每日简报

2026-06-20

← 历史归档

DeusData/codebase-memory-mcp

C · ★ 8,212 · 🍴 624 · 📈 1,058 stars today

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

中文介绍 高性能代码智能MCP服务器,将代码库索引为持久知识图谱,平均毫秒级处理仓库,支持158种语言,亚毫秒查询,减少99% token消耗。单静态二进制部署,适合快速理解大型代码库的开发者。

google-research/timesfm

Python · ★ 24,079 · 🍴 2,271 · 📈 1,510 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,专注于时间序列预测。可用于金融、气象、物联网等领域的预测任务,无需大量微调即可获得良好效果。

palmier-io/palmier-pro

Swift · ★ 1,892 · 🍴 187 · 📈 756 stars today

macOS video editor built for AI

中文介绍 专为macOS设计的AI视频编辑器,利用人工智能简化视频剪辑流程,提供智能剪辑、特效生成等功能,适合内容创作者和视频编辑者提升效率。

koala73/worldmonitor

TypeScript · ★ 57,230 · 🍴 9,136 · 📈 156 stars today

Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface

中文介绍 实时全球情报仪表盘,整合AI新闻聚合、地缘政治监控和基础设施跟踪,提供统一态势感知界面。适合分析师、记者及关注全球动态的机构实时掌握关键信息。

aishwaryanr/awesome-generative-ai-guide

HTML · ★ 27,613 · 🍴 5,738 · 📈 107 stars today

A one stop repository for generative AI research updates, interview resources, notebooks and much more!

中文介绍 一站式生成式AI资源仓库,汇集研究更新、面试准备资料、笔记等。适合AI从业者、求职者和学习者快速获取最新生成式AI相关知识。

BuilderIO/agent-native

TypeScript · ★ 1,037 · 🍴 117 · 📈 147 stars today

A framework for building agent-native applications.

中文介绍 构建代理原生应用的框架,提供工具和抽象层,帮助开发者轻松创建AI代理驱动的应用,支持自定义行为和工作流。适合需要集成AI代理的开发者。

chopratejas/headroom

Python · ★ 38,661 · 🍴 2,641 · 📈 4,005 stars today

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 60-95% fewer tokens, same answers. Library, proxy, MCP server.

中文介绍 压缩工具输出、日志、文件和RAG块的库/代理/MCP服务器,平均减少60-95% token消耗且不牺牲答案质量。适合LLM应用开发者降低API成本、提升推理速度。

calesthio/OpenMontage

Python · ★ 6,271 · 🍴 1,080 · 📈 156 stars today

World's first open-source, agentic video production system. 12 pipelines, 52 tools, 500+ agent skills. Turn your AI coding assistant into a full video production studio.

中文介绍 全球首个开源代理视频制作系统,集成12条流水线、52个工具和500+代理技能。可将AI编码助手转化为功能齐全的视频工作室,适合内容创作者和视频制作团队实现自动化制作。

zai-org/GLM-5

★ 4,577 · 🍴 461 · 📈 480 stars today

GLM-5: From Vibe Coding to Agentic Engineering

中文介绍 GLM-5致力于从“氛围编码”向“代理工程”转变,将AI代理整合进软件开发生命周期,适合探索AI驱动开发方法并提升工程效率的开发者。

withastro/flue

TypeScript · ★ 5,828 · 🍴 323 · 📈 309 stars today

The sandbox agent framework.

中文介绍 沙盒代理框架,提供安全隔离的执行环境来运行AI代理。适合开发需要沙箱化、可控代理行为的应用场景,如自动化测试、安全评估等。

n0-computer/iroh

Rust · ★ 10,246 · 🍴 466 · 📈 302 stars today

IP addresses break, dial keys instead. Modular networking stack in Rust.

中文介绍 用Rust编写的模块化网络栈,使用密钥替代IP地址进行可靠的点对点通信,解决传统IP易失效问题。适合去中心化、安全联网的应用程序。

obra/superpowers

Shell · ★ 233,334 · 🍴 20,725 · 📈 1,110 stars today

An agentic skills framework & software development methodology that works.

