每日简报

2026-08-11

Morning Brief

2026-08-11

今天在AI领域,广告创新和自动化代理推动科技变革,而金融市场则在中国企业盈利中经历轮换,政治上则面临重大人道挑战。

今日最重要的 5 件事

  1. 1
    技术动态OpenAI News

    OpenAI测试ChatGPT中的广告以支持免费访问

    为什么重要 此测试可能改变广告收入结构,使更多用户能够接触到AI服务,但也可能影响用户体验和信任。

  2. 2
    财经要点Bloomberg Markets

    中国大型科技公司盈利测试投资轮换

    为什么重要 这些数据将决定投资者的接下来策略,影响人工智能与互联网股票的资金流向。

  3. 3
    时政观察Al Jazeera

    哥伦比亚地震后搜救工作持续

    为什么重要 地震后的人道援助和救援工作突显了自然灾害对社会的影响力,亟需国际关注和援助。

  4. 4
    财经要点Bloomberg Markets

    全球投资者偏好台湾股票

    为什么重要 这种偏好反映出对市场稳定性和潜力的重新评估,对未来的投资决策具有指导意义。

  5. 5
    财经要点Financial Times

    意大利副总理提议对银行利润征超额税

    为什么重要 这一提议表明政府对经济不平等问题的关注,可能引发更广泛的财政策略讨论。

市场状态

Risk-Off
  • 5 个跟踪资产处于多头趋势,3 个处于空头趋势。
  • SPY 当前为 bullish,5 日 -0.10%,RSI 63.3。
  • 加密恐惧贪婪指数为 29(恐慌)。
  • 加密总市值 24h 变化为 -0.08%。

观察 观察 BTC 在 5 日 -1.06% 后能否企稳。

招聘信号

  • 升温 AI广告技术 OpenAI测试广告功能显示AI在广告领域的需求上升,预计将需要更多广告技术人才。
  • 升温 自动化代理开发 AI代理项目的推出暗示对开发和维护智能代理的高需求,提升相关技术人才的价值。
  • 升温 灾难响应与人道救助 哥伦比亚地震后救援行动开启了对应急管理人才的需求,尤其是在灾后二次恢复和社区重建方面。

论文趋势

X 讨论主题

Top Project

msitarzewski/agency-agents

Shell · ★ 143,148 · 🍴 23,271 · 📈 971 stars today

A complete AI agency at your fingertips - From frontend wizards to Reddit community ninjas, from whimsy injectors to reality checkers. Each agent is a specialized expert with personality, processes, and proven deliverables.

中文介绍 该项目提供了一整套AI代理,涉及前端开发、社群管理等多个领域,每个代理都是具有个性化和专业流程的专家。为什么重要:这表明AI正在被广泛应用于多种任务和领域,提高了自动化和效率。

Most Practical

semantica-agi/semantica

Python · ★ 4,843 · 🍴 526 · 📈 884 stars today

Graph-Native Infrastructure for Context and Accountable AI Systems

中文介绍 该项目旨在构建图结构化的基础设施,以支持上下文感知和可追溯的AI系统。为什么重要:它暗示了未来AI系统将在可解释性和负责性方面的趋势,推动更透明的决策过程。

Most Surprising

nvm-sh/nvm

Shell · ★ 94,483 · 🍴 10,353 · 📈 18 stars today

Node Version Manager - POSIX-compliant bash script to manage multiple active node.js versions

中文介绍 这是一个用于管理多个活动Node.js版本的工具,通过POSIX兼容的bash脚本进行操作。为什么重要:在开发过程中,它使得环境管理更加高效,适应不同版本的需求。

BDH-CQ: In-Context Learning with Recurrent Latent Reasoning

👍 241

We introduce BDH-CQ, a reasoning model that combines in-context learning with recurrent latent reasoning. Inputs presented at inference time continuously update the model's recurrent memory; the model then solves a query through iterative computation in a high-dimensional latent space, without verba

中文介绍 提出了 BDH-CQ 模型,将上下文学习与递归潜在推理相结合。该模型通过迭代计算在高维潜在空间中解决查询,并在推理时持续更新模型的递归记忆。核心创新在于不依赖于传统的监督方法,以实现更灵活的推理能力。意义:推动了在复杂推理任务中 Agent 能力的提升与应用。

Macaron-V1: Towards Open Continual Learning with Self-Improvement and Mixture-of-LoRA

👍 200

Macaron-V1 is an open agent-model family for experiential intelligence: learning from experience in real environments and continuing to learn after deployment. It is organized around two system goals. Adaptation is pursued through recursive improvement of versioned model-harness pairs, where experie

