要：提升了设计系统的能力与理解，对于程序员职业发展至关重要。
magnitudedev/magnitude
TypeScript · ★ 1,927 · 🍴 142 · 📈 130 stars today
Open source inference server that runs the best local models for your hardware, plugged into the agent you already use. Works with Pi, OpenCode, Hermes, OpenClaw, Codex, Claude Code, Oh My Pi, and Cline.
中文介绍 开源推理服务器，能够在本地硬件上运行最佳模型，适用于多种智能代理。为什么重要：支持灵活的AI模型使用，适合希望优化硬件资源和AI能力的开发者。
Imbad0202/academic-research-skills
Python · ★ 45,981 · 🍴 3,602 · 📈 498 stars today
Academic Research Skills for Claude Code: research → write → review → revise → finalize
中文介绍 提供学术研究技能，涵盖研究、写作、审阅等环节，支持Claude Code。为什么重要：为研究人员提供实用技能，促进学术工作的高效完成与质量提升。
Gitlawb/openclaude
TypeScript · ★ 32,331 · 🍴 9,018 · 📈 453 stars today
runs anywhere. uses anything
中文介绍 一种可在任何环境下运行的开源工具，支持多种技术栈。为什么重要：反映了开发工具的灵活性和跨平台能力，适合需要多样化工具链的开发者。
debpalash/VoiceStudio
Python · ★ 16,212 · 🍴 2,201 · 📈 1,738 stars today
VoiceStudio is the open-source, fully-local ElevenLabs alternative — voice cloning, voice design, video dubbing, dictation, transcription & audiobook creation in 646 languages.
中文介绍 开源的完全本地的替代EightLabs工具，提供语音克隆、设计、配音等多项功能。为什么重要：同步了最新的语音技术趋势，适合多语言内容创作者与配音工作者。
f/prompts.chat
HTML · ★ 168,965 · 🍴 21,761 · 📈 201 stars today
f.k.a. Awesome ChatGPT Prompts. Share, discover, and collect prompts from the community. Free and open source — self-host for your organization with complete privacy.
中文介绍 社区共享及收集提示的工具，支持完全隐私的自托管部署。为什么重要：推动了开放社区和隐私保护的理念，适合需要共享知识资源的组织。
obra/superpowers
Shell · ★ 281,320 · 🍴 25,202 · 📈 470 stars today
An agentic skills framework & software development methodology that works.
中文介绍 一个具有代理技能框架的软件开发方法论，旨在提升开发效率。为什么重要：强调了代理化开发的新思维，适合现代软件开发团队的工作流程改进。
Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills
👍 487
Autonomous agents are beginning to carry out machine-learning (ML) research end to end. These agents combine a model backbone with a harness for planning, execution, memory, and verification, but this architecture still leaves domain-specific know-how outside the agent. We call this missing layer op
中文介绍 提出Repo-To-Skill方法，将GitHub代码库提炼成AI4AI技能，以补充自主代理在机器学习研究中的知识缺口。该方法结合了模型主干与规划、执行、记忆和验证的架构。通过有效提取任务特定的知识，提升代理在特定领域的表现，对推动自主研究代理的发展具有重要意义。
HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?
👍 222
As agents move from research prototypes to deployed tools, their capability increasingly depends on model-external execution infrastructure, commonly termed the agent harness. Changing this harness while holding model weights fixed can substantially alter task performance. Current agent evaluations
中文介绍 通过HarnessDev研究大规模语言模型（LLMs）是否能够创建和演化自身的代理执行框架。论文表明，在保持模型权重不变的情况下，改变执行框架可以显著改善任务性能，揭示了模型外部基础设施对实现更强大代理至关重要的能力，关系到未来的AI工具演化与应用。
Aspire: Can Models Self-Evolve from Vague Goals?