中文介绍 一种实用的代理技能框架与软件开发方法论,通过定义“技能”组织和执行AI代理任务,提升开发效率和代码质量。适合团队探索AI辅助编程和自动化工作流。

penpot/penpot

Clojure · ★ 50,582 · 🍴 3,268 · 📈 85 stars today

Penpot: The open-source design tool for design and code collaboration

中文介绍 开源设计工具,支持矢量设计、原型制作,促进设计师与开发者高效协作。作为Figma的替代品,提供本地化部署和数据控制,适合开源社区和企业使用。

Kong/insomnia

TypeScript · ★ 38,971 · 🍴 2,300 · 📈 292 stars today

The open-source, cross-platform API client for GraphQL, REST, WebSockets, SSE and gRPC. With Cloud, Local and Git storage.

中文介绍 开源跨平台API客户端,支持GraphQL、REST、WebSockets、SSE和gRPC协议,提供云、本地和Git多种存储方式。适合开发者调试和测试API接口。

Lightricks/LTX-2

Python · ★ 7,668 · 🍴 1,220 · 📈 196 stars today

Official Python inference and LoRA trainer package for the LTX-2 audio–video generative model.

中文介绍 官方Python推理和LoRA训练器,用于LTX-2音频-视频生成模型,支持从文本/音频生成视频,并可通过LoRA微调定制风格。适合AI艺术创作者和研究者探索音视频生成。

Moebius: 0.2B Lightweight Image Inpainting Framework with 10B-Level Performance

👍 97

While 10B-level industrial foundation models have pushed the boundaries of image inpainting, their prohibitive computational costs severely hinder practical deployment. Constructing a highly optimized task-specific specialist offers a promising solution; however, extreme structural compression inevi

中文介绍 提出Moebius,0.2B参数的轻量级图像修复框架,通过极端结构压缩达到10B级性能,大幅降低计算成本,实现高效部署。

DragMesh-2: Physically Plausible Dexterous Hand-Object Interaction with Articulated Objects

👍 61

Dexterous interaction with articulated objects is important for household, assistive, and humanoid manipulation, where multi-finger hands can provide compliant contact patterns beyond parallel-jaw grasping. However, articulated-object manipulation differs from static-object manipulation: the target

中文介绍 提出DragMesh-2,实现物理合理的灵巧手与铰接物体交互,通过多指手提供合规接触模式,提升操控真实感和泛化能力。

MolmoMotion: Forecasting Point Trajectories in 3D with Language Instruction

👍 41

Motion forecasting is central to visual intelligence: agents must anticipate how objects will move in order to plan actions, reason about physical interactions, and synthesize realistic futures. We argue that 3D points in world coordinates provide a general representation that is class-agnostic, vie

中文介绍 提出MolmoMotion,基于语言指令预测3D世界坐标点的轨迹,以类无关通用表示实现运动预测,支持物理交互推理。

Playful Agentic Robot Learning

👍 35

Current agentic robot systems can write executable Code-as-Policy programs, observe feedback, and revise behavior across multiple attempts, but they remain largely task-driven: reusable skills are acquired only after explicit instructions. We study Playful Agentic Robot Learning, where an embodied c

中文介绍 研究Playful Agentic Robot Learning,让机器人通过「玩耍」自主学习可复用技能,无需明确指令,减少人工监督并提升效率。

S-Agent: Spatial Tool-Use Elicits Reasoning for Spatial Intelligence

👍 27

Real-world spatial intelligence requires reasoning over a continuous and evolving 3D world, yet existing VLMs and tool-augmented agents largely remain tied to static, stateless inference from isolated visual observations. We introduce \textsc{S-Agent}, a spatial tool-use agentic paradigm for underst

中文介绍 提出S-Agent,通过空间工具使用范式实现连续3D世界推理,突破VLMs和工具智能体在静态场景中的局限。

Multi-LCB: Extending LiveCodeBench to Multiple Programming Languages

👍 26

LiveCodeBench (LCB) has recently become a widely adopted benchmark for evaluating large language models (LLMs) on code-generation tasks. By curating competitive programming problems, constantly adding fresh problems to the set, and filtering them by release dates, LCB provides contamination-aware ev