中文介绍 Macaron-V1 是一个开放的智能体模型家族,旨在通过在真实环境中的经验学习实现持续自我改进。该方法通过对版本化模型与工具组合的递归改进进行适应,推动智能体在部署后不断学习。意义:引领持续学习和适应性智能体的发展,提升智能体在现实环境中的表现。

SWE-Bench ProMax: Benchmarking Agents on Large-Scale Multilingual Code Refactoring

👍 116

As AI coding agents take on increasingly complex, long-horizon software engineering tasks, existing benchmarks are rapidly saturating and their evaluation quality has come under serious scrutiny: a recent audit found that nearly 60% of unsolved SWE-bench Verified instances contain flawed tests -- ei

中文介绍 SWE-Bench ProMax 旨在评估在大型多语言代码重构任务中 AI 编码代理的表现。该基准检测了现有基准的不足,特别是存在的测试缺陷,提供更可靠的评估标准。意义:为软件工程中的 AI 应用提供标准化评估,提高了任务复杂性的应对能力。

Ouroboros: A Self-Developing Frontier Coding Agent with Reviewed Core Evolution

👍 66

We present Ouroboros, a self-developing agent harness whose tools, prompts, context assembly, and core implementation improve through reviewed commits that become the runtime for later work. Core evolution proceeds in two modes. In recursive free evolution, improvement is itself a task, and completi

中文介绍 Ouroboros 是一个自我发展的编码代理,通过审查的提交来改进其工具、提示和核心实现,进而提升效率。该模型采用递归自由演化与任务驱动的改进方式,对智能体能力的不断增强提供了新的视角。意义:有效推动智能体自我优化技术的发展,适用于复杂编码任务。

On-Policy Self-Distillation without Any Supervision

👍 52

On-policy (Self-)Distillation (OPD / OPSD) has shown strong potential for post-training large language models (LLMs). However, existing methods still rely heavily on external supervision, including ground-truth signals, environmental feedback, or guidance from larger models, and therefore fall short

中文介绍 提出了一种无监督的 On-Policy Self-Distillation (OPSD) 方法,显示出对后训练大语言模型的强大潜力。该方法突破了依赖外部监督的限制,旨在提高模型性能,从而提升应用效果。意义:推动了无监督学习与模型自我优化方向的探索,可能影响多种智能应用。

Motif 3: Technical Report

👍 33

We introduce Motif 3, a decoder-only Mixture-of-Experts language model with 314 billion total parameters and 13.2 billion activated per token. Each sparse MoE layer contains 384 routed experts, with eight selected per token. This fine-grained sparsity provides substantial expert capacity while limit

中文介绍 Motif 3 是一个解码器-仅 Mixture-of-Experts 语言模型,拥有 3140 亿个参数,并在每个 token 上激活 132 亿个参数。该模型通过细粒度的稀疏性增强了专家容量,并提高了计算效率。意义:为大规模语言模型的技术发展提供新的范例,改善模型的推理与生成能力。

Stealing Reasoning Traces from Proprietary LLM APIs

👍 31

Leading large language model providers now conceal their models' step-by-step reasoning, or chain-of-thought, to protect intellectual property and limit information leakage. Rather than storing these traces server-side, providers return them to the client as blocks of encrypted text, which the clien

中文介绍 研究了从专有 LLM API 中提取推理痕迹的方法,挑战了当前模型提供商的知识产权保护。该方法强调了通过加密文本形式返回推理过程,突破了传统束缚。意义:推动机器学习模型透明度及开放研究方向,影响智能应用和知识产权管理的策略。

Agent Memory Distillation: Empowering Small LLM Agents with Hierarchical Teacher Memory

👍 31

Memory systems have shown promise for improving agent performance, but their potential remains largely unexplored for small language models, which struggle to generate sufficient successful trajectories on their own. We propose Agent Memory Distillation (AMD), a training-free framework that transfer

中文介绍 提出了 Agent Memory Distillation (AMD) 框架,用于改善小型语言模型的代理性能。AMD 在不需要额外训练的情况下,通过传递层次化教师记忆来提高智能体生成有效轨迹的能力。意义:为小型模型的性能提升提供新思路,有助于增强智能体在复杂任务中的适应能力。

MatrAIx: Simulating the World with 8.3 Billion Persona Agents

👍 27

Human evaluation of AI systems and digital products is costly, slow, and difficult to scale. Offline evaluations are more scalable but often abstract away human diversity and interactive behavior. We therefore introduce MatrAIx, a population-scale simulated-user evaluation infrastructure for testing

中文介绍 MatrAIx 是一个用于测试 AI 系统与数字产品的人类评估基础架构,解决了传统人类评估的规模与成本问题。通过模拟用户的方法,该系统可以更好地反映人类行为的多样性。意义:为模拟人类评估提供可扩展的解决方案,适用于 AI 和人机交互领域。