👍 172
Many important forms of human learning begin with a vague goal, such as "become a better physicist" or "improve at research." Learners must interpret the goal, identify capability gaps, decide how to learn, and determine whether they have actually improved. In contrast, existing work on LLM self-evo
中文介绍 Aspire框架探讨模型如何从模糊目标中自我进化，强调学习者需解释目标、识别能力差距及学习方法等。与现有方法不同，该研究试图通过自我驱动方式提升模型自主学习能力，对推动自适应学习和目标导向研究的发展具有重要启示。
EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction
👍 110
Evaluating LLM agents is essential for guiding their development, yet it has grown prohibitively expensive: a single pass of a frontier model over an agentic benchmark can cost hundreds to thousands of dollars, a price paid repeatedly across iterative development cycles. Prior efforts, centered on b
中文介绍 提出EarlyEval方法，以降低大规模语言模型（LLM）代理评估的成本。在当前复杂代理系统的开发中，评估费用高达数千美元，而此方法通过早期结果预测实现更经济的评估方式，有助于加速迭代开发，推动智能代理在多领域的应用。
It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning
👍 64
Retrieval is the first stage of modern search and advertising systems, selecting a candidate set from a large item universe for downstream ranking and auction. Recent work increasingly leverages LLMs to improve retrieval through query expansion, data synthesis, and retrieval-feedback training. Howev
中文介绍 该研究探讨了强化学习（RL）与生成检索器的共演化，以提升信息检索效果。通过结合查询扩展和数据合成，该方法优化了检索过程，提升了系统对用户查询的响应准确性，表明RL在改进信息检索系统中的潜力，影响广告和搜索引擎等应用。
Language Models Can Control Their Own Attention
👍 51
Language models spend most of their attention on a small fraction of context, yet they read the entire KV cache to find the few tokens that matter. If the user asks about a previous detail in a 1M-token conversation, global attention layers must scan the full context to generate each token of the re
中文介绍 本研究揭示语言模型在信息获取过程中如何自我控制注意力。模型在处理大量上下文时通常仅对少部分关键信息进行注意，此特性提升了长文本对话生成的效率，具有优化对话系统和交互式应用的潜力，属于AI推理与理解领域的重要研究方向。
From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix
👍 33
Data-residency constraints force enterprises to self-host LLMs, but continuous adoption of newer models without decommissioning their predecessors expands the serving fleet, fragmenting a finite GPU pool. We consolidate traffic from over 200 internal applications onto a single model by closing quali
中文介绍 研究自托管大规模语言模型（LLM）的需求，提出在多种内部应用间整合流量的解决方案。通过压缩模型数量，确保在固定GPU资源下保持高效性，推动企业在AI部署和应用中应对数据隐私的挑战，获得高效的AI服务。
On the Design Fundamentals of Pixel Text Representation Learning
👍 28
Text-rich visual inputs require models that can read, retrieve, and compress language directly in pixel space, yet existing pixel-text encoders struggle with fixed resolution pretraining, visual shortcut learning, weak visual grounding, and multilingual visual text understanding. In this work, we in
中文介绍 本论文专注于像素文本表示学习的设计基础，指出现有编码器在多语言视觉文本理解和固定分辨率预训练方面面临的挑战。研究提供了一种新的框架，目标在视觉系统中实现更准确的文本解析与理解，为视觉-语言伴随任务的发展奠定基础。
Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering
👍 26
Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employ CLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semanti
中文介绍 提出了一种基于实体对齐的检索机制，以提升知识基础视觉问答（KB-VQA）的表现。研究指出，传统检索系统过于依赖表面视觉相似度，无法有效理解长尾实体的语义，强调重视实体层面语义信息在复杂问答系统中的重要性，推动更智能的视觉问答技术发展。
S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?
👍 26
Large language models (LLMs) increasingly interact with external environments and accumulate substantial behavioral experience, yet existing agent benchmarks largely evaluate them as fixed policies. It therefore remains unclear whether an agent can actively test its behavior, judge the resulting exp
中文介绍 S3Gym方法探讨大规模语言模型（LLMs）能否通过自我测试与自我评估实现自我改善。研究显示，模型在不断的外部环境交互中，有潜力在动态任务中自我优化，推动自适应AI代理的发展，具有重要应用于智能体的潜力。
ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes
👍 20
Compact token sequences are essential for efficient 3D generation. However, existing 3D tokenizers typically organize latent representations either over spatial regions or as fixed-size sets of global tokens, both suffering sharp reconstruction degradation when compressed to extremely low token budg
中文介绍 ZipTok3D提出了一种高保真3D令牌化方法，通过紧凑令牌前缀实现高效3D生成。该方法解决了现有3D令牌化器在极低令牌预算下的重构损失问题，为提升3D生成效率和精度提供了新思路，具有重要的实际应用价值。
AgentJudgeBench: A Multi-Difficulty Benchmark for Evaluating LLM Judges on Agentic Tool-Calling
👍 19