中文介绍 扩展LiveCodeBench至多编程语言(Multi-LCB),持续更新竞编程题目并过滤污染,提供更全面的跨语言代码生成评估基准。

Beyond Static Leaderboards: Predictive Validity for the Evaluation of LLM Agents

👍 21

Agent benchmarks are growing fast, but no single benchmark touches more than four or five of the dimensions that deployment exposes. This paper aggregates the largest coordinated deep-dive of one MCP-based industrial-agent benchmark to date: fourteen parallel implementation studies covering new asse

中文介绍 综合14项并行实现研究,提出超越静态排行榜的评估框架,强调LLM智能体评测的预测有效性,覆盖部署关键维度。

FreeStyle: Free Control of Style-Content Dual-Reference Generation from Community LoRA Mining

👍 18

Style-content dual-reference generation aims to synthesize an image that preserves the structure and semantics of a content reference while adopting the style of a separate style reference.Despite recent progress, this setting remains challenging because models must balance content fidelity, style a

中文介绍 提出FreeStyle,从社区LoRA挖掘实现风格-内容双参考生成,平衡内容保真度与风格迁移,提升生成质量。

DF3DV-1K: A Large-Scale Dataset and Benchmark for Distractor-Free Novel View Synthesis

👍 9

Advances in radiance fields have enabled photorealistic novel view synthesis. In several domains, large-scale real-world datasets have been developed to support comprehensive benchmarking and to facilitate progress beyond scene-specific reconstruction. However, for distractor-free radiance fields, a

中文介绍 构建DF3DV-1K大规模无干扰新视角合成数据集,含1000个场景,为无干扰辐射场提供全面基准,推动场景无关重建进展。

MaineCoon: Pursuing A Real-Time Audio-Visual Social World Model

👍 8

As an increasing majority of global video content is consumed on social platforms for interactive social purposes, video generation models built for social worlds are important but largely overlooked by previous studies. In this work, we define the position of social world models and build a prototy

中文介绍 定义社交世界模型及原型MaineCoon,针对社交平台交互视频生成,实现实时音视频一致的社交内容合成。

ENPIRE: Agentic Robot Policy Self-Improvement in the Real World

👍 7

Achieving dexterous robotic manipulation in the real world heavily relies on human supervision and algorithm engineering, which becomes a central bottleneck in the pursuit of general physical intelligence. Although emerging coding agents can generate code to automate algorithm search, their successe

中文介绍 提出ENPIRE,实现真实世界中机器人策略自改进,通过编码智能体自动搜索算法,减少灵巧操作人工监督瓶颈。

ViT-Up: Faithful Feature Upsampling for Vision Transformers

👍 6

Vision Transformers (ViTs) have become a dominant architecture for visual representation learning, providing exceptionally strong and broadly reusable backbone features. However, ViTs are commonly operated on relatively small patch-token grids due to the quadratic cost of global self-attention, whic

中文介绍 提出ViT-Up,对Vision Transformer特征进行忠实上采样,克服全局自注意力低分辨率限制,保留细节并提升可重用性。

Context-Aware RL for Agentic and Multimodal LLMs

👍 4

Large language models (LLMs) often fail when answering requires identifying a small but decisive piece of evidence within a long or complex context, such as a single line in a tool trace or a subtle detail in an image. We propose ContextRL, a context-aware reinforcement learning (RL) method that imp

中文介绍 提出ContextRL,上下文感知强化学习方法,提升LLM在长复杂语境中识别关键证据的能力,改进智能体与多模态问答。

Understanding the Behaviors of Environment-aware Information Retrieval

👍 4

Recent retrieval-augmented generation (RAG) approaches have demonstrated strong capability in handling complex queries, yet current research overlooks a critical challenge: different retrievers require fundamentally different query formulation strategies for optimal performance. In this work, we pre