Sci-VBench: Evaluating Knowledge- and Reasoning-Intensive Video Generation in Science Domains

👍 23

We introduce Sci-VBench, a comprehensive benchmark for evaluating knowledge- and reasoning-intensive video generation across scientific domains. It contains 1,253 expert-annotated examples spanning 60 subjects across four core disciplines: Natural Science, Healthcare, Humanities & Social Sciences, a

中文介绍 Sci-VBench 是一个评估知识与推理密集型科学视频生成的综合基准,涵盖了多个学科的专家注释示例。其目标是提升科学领域视频生成模型的评估标准与应用潜力。意义:推动科学领域智能生成技术的发展,助力新兴的视频生成应用。

What to Edit Next: Visually Aligned Image-Editing Follow-Up Suggestions in Conversational Systems

👍 22

Conversational assistants increasingly recommend follow-up edits to help users continue a task. Existing systems primarily target text-only interactions, leaving image-creation conversations underexplored. In image-creation tasks, useful follow-up edit suggestions must reflect user preferences, offe

中文介绍 探讨了在图像创建任务中,如何根据用户偏好提供有用的后续编辑建议。该系统旨在提升现有对话助手在图像编辑中的效果,以适应用户的实时需求。意义:促进图像生成任务中的个性化指南的发展,影响用户体验的提升。

OasisKV: Scaling In-Decode KV Cache Beyond HBM with Lookahead Sparse Prefetching

👍 16

Large language model (LLM) inference serving is increasingly constrained by memory rather than compute. As long-context and long-form reasoning workloads become more prevalent, the key-value (KV) cache dominates both memory footprint and memory traffic during LLM token generation, i.e., decode. In p

中文介绍 OasisKV 提出了利用前瞻性稀疏预取方法来扩展关键值缓存的能力,以解决长上下文推理中的内存限制问题。该方法优化了在 LLM 解码过程中记忆流量的管理,从而提升了效率。意义:推动长上下文模型性能的提升,有助于复杂推理任务的实现。

SPOT: Sparse Probing and Outcome Calibration for On-Policy Distillation

👍 16

On-policy distillation (OPD) provides dense teacher supervision on student-generated trajectories, but standard reverse-KL training can assign insufficient probability to other plausible continuations. Teacher entropy alone does not reveal whether uncertainty is concentrated among a few plausible ne

中文介绍 SPOT 评估了 On-Policy Distillation (OPD) 中的稀疏探测与结果校准问题,以提高学生生成轨迹的监督质量。该研究识别出标准方法中的不足之处,提供了改进方向。意义:为强化学习中的优化方法提供新思路,有助于提高智能体的学习能力。

DCAS: Decoupling CLI Agent Scaffolding to Internalize Planning across Scaffolds

👍 16

CLI-based software-engineering agents have matured rapidly, yet the open ecosystem has converged on a single training environment: trajectory datasets used to fine-tune open models are collected almost exclusively under OpenHands. Models fine-tuned on this data score well under OpenHands but degrade

中文介绍 DCAS 通过解耦 CLI 代理支撑结构,提升了 Planning 跨支撑的内化能力。尽管 CLI 基础的智能体在快速发展,然而对训练环境的局限性却使模型适应性降低。意义:为软件工程代理技术的灵活性提供新路径,促进开放生态的进步。

Efficient Knowledge Distillation for LLMs: Offline Top-K Logits and a Fused Chunked KL Loss

👍 11

Small language models are often the only option for deployment under tight latency, cost, and on-premises constraints, but they are rarely trained from scratch: a compressed model is usually recovered through knowledge distillation (KD). This recovery step largely decides the final quality, yet it i

中文介绍 研究了高效的知识蒸馏方法,针对 LLM 在严格的延迟、成本和本地化约束下的应用,提出 Offline Top-K Logits 和融合分块 KL 损失。新方法改善了压缩模型的质量,推动模型在资源受限环境中的部署。意义:为低资源环境中的大语言模型应用提供支持,拓宽了 LLM 的应用场景。

Evo-Bench: Can Language Models Improve Agent Harness?