LLM judges are widely used to evaluate agentic tool-calling systems, yet their reliability on structured, dependency-driven workflows remains largely unexamined. We present AgentJudgeBench, the first benchmark to systematically study LLM-as-a-judge reliability for agentic tool-calling over workflow
中文介绍 AgentJudgeBench是针对评估LLM作为评审在代理工具调用中的可靠性而设计的基准测试。研究重点考察结构化依赖驱动工作流的可靠性问题，为评估LLM在复杂任务中的表现提出了标准，推动智能工具评估与安全性研究的进展。
DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory
👍 18
Self-play is an effective paradigm for language-model self-evolution, but without guidance, solver performance can plateau or decline across rounds. Unguided methods steer question generation with signals such as difficulty, learnability, or diversity. These signals keep questions challenging and va
中文介绍 DiagEvo方法通过层次错误记忆引导自我进化，旨在提升语言模型的自我学习能力。该方法通过有针对性的问题生成，确保模型在自我游戏中的表现始终处于挑战状态，推动自我学习系统的有效改进，对理性推理研究具有启示意义。
Cliff: Learning Process Rewards from the First Mistake
👍 15
Reinforcement learning with verifiable rewards (RLVR) has emerged as a powerful paradigm for large language model (LLM) post-training, but its reliance on coarse outcome rewards leads to limited guidance on intermediate reasoning processes. Existing approaches such as process reward modeling and on-
中文介绍 Cliff探讨了如何从初次错误中学习过程奖励，以提高强化学习后的模型表现。研究强调过程奖励建模对中间推理过程的重要性，提出了新的方法来优化其指导能力，具有推动RL与推理算法结合的广泛应用潜力。
Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation
👍 15
Sampled-token on-policy distillation (OPD) efficiently transfers capabilities from teacher to student using student-generated tokens, requiring teacher probabilities only for sampled tokens. Yet it frequently suffers from diversity distillation failure: the student's pass@1 improves while its pass@k
中文介绍 Influence-Directed Distillation方法通过解决抽样令牌政策蒸馏的多样性瓶颈，提升教师与学生模型间的能力转换。该方法为改善蒸馏产生的多样性问题提供了解决方案，推动高效模型转换和提升学习效率，具有重要的应用前景。
Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement
👍 13
This paper studies autonomous software development, in which LLM-based coding agents transform high-level requirements into complete, functional, and usable software systems without human intervention. We introduce Harness-of-Harness (HoH), a framework that enables coding agents to continually impro
中文介绍 Harness-of-Harness研究无人工干预的自主软件开发，提出框架使得LLM基础的编程代理能够持续改善。该方法强调在编码需求与功能实现之间形成闭环，推动了自动化软件开发的前沿技术应用，对软件工程领域的发展具有重要意义。
VibeVoice-ASR-Streaming Technical Report
👍 10
Traditional speaker-attributed ASR systems treated ASR and speaker diarization as two separate tasks. Recently, end-to-end models such as VibeVoice-ASR have unified the two tasks within a single model. However, existing unified models still mainly support offline recognition, making it difficult to
中文介绍 VibeVoice-ASR通过将说话人归属ASR与说话人分离化整合为一个模型，旨在提升实时语音识别的效率。该方法克服了传统模型离线识别的局限性，将为语音助手和实时交互应用的发展提供更多支持，推动语音技术的进步。
WHALE: A Simple Recipe for Joint Harness-Weight Optimization
👍 9
Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness u
中文介绍 WHALE提出了联合优化执行环境和模型权重的新方法，强调二者的相互依赖性。通过优化两个组件的结合，提升了智能代理在工作中的表现，推动了AI系统开发和执行效率的研究，具有重要的应用前景。
Agents in the Large: Perception-Centered Architecture for Persistent Agents
👍 9
Cognitive language agents have achieved substantial progress by equipping language models with memory, tools, and decision-making procedures, enabling agents to reason and act in interactive environments. Existing frameworks largely cast these agents as systems for solving user-specified, bounded ta
中文介绍 本研究关注持久性代理的感知中心架构，探讨如何通过结合记忆、工具及决策机制，提升认知语言代理在动态环境中的表现。该研究推动了智能体的互动式应用与复杂任务处理能力，对未来的动态AI系统设计具有深远影响。
CRISP: Cliff-awaRe Input-adaptive Sparse Prefilling with Structural-Mass-Motivated Routing
👍 8
The attention prefilling phase of long-context LLM inference scales quadratically, making self-attention a severe computational bottleneck. Traditional sparse attention methods mitigate this through fixed patterns or offline profiling, but lack the flexibility to adapt to input-dependent attention s
中文介绍 CRISP方法提出了一种输入自适应的稀疏预填充机制，旨在缓解长上下文LLM推理中的计算瓶颈。研究探索了基于结构化质量动机路由的新范式，为提升LLM在处理长文本时的灵活性和效率提供了有益的思路，具有广泛应用潜力。
Nobody is talking seriously about AI demand
@giovannicatt3 · 1.4K 粉丝 · 142.0K 阅 · 507 赞 · 56 转
AI capex for 2028 is forecast to be larger than the budget of France. Frontier AI labs’ revenue ramp justifies almost any number. Reflexivity in AI demand is a double-edged sword, and it's now a good
中文介绍 根据预测，2028年人工智能的资本支出将超过法国的预算，前沿AI实验室的收入大幅增加几乎可以支持任何数字。人工智能需求的反身性是把双刃剑，当前对此讨论的声音较少。
Prediction: AI will collapse
@wordgrammer · 26.1K 粉丝 · 120.8K 阅 · 602 赞 · 33 转
The dot com bubble got some things right. It noticed that the internet was going to be big. It noticed that the value generated by the internet would roughly scale in proportion with the amount of
中文介绍 回顾互联网泡沫，虽然预见了互联网的巨大潜力，但也反映出价值创造与用户规模之间的关系并非简单。此条帖子指向对人工智能未来可能崩溃的担忧，揭示行业对过去经验的反思。
Compilers 2.0: AI as stochastic optimizer
@cdleary · 2.3K 粉丝 · 100.3K 阅 · 533 赞 · 66 转
There has been a lot of discussion following the presentation of the Jalapeño MLA kernel at HotChips and subsequent commentary by SemiAnalysis. As OpenAI’s hardware team, we just barely touched on
中文介绍 讨论围绕Jalapeño MLA内核展开，强调人工智能作为随机优化器的潜力。OpenAI硬件团队的相关评述指出，与该领域发展相对应的技术进步应该引起重视。
7 Grok Bots for Marketing
@irabukht · 18.4K 粉丝 · 63.3K 阅 · 536 赞 · 28 转
18 things to know and 7 paste-in prompts for running ads, SEO and GEO on Grok Bot. Collected from 1,000+ marketers in our community. Most of us use Grok Bot to run marketing work end to end: Weekly
中文介绍 为营销活动提供17个重要信息和7个粘贴式提示，基于1,000多名市场营销人员的反馈，分享了使用Grok Bot进行广告、SEO和地理定位的经验，展示了在行业中普遍采用的工具和工作流。
Introducing agentic video understanding with Gemini
@GoogleAIStudio · 198.5K 粉丝 · 37.4K 阅 · 522 赞 · 48 转
Our new agentic feature for video analysis cuts costs by up to 66% and reduces token consumption by up to 88% while boosting accuracy. Today, we’re launching agentic video understanding across our
中文介绍 推出的新视频分析功能可将成本降低高达66%，令令牌消耗减少88%，同时提升准确率。这反映了产业对高效视频理解工具的迫切需求，展示了技术进步的实际应用。
Using Grok Bot: 8 templates to get inspired
@mattyp · 57.4K 粉丝 · 29.7K 阅 · 508 赞 · 40 转
Templates are now available for Grok Bot, so I wanted to share my favorites. Templates contain the skills, memories, and official @bot plugins from your bots. When you use a template, you get a copy
中文介绍 提供Grok Bot的8个模板以激发灵感，模板中包含技能、记忆和官方插件，方便用户快速上手。此内容体现了对标准化工作流工具的需求日益增长。
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
@GoogleAIStudio · 198.5K 粉丝 · 26.2K 阅 · 507 赞 · 59 转
Our newest Gemini models deliver next-generation intelligence for agentic workflows and cybersecurity. Building on the momentum of 3.7 Flash from three weeks ago and marking our third Flash release in
中文介绍 最新推出的Gemini 3.8 Flash系列模型增强了代理工作流与网络安全领域的智能化，标志着该系列持续的技术迭代，为市场提供了更强的竞争力。
From rg to zg: Local Search Beyond Keywords
@QwenDevs · 10.9K 粉丝 · 21.4K 阅 · 509 赞 · 55 转
Summary: The information humans and agents need is often scattered across large numbers of local files, making it difficult to locate accurately and efficiently. zg (zvec-grep) is local-first search
中文介绍 针对文件散布信息的搜索难题，提出zg（zvec-grep）作为一种新场景下的本地优先搜索解决方案。此内容反映了对高效信息检索工具的市场需求愈发明显。
GPT-6 Astra: an automated AI Engineer you can hire for <$6 an hour
We spent 20B+ tokens of GPT-6 Astra to explore everything. Here’s our learnings.
中文介绍 GPT-6 Astra是一个自动化的AI工程师，雇佣成本低于每小时6美元。项目使用了超过200亿个标记进行探索，取得了重要发现。
Introducing WeatherNext 3, our most advanced and accurate global weather AI model
中文介绍 DeepMind推出WeatherNext 3，这是其最先进和精准的全球天气AI模型。
Daybreak for Frontline Defenders: $1B to protect essential services
OpenAI introduces Daybreak for Frontline Defenders. A $1 billion commitment expands access to frontier cyber AI, training, and support for essential services.
中文介绍 OpenAI推出“前线防御者的曙光”计划，承诺提供10亿美元资金，以扩大前沿网络AI、培训和必要服务的获取。
NeoMME: an efficient Multimodal-native and Multilingual Encoder
中文介绍 Hugging Face发布了NeoMME，一个高效的多模态本地化和多语言编码器。
Legora reviewed 41 documents in minutes with GPT-6 Astra
Legora used GPT-6 Astra to review 41 documents in minutes, find all four planted errors, and improve performance by nearly 40% in this financial-review workflow.
中文介绍 Legora利用GPT-6 Astra在几分钟内审查了41份文件，发现了所有四个植入的错误，提高了近40%的性能。
Playco cut manual fixes 50% prototyping games with GPT-6 Astra
Using GPT-6 Astra, Playco built three themed game prototypes from one grey box foundation and reported 50% fewer manual fixes than with the previous model.
中文介绍 Playco使用GPT-6 Astra构建了三个主题游戏原型，手动修复减少了50%，相比之前的模型有明显优化。
[AINews] Muse Spark 1.3 matches GPT-5.6-Sol, confirming Meta Superintelligence as the newest Frontier Lab, >90% discount for training
an epic comeback story for Meta
中文介绍 AINews称，Muse Spark 1.3与GPT-5.6-Sol匹配，Meta的超级智能被确认为最新的前沿实验室，培训费用超过90%优惠。
Safety overview: GPT-6 Astra
GPT-6 Astra is our most capable broadly deployed model and our first to reach the Critical level of cybersecurity capability under our Preparedness Framework.