👍 10

Large Language Models (LLMs) have driven rapid progress in autonomous agents, yet standard evaluations remain confined to static task solving. An emerging frontier is harness evolution---the agent's capacity to autonomously optimize its own operating harness. However, systematically benchmarking thi

中文介绍 Evo-Bench 关注于大型语言模型 (LLM) 如何改善智能体创建与优化的能力,探索代理自身操作环境的自动优化。该研究提出了系统性的基准评估智能体的进化能力。意义:促进智能体自我优化的研究,推动任务复杂性应对能力的提升。

RoMeRL: Balancing Feedback Coverage and the Memory-Reward Trap in Self-Evolving Agent Memory via Reduced-Order Utility States

👍 9

Learning-based memory systems for self-evolving LLM agents face two tightly coupled challenges. First, trajectory-indexed utilities grow with the interaction history, thereby dispersing limited feedback over an ever-expanding state space. Second, because trajectory-level rewards are jointly assigned

中文介绍 RoMeRL 解决了有关学习基础的记忆系统的两大挑战,提出通过降阶效用状态来平衡反馈覆盖与记忆奖励的困境。该方法有助于提升智能体在复杂状态空间中的学习效率。意义:为自我演化智能体的学习机制提供理论基础,影响未来的智能体研发。

Evidence-RL: Towards Evidence-intensive Visual Reasoning

👍 9

Vision-Language Models (VLMs) should answer from concrete image evidence rather than language priors, dataset shortcuts, or irrelevant visual context. Existing perception-aware post-training methods encourage image use through global perturbations or attention proxies, but they do not test whether a

中文介绍 Evidence-RL 旨在增强视觉语言模型 (VLM) 基于具体图像证据的推理能力,从而提升模型对图像的理解与应用效果。该研究挑战现有依赖语言推理的流行方法,强调了基于证据的重要性。意义:推动视觉推理领域的进步,提高模型在处理实际应用中的能力。

Multi-Agent Forensic Reasoning for Generalizable Deepfake Video Detection

👍 9

The malicious use of generative artificial intelligence to create highly realistic deepfake videos raises serious ethical concerns and poses substantial challenges to AI safety. However, existing deepfake video benchmarks provide limited coverage of recent synthesis methods and generally lack reliab

中文介绍 提出了多智能体取证推理方法,以应对深度伪造视频所带来的挑战。通过集成化的评估框架,该研究强调了检测新型合成方法的重要性与必要性。意义:提升对深度伪造检测方法的有效性与实用性,增强 AI 安全性研究的方向。

Relevant but Incomplete: Referential Dangling as a Paradigm-Level Failure Mode in Hard Prompt Compression

👍 9

Hard prompt compression reduces long-context inference cost by independently scoring tokens, sentences, or chunks and retaining the highest-scoring units under a budget. We identify a structural failure in this procedure: independent selection can split dependent evidence pairs, retaining one member

中文介绍 探讨了硬提示压缩的缺陷,揭示了在高分散的选择过程中可能导致的参考未完整问题。该文提出了改进的方法,以优化选择机制并增强信息保留的稳定性。意义:为提高提示压缩技术的性能指明方向,影响多模态学习的研究。

How to Build a Shared AI Harness for Your Team

@VibeMarketer_ · 30.1K 粉丝 · 107.1K 阅 · 528 赞 · 29 转

I am going to show you how to turn the models, agents, skills, and automations living across your team's tools and workflows into one coordinated system. What most teams have today looks very

中文介绍 博主介绍如何将团队中的模型、代理、技能和自动化工具整合成一个协调一致的系统,帮助团队提升工作效率。这种方法针对目前很多团队面临的协调性不足的问题,反映出在团队协作中利用 AI 工具整合的趋势日益显著。

NVIDIA Nemotron 3.5 Lightning and NeMo Switchyard Deliver Faster, Smarter, More Efficient Agentic AI

@nvidia · 2.6M 粉丝 · 59.5K 阅 · 548 赞 · 66 转

The new lightweight open model and routing library delivers greater control over AI, data and workflows across edge devices, PCs, workstations, data centers and the cloud. Source: NVIDIA Blog By Kari

中文介绍 NVIDIA 发布了新一代轻量级开放模型和路由库,旨在提高 AI、数据和工作流的控制能力,适用于边缘设备、PC、工作站、数据中心和云计算。这一进展标志着 AI 控制技术的发展,推动了智能化工作的效率提升。

7 rules for self-improving agent loops every AI engineer should know

@GoogleCloudTech · 1.3M 粉丝 · 46.8K 阅 · 505 赞 · 63 转

By @Saboo_Shubham_ and @secchi_elia When coding agents write the code, defining what's good is the real engineering job. Coding agents can now build and improve other agents. Write the instructions,

中文介绍 博主分析了自我改进的代理循环的重要性,指出在编码代理时,定义“好”的标准是核心工程任务。这一概念在 AI 工程师中引发关注,展现了自动化生成和改进自身代码的能力正在成为重要的发展方向。

Intro to Grok Bot

@mattyp · 44.5K 粉丝 · 33.5K 阅 · 651 赞 · 48 转

I’m always chasing better tools. Notes apps, workflows, optimizations, and now, personal agents. It’s an easy trap to fall into - always building the “custom” thing and sacrificing the work as a

中文介绍 博主探讨了个人代理工具的优化,强调在追求更高效的工作流时需避免过于追求“定制化”而影响整体工作效率的陷阱。这反映出在个人乃至团队管理中,平衡工具使用和工作效率的重要性不断被讨论。

🔬The BioAI Phase Shift - Matthew McPartlon & Neil Patil, Chai Discovery

Pharma is suddenly paying for Bio × AI tools, and Chai is leading the pack with four deals closed this summer. Cofounder Matt McPartlon and Product leader Neil Patil explain why.