中文介绍 GPT-6 Astra是OpenAI部署能力最强的模型，也是首个在安全准备框架下达到关键网络安全能力级别的模型。
Fine-tuning a 350M Model for Better Structured Outputs in 100 GRPO Steps
中文介绍 Hugging Face针对一个350M模型进行了微调，以更好地输出结构化结果，采用100个GRPO步骤。
Give Your Coding Agents a Memory You Own
中文介绍 Hugging Face允许开发者为代码代理赋予属于自己的记忆。
Training a coding model to paint watercolours with TRL and OpenEnv
中文介绍 Hugging Face通过TRL和OpenEnv训练一个编码模型，使其能够绘制水彩画。
Muse Spark 1.3 ✨, Gemini 3.8 Flash ⚡, intelligence vs cost 📊
中文介绍 Muse Spark 1.3和Gemini 3.8 Flash推动了AI的智能与成本之间的比较。
Proactive cyber defense for governments and enterprises
中文介绍 DeepMind提供针对政府和企业的主动网络防御解决方案。
Introducing Gemini 3.8 Flash and 3.8 Flash Cyber
中文介绍 DeepMind发布了Gemini 3.8 Flash和3.8 Flash Cyber解决方案。
Facilitating AI integration with simplicity at scale
As companies scale, the technology supporting operations can become a liability just as quickly as it becomes an asset. Disconnected systems, site-specific tools, spreadsheets, and manual workarounds can create data silos that make it harder to spot problems early, coordinate responses, and make dec
中文介绍 MIT科技评论讨论了在企业规模扩展中实现AI集成的简便性与复杂性，强调如何避免数据孤岛。
根据当前市场情况，美股表现整体偏强，特别是像 Microsoft 和 Apple 这样的科技股显示出上行趋势。加密市场也较为活跃，Bitcoin 的当前市值和主导率分别为 81185.91 美元和 59.4%，呈现出一定的贪婪情绪，恐慌贪婪指数为 65。尽管部分资产如 Bitcoin 出现超买风险，但整体加密总市值达到 2.74 万亿美元，显示市场仍有强劲活力。在商品市场，黄金和原油期货维持稳定，WTI 原油期货表现稍好，呈现出多头走势。总体来看，市场走势较为复杂，仍需关注各类资产的不同信号。
今日关注
MSFT
Microsoft
偏上行
Microsoft 当前价格为 510.12 美元，RSI 为 66.1，接近超买区域，同时 MACD 指标显示金叉，且其 SMA 列表呈现多头排列，显示出强势上行趋势。
BTC-USD
Bitcoin
偏下行
Bitcoin 当前价格为 81185.91 美元，随着 RSI 达到 72.9，已进入超买状态，同时其 MACD 显示出红柱减小的趋势，表明价格可能面临回调风险。
AAPL
Apple
中性
Apple 当前价格为 328.21 美元，RSI 为 63.4，接近超买的上限，同时价格处于当前 SMA 列表的多头排列中，短期出现调整幅度的可能性增加，因此保持中性态度。
全部资产
^VIX
VIX 恐慌指数
$14.32
-5.79%
- 5 日
- -1.31%
- 距 52w 高
- -59.4%
- RSI(14)
- 43.8
- 趋势
- 空头
- SMA 20 / 50 / 200
- 15.12 / 16.29 / 18.30
- MACD / 信号
- -0.392 / -0.432
MACD 金叉 (3 天前)空头排列
^TNX
10Y 美债收益率 (%)
$4.76
-0.71%
- 5 日
- +1.93%
- 距 52w 高
- -0.9%
- RSI(14)
- 58.1
- 趋势
- 多头
- SMA 20 / 50 / 200
- 4.70 / 4.63 / 4.36
- MACD / 信号
- 0.037 / 0.031
MACD 金叉 (2 天前)接近 52 周高多头排列
DX-Y.NYB
美元指数 DXY
$99.00
-0.56%
- 5 日
- -0.16%
- 距 52w 高
- -2.7%
- RSI(14)
- 39.9
- 趋势
- 中性
- SMA 20 / 50 / 200
- 99.42 / 100.28 / 99.17
- MACD / 信号
- -0.243 / -0.297
MACD 金叉 (4 天前)接近 52 周高
SPY
S&P 500 ETF
$773.17
+1.05%
- 5 日
- +0.27%
- 距 52w 高
- -0.8%
- RSI(14)
- 58.8
- 趋势
- 多头
- SMA 20 / 50 / 200
- 769.21 / 756.14 / 711.63
- MACD / 信号
- 3.071 / 3.900
接近 52 周高多头排列
QQQ
Nasdaq 100 ETF
$717.67
+1.19%
- 5 日
- -0.48%
- 距 52w 高
- -4.1%
- RSI(14)
- 52.9
- 趋势
- 多头
- SMA 20 / 50 / 200
- 717.71 / 711.03 / 657.13
- MACD / 信号
- 0.863 / 1.534
多头排列
AAPL
Apple
$328.21
+1.00%
- 5 日
- +4.33%
- 距 52w 高
- -4.7%
- RSI(14)
- 63.4
- 趋势
- 多头
- SMA 20 / 50 / 200
- 312.81 / 314.25 / 283.61
- MACD / 信号
- 2.672 / 0.644
多头排列
MSFT
Microsoft
$510.12
+2.68%
- 5 日
- +1.00%
- 距 52w 高
- -7.9%
- RSI(14)
- 66.1
- 趋势
- 多头
- SMA 20 / 50 / 200
- 495.72 / 441.23 / 431.09
- MACD / 信号
- 16.345 / 18.236
金叉(SMA50↑SMA200) (3 天前)多头排列
NVDA
Nvidia
$228.45
+1.80%
- 5 日
- +0.21%
- 距 52w 高
- -3.4%
- RSI(14)
- 59.2
- 趋势
- 多头
- SMA 20 / 50 / 200
- 219.76 / 209.87 / 196.46
- MACD / 信号
- 3.255 / 2.785
MACD 金叉 (今天)多头排列
GOOGL
Alphabet
$342.48
+1.59%
- 5 日
- +0.54%
- 距 52w 高
- -16.2%
- RSI(14)
- 47.8
- 趋势
- 中性
- SMA 20 / 50 / 200
- 344.40 / 348.67 / 335.67
- MACD / 信号
- -2.906 / -2.498
TSLA
Tesla
$376.37
+5.42%
- 5 日
- +6.08%
- 距 52w 高
- -24.6%
- RSI(14)
- 60.5
- 趋势
- 中性
- SMA 20 / 50 / 200
- 347.16 / 358.37 / 399.83
- MACD / 信号
- 4.069 / 0.002
META
Meta
$610.68
+3.01%
- 5 日
- +6.93%
- 距 52w 高
- -22.8%
- RSI(14)
- 60.7
- 趋势
- 中性
- SMA 20 / 50 / 200
- 575.65 / 594.01 / 622.22
- MACD / 信号
- -1.070 / -6.688
BTC-USD
Bitcoin
$81,185.91
+5.03%
- 5 日
- +3.76%
- 距 52w 高
- -35.7%
- RSI(14)
- 72.9
- 趋势
- 中性
- SMA 20 / 50 / 200
- 74,774.97 / 68,456.89 / 69,567.77
- MACD / 信号
- 3,487.049 / 3,477.935
MACD 金叉 (今天)RSI 超买
ETH-USD
Ethereum
$2,502.02
+4.62%
- 5 日
- +1.80%
- 距 52w 高
- -47.5%
- RSI(14)
- 67.4
- 趋势
- 多头
- SMA 20 / 50 / 200
- 2,330.14 / 2,065.20 / 2,031.31
- MACD / 信号
- 129.297 / 138.370
金叉(SMA50↑SMA200) (3 天前)MACD 死叉 (2 天前)多头排列
SOL-USD
Solana
$104.06
+3.66%
- 5 日
- -1.51%
- 距 52w 高
- -58.9%
- RSI(14)
- 67.6
- 趋势
- 多头
- SMA 20 / 50 / 200
- 94.22 / 82.68 / 82.15
- MACD / 信号
- 6.651 / 6.587
金叉(SMA50↑SMA200) (1 天前)多头排列
BABA
阿里巴巴 (BABA)
$111.81
+0.00%
- 5 日
- -3.87%
- 距 52w 高
- -42.0%
- RSI(14)
- 38.1
- 趋势
- 空头
- SMA 20 / 50 / 200
- 121.74 / 115.80 / 135.68
- MACD / 信号
- -1.872 / -0.094
空头排列
PDD
拼多多 (PDD)
$81.63
-0.75%
- 5 日
- -3.61%
- 距 52w 高
- -41.4%
- RSI(14)
- 36.9
- 趋势
- 空头
- SMA 20 / 50 / 200
- 86.93 / 85.43 / 98.28
- MACD / 信号
- -1.070 / -0.249
空头排列
JD
京东 (JD)
$27.74
-0.32%
- 5 日
- -2.63%
- 距 52w 高
- -24.7%
- RSI(14)
- 33.1
- 趋势
- 中性
- SMA 20 / 50 / 200
- 29.50 / 29.41 / 29.14
- MACD / 信号
- -0.670 / -0.434
金叉(SMA50↑SMA200) (8 天前)
0700.HK
腾讯控股 (0700.HK)
HK$433.00
-1.19%
- 5 日
- -3.31%
- 距 52w 高
- -36.6%
- RSI(14)
- 39.7
- 趋势
- 空头
- SMA 20 / 50 / 200
- 450.70 / 454.78 / 517.93
- MACD / 信号
- -5.401 / -3.870
空头排列
GC=F
黄金期货
$4,519.50
+3.51%
- 5 日
- -1.96%
- 距 52w 高
- -19.1%
- RSI(14)
- 58.3
- 趋势
- 中性
- SMA 20 / 50 / 200
- 4,464.09 / 4,231.55 / 4,518.79
- MACD / 信号
- 65.385 / 85.988
MACD 死叉 (3 天前)
CL=F
WTI 原油期货
$91.85
+0.92%
- 5 日
- +9.96%
- 距 52w 高
- -23.1%
- RSI(14)
- 65.8
- 趋势
- 多头
- SMA 20 / 50 / 200
- 84.80 / 80.58 / 78.66
- MACD / 信号
- 2.077 / 1.341
MACD 金叉 (2 天前)多头排列
USDCNY=X
美元 / 人民币
¥6.71
-0.16%
- 5 日
- -0.19%
- 距 52w 高
- -6.7%
- RSI(14)
- 25.7
- 趋势
- 空头
- SMA 20 / 50 / 200
- 6.73 / 6.76 / 6.86
- MACD / 信号
- -0.011 / -0.011
MACD 金叉 (2 天前)RSI 超卖接近 52 周低空头排列
本报告仅供技术指标解读参考，过去走势不代表未来表现。投资者在做出投资决策时应谨慎行事，考虑市场风险。
Rogers Stadium in Toronto Is Severely Damaged in Thunderstorm
The music venue, which opened in 2025, was built specifically for large concerts. The fate of its next show, featuring AC/DC, is unclear.