中文介绍 制药公司对于生物与人工智能工具的需求正在急剧增加,Chai Discovery在这个夏天达成了四项交易。共同创始人马修·麦克帕特隆和产品负责人尼尔·帕特尔解释了原因。

Testing ads in ChatGPT

OpenAI begins testing ads in ChatGPT to support free access, with clear labeling, answer independence, strong privacy protections, and user control.

中文介绍 OpenAI开始在ChatGPT中测试广告,以支持免费使用,广告将明确标识,确保答案的独立性,并保护用户隐私以及提供用户控制权。

Daybreak models are now available on AWS

OpenAI and AWS are making Daybreak cybersecurity capabilities available through Amazon Bedrock to support enterprise security workflows.

中文介绍 OpenAI与AWS合作,将其Daybreak网络安全能力通过Amazon Bedrock推向市场,旨在支持企业安全工作流程。

AI professors are negotiating the new realities of academic research

This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. Last week, I headed 30 miles south of San Francisco to a hotel in Mountain View, California, to join some of the most accomplished, and some of the most promising

中文介绍 多位学者在加州圣克拉拉讨论AI对学术研究新现实的影响,强调学术界面临的变革与挑战。

What building an AI-native finance function taught me

OpenAI CFO Sarah Friar shares five lessons for building an AI-native finance function, from automated forecasting to stronger controls and AI ROI.

中文介绍 OpenAI首席财务官莎拉·弗赖尔分享了五个建立AI原生财务功能的经验教训,涵盖自动预测、加强控制和提升AI投资回报等方面。

OpenAI’s letter to Governor Abbott on responsible AI infrastructure in Texas

OpenAI sent Governor Greg Abbott a letter outlining its commitment to responsible AI infrastructure in Texas. The letter supports reliable, transparent growth that benefits Texans.

中文介绍 OpenAI向德克萨斯州州长格雷格·阿博特发信,强调其对德州负责的AI基础设施的承诺,支持可靠、透明的发展。

Model ML completes finance work more efficiently with GPT-5.6 Sol

Model ML uses GPT-5.6 Sol to carry finance work from research and analysis through editable, traceable PowerPoint decks and Excel workbooks.

中文介绍 Model ML利用GPT-5.6 Sol高效完成金融工作,从研究和分析到可编辑、可追溯的PowerPoint和Excel文档。

Expanding Daybreak as the Cyber Defense Window Narrows

Meet GPT-5.6-Cyber, OpenAI’s cybersecurity-specific model available through Daybreak Red for authorized vulnerability research, exploit validation, and security testing.

中文介绍 OpenAI推出GPT-5.6-Cyber模型,通过Daybreak Red为授权的漏洞研究、利用验证和安全测试提供支持。

Putting frontier cyber models in more trusted hands

Approved Daybreak partners can use OpenAI’s frontier cyber models to deliver authorized, governed cybersecurity services to customers.

中文介绍 获得批准的Daybreak合作伙伴可使用OpenAI前沿网络安全模型,为客户提供授权和管理的网络安全服务。

AI for science needs reasoning, not just data

Every few decades, someone announces that science has reached its end. In 1903, the revered physicist Albert Michelson wrote that the “facts of physical science have all been discovered.” In the 1980s, Stephen Hawking predicted that theoretical physics might be finished by the end of the century. Wi

中文介绍 科学界不断呼吁需要AI进行推理而不仅仅是数据处理,强调科学探索依然在持续进行中。

市场总览

当前市场整体呈现出混合的技术状态。在美股方面,S&P 500 ETF(SPY)和Nasdaq 100 ETF(QQQ)均显示出了上行动能,但均存在一定的回调风险。加密市场的恐慌贪婪指数为29,表明市场处于恐慌状态,且比特币(BTC-USD)面临MACD死叉,暗示下行风险。商品外汇中,WTI原油期货(CL=F)表现良好,短期内呈现出多头排列,支持上行趋势。虽然存在一些强劲的标的,但市场的整体情绪仍较为谨慎。

今日关注

SPY S&P 500 ETF
偏上行

当前价格为770.56,RSI14为63.3,显示正常状态;MACD为7.3125,且多头排列,表明短期内有上行的趋势。

SOL-USD Solana
偏上行

当前价格为76.22,RSI14为55,显示正常状态;同时MACD金叉,提示可能的上行动能。

MSFT Microsoft
中性

当前价格为503.81,RSI14为77.7,超买状态;虽然MACD为30.2958,但由于RSI超买,需要警惕潜在的回调风险,因此状态为中性。

全部资产

^VIX

VIX 恐慌指数

$15.28 -1.16%
5 日
-7.39%
距 52w 高
-56.7%
RSI(14)
43.7
趋势
空头
SMA 20 / 50 / 200
17.02 / 17.27 / 18.56
MACD / 信号
-0.550 / -0.274
空头排列

^TNX

10Y 美债收益率 (%)

$4.68 -0.32%
5 日
+1.23%
距 52w 高
-1.3%
RSI(14)
57.1
趋势
多头
SMA 20 / 50 / 200
4.64 / 4.55 / 4.30
MACD / 信号
0.038 / 0.041
MACD 死叉 (4 天前)接近 52 周高多头排列

DX-Y.NYB

美元指数 DXY

$99.81 -0.00%
5 日
-0.08%
距 52w 高
-2.0%
RSI(14)
40.1
趋势
中性
SMA 20 / 50 / 200
100.52 / 100.52 / 99.18
MACD / 信号
-0.266 / -0.132
接近 52 周高

SPY

S&P 500 ETF

$770.56 -0.32%
5 日
-0.10%
距 52w 高
-0.8%
RSI(14)
63.3
趋势
多头
SMA 20 / 50 / 200
752.30 / 747.84 / 703.99
MACD / 信号
7.312 / 4.402
接近 52 周高多头排列

QQQ

Nasdaq 100 ETF

$718.45 -0.34%
5 日
-0.75%
距 52w 高
-4.0%
RSI(14)
55.1
趋势
多头
SMA 20 / 50 / 200
700.73 / 713.88 / 648.96
MACD / 信号
2.134 / -2.127
多头排列

AAPL

Apple

$304.91 -1.09%
5 日
-1.44%
距 52w 高
-11.5%
RSI(14)
41.9
趋势
中性
SMA 20 / 50 / 200
322.48 / 309.56 / 279.87
MACD / 信号
-0.827 / 2.444

MSFT

Microsoft

$503.81 -0.44%
5 日
+2.23%
距 52w 高
-9.0%
RSI(14)
77.7
趋势
中性
SMA 20 / 50 / 200
435.53 / 409.66 / 432.94
MACD / 信号
30.296 / 21.168
RSI 超买

NVDA

Nvidia

$217.50 -0.02%
5 日
+2.62%
距 52w 高
-8.0%
RSI(14)
58.0
趋势
多头
SMA 20 / 50 / 200
207.80 / 206.27 / 194.35
MACD / 信号
3.651 / 1.629
多头排列

GOOGL

Alphabet

$343.80 -3.84%
5 日
-8.96%
距 52w 高
-15.9%
RSI(14)
45.9
趋势
中性
SMA 20 / 50 / 200
348.23 / 355.12 / 330.11
MACD / 信号
0.261 / -1.070

TSLA

Tesla

$332.81 +0.58%
5 日
+1.67%
距 52w 高
-33.3%
RSI(14)
43.5
趋势
空头
SMA 20 / 50 / 200
336.97 / 376.16 / 407.71
MACD / 信号
-16.385 / -19.469
MACD 金叉 (2 天前)空头排列

META

Meta

$599.12 +0.71%
5 日
+1.90%
距 52w 高
-24.8%
RSI(14)
49.9
趋势
中性
SMA 20 / 50 / 200
606.08 / 598.28 / 629.93
MACD / 信号
-5.702 / -5.461
加密恐慌贪婪
29
恐慌
加密总市值
$2.27 T
-0.08% / 24h
BTC 主导率
56.2%
ETH 10.0%
24h 成交量
$51.0 B
活跃币 18,377

BTC-USD

Bitcoin

$63,579.30 -0.52%
5 日
-1.06%
距 52w 高
-49.6%
RSI(14)
46.4
趋势
中性
SMA 20 / 50 / 200
64,129.02 / 63,323.42 / 69,872.57
MACD / 信号
47.527 / 73.442
MACD 死叉 (今天)

ETH-USD

Ethereum

$1,882.27 +0.59%
5 日
-1.04%
距 52w 高
-62.0%
RSI(14)
51.3
趋势
中性
SMA 20 / 50 / 200
1,890.65 / 1,808.25 / 2,039.31
MACD / 信号
15.243 / 20.658