中文摘要 多伦多的罗杰斯体育场在雷暴中遭受严重损坏，该场馆于2025年开幕，专门用于大型音乐会。AC/DC乐队的下一场演出是否能够如期进行尚不明朗。
Tunisia’s top court upholds conviction of opposition figures, activists
Critics claim Tunisia’s judiciary acts under political pressure, as opposition leaders face lengthy prison terms.
中文摘要 突尼斯最高法院维持了对反对派人士和活动家的定罪，批评者认为突尼斯司法系统受到政治压力，反对派领导人面临较长的监禁。
News live: Two more Australians confirmed safe in Nepal-Tibet floods; Bowen calls joint statement on China ‘strong’
Follow the day’s news live. Get our breaking news email, free app or daily news podcast ‘Gut-wrenching’: Bowen to meet with Nepali community today Chris Bowen said he would meet members of the Nepali Australian community later today. Every person who’s been rescued or found is a relief to their fami
中文摘要 跟踪当天新闻。所有被救出或找到的人都被认为是安全的，澳大利亚副总理Chris Bowen今天将与尼泊尔社区会面。
DR Congo joins countries pledging embassy move to Jerusalem
DR Congo has renewed its pledge to move its embassy in Israel from Tel Aviv to Jerusalem.
中文摘要 刚果民主共和国重申将其驻以色列大使馆从特拉维夫迁移至耶路撒冷的承诺。
Transplanted pig kidney works in US man's body for record 271 days
A man has lived for a record nine months with a transplanted pig kidney while waiting for a human transplant.
中文摘要 一名男子在等待人类器官移植期间，成功接受了猪肾移植并在体内存活了创纪录的271天。
Volkswagen announces it will cut 100,000 jobs by 2030
Under-pressure carmaker will shed 15% of workforce and halve its product line in sector’s biggest ever restructure The car company Volkswagen has announced it will shed 100,000 jobs by the end of the decade after being hit by US tariffs and fierce competition from Chinese rivals. The German manufact
中文摘要 大众汽车宣布到2030年将裁减10万个工作岗位，占其员工的15%，并将减半其产品线，此次重组是汽车行业最大规模的调整之一。
US investigating if missile hit Iran wedding, Vance says
The Iranian Red Crescent Society (IRCS) said shrapnel from a missile hit the ceremony and killed four people on Tuesday.
中文摘要 伊朗红新月会表示，星期二一场婚礼庆典遭到导弹碎片袭击，造成四人遇难，现正对事件进行调查。
Serbians targeted with spyware in country’s ‘largest documented wave of surveillance’
Student protesters among those hacked, says digital rights group, though Aleksandar Vučić’s government denies spying At least 14 people from across Serbian civil society were targeted with advanced spyware earlier this year in what the digital rights group Share Foundation said was the largest docum
中文摘要 数字权利组织称，今年早些时候，至少14名塞尔维亚公民在该国遭到先进间谍软件的攻击，尽管亚历山大·武契奇政府否认进行监控。
Why are people calling Venezuela oil deal a new form of US colonialism?
Critics call the US multi-billion-dollar oil deal in Venezuela modern-day colonialism.
中文摘要 批评者称美国在委内瑞拉的数十亿美元石油交易为现代殖民主义。
Burnham, Macron talk border security and stronger Europe-UK ties
New UK Prime Minister Andy Burnham discussed shared challenges in his first official meeting with the French president.
中文摘要 英国新任首相安迪·伯纳姆与法国总统马克龙首次正式会面，讨论边境安全和加强英欧关系的共同挑战。
NYC’s AI ban through grade 8 sets stage for rest of US, experts say
The ban will last for one year and affect 600,000 students.
中文摘要 纽约市对八年级及以下学生的人工智能禁令将持续一年，影响60万名学生，被专家认为将为美国其他地区设定标准。
Norway Seizes Russian Ship, Testing Moscow in the High Arctic
Empowered by an international court order, Norway captured the vessel as part of a global effort to seize Russian assets. The Russians called it “piracy.”
中文摘要 挪威依靠国际法庭的命令拘留了一艘俄罗斯船只，这是全球努力扣押俄罗斯资产的一部分，俄罗斯方面称其为”海盗行为“。
Nepal flood victims to be identified through DNA
Nepal flood victims to be identified through DNA
中文摘要 尼泊尔将通过DNA鉴定来识别洪水受害者。
El Nino: A phenomenon supercharged by climate change
Scientists warn the 2026-27 El Nino could become the strongest ever, exacerbating global heat, droughts and floods.
中文摘要 科学家警告，2026-27年的厄尔尼诺现象可能成为有史以来最强，可能加剧全球的高温、干旱和洪水。
US Vice President JD Vance declines to offer timeline for end of Iran war
The Trump official expressed scepticism at White House briefing about the conflict being a war despite recent flare-up.
中文摘要 美国副总统JD·万斯在白宫简报会上对伊朗战争未来结束的时间表示怀疑，尽管近期冲突加剧。
From cake to pizza: How to get the best birthday freebies (and why there's a catch)
Cake, coffee, make-up and burgers are just some of the free things you can get on your birthday.