SOL-USD

Solana

$76.22 +0.36%
5 日
+5.02%
距 52w 高
-69.9%
RSI(14)
55.0
趋势
中性
SMA 20 / 50 / 200
74.33 / 75.46 / 83.01
MACD / 信号
-0.034 / -0.402
MACD 金叉 (3 天前)

BABA

阿里巴巴 (BABA)

$127.85 -3.38%
5 日
-0.88%
距 52w 高
-33.6%
RSI(14)
62.5
趋势
中性
SMA 20 / 50 / 200
120.76 / 114.01 / 139.73
MACD / 信号
4.689 / 3.515

PDD

拼多多 (PDD)

$90.50 -2.69%
5 日
-0.58%
距 52w 高
-35.1%
RSI(14)
59.2
趋势
中性
SMA 20 / 50 / 200
87.51 / 84.08 / 102.50
MACD / 信号
2.072 / 1.570

JD

京东 (JD)

$31.92 -4.63%
5 日
-3.18%
距 52w 高
-13.4%
RSI(14)
57.8
趋势
中性
SMA 20 / 50 / 200
31.50 / 29.18 / 29.43
MACD / 信号
1.142 / 1.129

0700.HK

腾讯控股 (0700.HK)

HK$470.80 -2.20%
5 日
-3.45%
距 52w 高
-31.1%
RSI(14)
51.8
趋势
中性
SMA 20 / 50 / 200
468.79 / 456.15 / 533.89
MACD / 信号
7.266 / 6.558

GC=F

黄金期货

$4,427.20 +1.50%
5 日
+8.10%
距 52w 高
-20.7%
RSI(14)
69.0
趋势
中性
SMA 20 / 50 / 200
4,121.31 / 4,163.86 / 4,481.96
MACD / 信号
49.999 / 1.902

CL=F

WTI 原油期货

$83.47 +1.63%
5 日
+10.16%
距 52w 高
-30.1%
RSI(14)
54.3
趋势
多头
SMA 20 / 50 / 200
82.22 / 80.26 / 76.53
MACD / 信号
0.044 / 0.142
多头排列

USDCNY=X

美元 / 人民币

¥6.74 -0.04%
5 日
-0.13%
距 52w 高
-6.2%
RSI(14)
30.0
趋势
空头
SMA 20 / 50 / 200
6.76 / 6.77 / 6.89
MACD / 信号
-0.009 / -0.008
接近 52 周低空头排列
风险提示

过去走势不代表未来表现,本报告仅供技术指标解读参考,投资需谨慎。

Colombia scrambles to find survivors after deadly earthquake

At least 254 deaths have been reported so far, as rescue workers continue to sift through collapsed buildings.

中文摘要 哥伦比亚发生致命地震,目前已报告254人遇难,救援人员仍在瓦砾中寻找幸存者。

Australia politics live: Tim Wilson questions energy efficiency standards for housing, saying ‘some people call it putting a jumper on’

Coalition promises to drastically cut net overseas migration and housing construction code. Follow today’s news live Get our breaking news email, free app or daily news podcast Education minister says funding agreement with Victoria ‘very close’ The education minister, Jason Clare, says a funding de

中文摘要 澳大利亚反对党承诺大幅减少海外移民和住房建设法规。教育部长杰森·克拉尔表示与维多利亚州的资金协议“非常接近”。

Russia releases former US Marine reported in poor health

Robert Gilman, who was arrested in 2022, was left in "life-threatening condition" after being subjected to "harsh and abusive treatment", his family says.

中文摘要 前美国海军陆战队员罗伯特·吉尔曼在2022年被捕,因遭受“严酷和虐待性待遇”而健康堪忧,家属称其目前处于“危及生命状态”。

How a 97-year-old escaped deadly Indian floods on a banana raft

A 97-year-old's dramatic rescue captures how Assam's floods are reaching places once thought safe.

中文摘要 97岁老人用香蕉筏成功逃离印度阿萨姆邦的洪水,其被救援的戏剧性表现出该地区洪水的蔓延情况。

Male redback spiders invite cannibalism through a ‘self-sacrificial somersault’ during sex – but why?

New research sheds light on the genetic basis of the venomous insect’s ‘weird mating behaviour’ Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast If you think you’ve had a bad one-night stand, spare a thought for the male redback spide

中文摘要 新的研究揭示了红背蜘蛛雄性在交配时通过“自我牺牲翻筋斗”引发同类相食行为的遗传基础。

Father and sons forced to manually clear rubble in Gaza

Forced to rely on basic tools, a father and his sons are clearing the rubble of their destroyed home in Gaza.

中文摘要 加沙一名父亲与他的儿子们被迫使用简单工具手动清理被摧毁的家园废墟。

Searching for Survivors After Colombia’s Earthquake

Our Colombia-based reporter Genevieve Glatsky witnesses the search for survivors after a five-story apartment building in the city of Cali collapsed in an earthquake.

中文摘要 在哥伦比亚,记者见证了因地震而倒塌的五层公寓楼的幸存者搜救工作。

Deported From the U.S., Trapped in a Prison in Africa

Roberto Mosquera, a Cuban immigrant who has lived in the United States since 1980 and completed a prison sentence in 1996, was arrested by ICE last year and sent to a prison in Eswatini, a country in Africa he’d never heard of. After a hunger strike, Mosquera obtained a phone and talked with Nichola

中文摘要 古巴移民罗伯托·莫斯克拉因被驱逐至非洲淳朴地区监狱而感到困惑,经过绝食抗议后获得了打电话的机会。

Relatives mourn victims of Pakistan avalanche

Relatives gathered to mourn the three Nepali mountaineers, Kili Pemba Sherpa, Gyalu Sherpa, and Nima Sherpa.

中文摘要 巴基斯坦山区雪崩的遇难者亲属聚集哀悼三名尼泊尔登山者吉利·彭巴·舍尔巴、贾尤·舍尔巴和宁玛·舍尔巴。

Pakistan interior minister in Tehran amid talks to reopen Hormuz

Pakistan’s interior minister Mohsin Naqvi has held talks with Iranian Foreign Minister Abbas Araghchi in Tehran.

中文摘要 巴基斯坦内政部长莫欣·纳克维在德黑兰与伊朗外长阿巴斯·阿拉赫奇举行会谈,讨论重启霍尔木兹海峡。

Total Eclipse

If you’re in Europe, get ready to gaze at the sky. But not without the right glasses.

中文摘要 如果您在欧洲,准备好欣赏天象,但不要忘记准备合适的眼镜。

China Earnings to Test Rotation From AI to Internet Stocks

A rotation into China’s biggest tech firms from AI infrastructure stocks will be tested this week as earnings are due.

中文摘要 本周,中国大型科技公司的盈利将测试从人工智能基础设施股票到互联网股票的投资轮换。

Tokenomics: Why making AI pay is tricky

Buyers of AI services are struggling to control costs and sellers are not sure how much to charge.

中文摘要 购买人工智能服务的客户面临控制成本的困难,而销售方则不确定定价策略.

Global Investors Favor Taiwan Over Korea Stocks After AI Selloff

Taiwanese stocks are gaining favor over South Korea’s on steadier earnings prospects, as the two AI-bellwether markets look to recover from July’s rout.

中文摘要 全球投资者在人工智能股票回调后普遍偏好台湾股票,相比之下韩国股票的盈利前景不稳定。

Singapore Exchange Eyes Single-Stock ETFs, Tie-Ups for Growth

Singapore Exchange Ltd. is exploring offering a broader suite of exchange-traded funds and stepping up partnerships across markets as it seeks to bolster its position as Southeast Asia’s key multi-asset exchange.

中文摘要 新加坡交易所正探索扩展交易所交易基金,并加强与其他市场的合作,以巩固其东南亚多资产交易所的地位。

Oil Gains With Iran in Focus, CoreWeave Rallies: Markets Wrap

Crude oil posted a modest gain as traders remained wary over the prospects for a Middle East deal, tempering the mood ahead of crucial US inflation data. Sentiment toward technology stocks remained buoyant, with CoreWeave Inc. shares surging after the company bolstered its outlook.

中文摘要 原油小幅上涨,因交易者对中东交易的前景仍持谨慎态度,在重要的美国通胀数据发布前,科技股情绪保持乐观。

Latest Oil Market News and Analysis for Aug. 12

Oil held a four-day gain as traders monitored efforts to reach a deal that could help restore energy flows through the Strait of Hormuz.

中文摘要 油价维持四天的涨幅,交易者关注恢复霍尔木兹海峡能源流动的谈判进展。

Blackstone Strikes $1.8 Billion Deal for Air Canada Loyalty Unit

Blackstone Inc. and a group of Canadian pension funds are buying a 25% stake in Air Canada’s Aeroplan loyalty program for C$2.5 billion ($1.8 billion).

中文摘要 黑石集团与一组加拿大养老金基金以25亿加元(18亿美元)收购了加拿大航空的Aeroplan忠诚度计划25%的股权。

The Chinese AI Model Rattling Silicon Valley and Washington

A new Chinese AI model is sending shockwaves through Silicon Valley and Washington. Today on Big Take Asia: What Moonshot’s controversial Kimi K3 means for the future of the global AI race. (Source: Bloomberg)

中文摘要 中国新推出的人工智能模型正对硅谷和华盛顿产生震荡,其影响将对全球人工智能竞争未来产生深远影响。

该源今日无内容。