中文摘要 在生日当天，消费者可以免费获得蛋糕、咖啡、化妆品和汉堡等多种礼品。
From cake to pizza: How to get the best birthday freebies (and why there's a catch)
Cake, coffee, make-up and burgers are just some of the free things you can get on your birthday.
中文摘要 在生日当天，消费者可以免费获得蛋糕、咖啡、化妆品和汉堡等多种礼品。
Controversial Jackdaw gas field set to be approved in weeks, sources say
The government is set to give the go-ahead for the gas field off the coast of Aberdeen, the BBC understands.
中文摘要 英国政府即将批准阿伯丁海岸的杰克道天然气田，预计该决定将在数周内作出。
Hong Kong’s Picky IPO Issuers Rewrite Playbook for Allocations
Hong Kong’s boom in initial public offerings has turned getting shares in the hottest deals into something of an invitation-only affair.
中文摘要 香港的首次公开募股热潮使得热门交易的股份获取变得愈加困难，很多交易采取邀请制。
Anthropic Nears Finalizing $15 Billion Pre-IPO Credit Facility
Anthropic PBC is set to finalize expanding its revolving credit facility to $15 billion, according to people familiar with the matter, clearing a hurdle before the artificial intelligence firm’s public filing for its highly anticipated IPO.
中文摘要 Anthropic PBC将推进其循环信贷额度扩大至150亿美元，此举为该公司即将进行的IPO扫清障碍。
Asian Stocks to Climb as Fed Bets Cool, Yen Gains: Markets Wrap
Asian stocks were poised to rise, tracking gains on Wall Street as investors pared bets on a Federal Reserve interest-rate hike this month. The yen held its biggest advance in more than a month.
中文摘要 亚洲股市有望上涨，因投资者减弱了对美联储本月加息的押注，日元也迎来了一个多月来的最大涨幅。
Latest Oil Market News and Analysis for Sept. 4
Oil headed for its biggest weekly gain since July as renewed US-Iran hostilities heightened concerns about prolonged disruptions to energy flows through the Strait of Hormuz.
中文摘要 由于美国与伊朗的紧张局势加剧，油价本周有望实现自七月以来的最大涨幅，市场担忧霍尔木兹海峡的能源运输受阻。
Lululemon Needs a Full Product Overhaul, Widlitz Says
SW Retail Advisors President and founder Stacey Widlitz discusses Lululemon Athletica lowering its full-year outlook for a second straight quarter on "Bloomberg The Close." (Source: Bloomberg)
中文摘要 Lululemon Athletica连续第二个季度下调全年业绩预测，SW Retail Advisors创始人Stacey Widlitz对此进行了讨论。
SEC Moves to Nix Rule on Adviser Political Donations
Investment advisers would no longer be prohibited from offering services to public pension funds if they made recent political donations to state and local elected officials, the Securities and Exchange Commission said in a proposed rule change Thursday.
中文摘要 美国证券交易委员会提出的新规则将取消对投资顾问向公共养老金捐赠的禁令，允许其提供相关服务。
Toyota’s Sato on How Japanese Automakers Stay Relevant
Toyota Vice Chairman and Japan Automobile Manufacturers Association Chairman Koji Sato sits down for an exclusive Bloomberg Television interview with Shery Ahn to discuss how Japanese automakers can stay relevant globally, as well as Toyota's hydrogen business. (Source: Bloomberg)
中文摘要 丰田副董事长佐藤浩司在采访中讨论了日本汽车制造商保持全球竞争力的方法，以及丰田的氢能业务。
The Hottest New Trade Is Your Own Electricity
Could the hottest new trade be your own electricity? Zero host Akshat Rathi meets the vibe coders making money from their excess energy. Listen to the Big Take podcast. (Source: Bloomberg)
中文摘要 新兴的交易方式是个人的电力利用，Zero节目的Akshat Rathi探讨了如何从多余的电力中获利。
Australia’s Battered Bonds Have Reasons to Rebound, Funds Say
Australia’s bonds, caught up in the global fixed-income selloff, have seen their yields rise more than any of their peers in the past month, despite arguments that the country’s finances are in much better shape.
中文摘要 澳大利亚债券在全球固定收益市场大跌中受到影响，尽管财富状况较好，但收益率在过去一个月内升幅超过同行。
一语成美图的 SKILL
5 回复 · 程序员 节点
为了让团队里每个人的 AI 之间可以更好的协作和共享业务上下文，我做了一个开源产品
6 回复 · 程序员 节点
如何让多不同模型在同一个 session 中与自己一起讨论问题？
10 回复 · 程序员 节点
无法使用 google one 怎么解决
7 回复 · 程序员 节点
感觉现在关于 github copilot 的讨论少了，还有人在主力使用的吗？想听听实际体验
33 回复 · 程序员 节点
电子产品大涨价，大家还会正常更新数码设备吗？
34 回复 · Apple 节点
MAC Studio M5 Ultra 被延期发货
30 回复 · Apple 节点
有人用过 OCLP 给老电脑强制升级 macos 吗
12 回复 · Apple 节点
各位 Apple 的 Inteler 是时候说再见了～
14 回复 · Apple 节点
美区苹果账户如何解封
11 回复 · Apple 节点
该源今日无内容。