中文介绍 研究环境感知信息检索中不同检索器对查询策略的需求差异,揭示适配策略对RAG性能的关键影响。

FAPO: Fully Autonomous Prompt Optimization of Multi-Step LLM Pipelines

👍 4

Multi-step LLM pipelines fail through interactions among retrieval, reasoning, and formatting steps, so prompt-only optimization can miss bottlenecks in the chain. We present FAPO (Fully Autonomous Prompt Optimization), a framework that lets Claude Code optimize an LLM pipeline inside a standardized

中文介绍 提出FAPO框架,让Claude Code全自动优化多步LLM流水线中的提示,解决检索、推理、格式化步骤间交互失败问题。

Thinking with Visual Grounding

👍 4

Visual thinking should not only sound right; it should show its evidence. While recent vision-language models (VLMs) can produce natural-language reasoning traces, these traces often leave the supporting image regions implicit, making them hard to verify and difficult to supervise. We introduce visu

中文介绍 引入视觉接地推理,让VLM产生推理轨迹时显式定位支持图像区域,提升可验证性和可监督性。

LedgerAgent: Structured State for Policy-Adherent Tool-Calling Agents

👍 3

Policy-adherent tool-calling agents in customer-service domains must maintain task states across turns while calling tools and obeying domain policies. Task states consist of relevant facts, identifiers, constraints, and conditions observed through user interaction and tool calls. In standard agents

中文介绍 提出LedgerAgent,用结构化状态维护客服工具调用智能体中的任务事实与约束,提高策略遵循准确性。

Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States

👍 3

Progress in legal AI increasingly depends on access to authoritative legal text at scale. Yet one of the most consequential layers of American law remains largely absent from existing machine-readable corpora: local ordinances. Local codes govern zoning, housing, business licensing, public health, n

中文介绍 构建LOCUS,首个大规模机器可读美国地方法规语料库,涵盖分区、住房、商业许可等,填补法律AI数据空白。

HumanScale: Egocentric Human Video Can Outperform Real-Robot Data for Embodied Pretraining

👍 3

Embodied foundation models are expected to benefit from data scaling like large language models, but face a much tighter data bottleneck. Teleoperated real-robot trajectories remain the dominant pretraining source due to their precise action supervision and embodiment alignment, yet their scalabilit

中文介绍 证明第一人称人类视频在具身预训练中可超越真实机器人数据,提升数据可扩展性,降低遥操作依赖。

When Does Trajectory-Level Supervision Permit Efficient Offline Reinforcement Learning?

👍 3

Offline reinforcement learning is typically analyzed under process-level reward supervision, yet many sequential decision datasets record only trajectory-level outcomes. We develop a statistical theory for offline policy optimization from such outcome-level supervision. We first study the canoni

中文介绍 开发离线强化学习统计理论,在仅有轨迹级结果监督下分析可识别性和样本效率,指导高效策略优化。

How Quants Use Loop Engineering to Build Alpha (Full Framework)

@horizon_trade_x · 4.4K 粉丝 · 1.3M 阅 · 507 赞 · 59 转

Your backtest looked flawless. You went live. Two weeks later, the strategy was bleeding. Every quant has lived this. The answer is a loop: generate a strategy, test it, score it, feed the result

中文介绍 量化交易中回测完美实盘却亏损,博主提出「循环工程」框架:生成策略→测试→评分→反馈结果,形成闭环以持续优化Alpha,适合量化从业者参考。

Context Engineering for AI Agents: The Complete Playbook

@sairahul1 · 117.4K 粉丝 · 511.9K 阅 · 500 赞 · 84 转

Your AI agent works great for the first 10 steps. Then somewhere around step 15, it starts getting sloppy. Wrong tool calls. Forgetting your original instructions. Low-quality outputs. Most people

中文介绍 AI Agent运行15步后常出现工具调用错误、遗忘指令等问题。博主总结「上下文工程」完整方法论,教你在长任务中维持Agent的准确性和一致性。

The Agent Loop Architecture

@djfarrelly · 3.8K 粉丝 · 344.7K 阅 · 501 赞 · 61 转

Everyone's asking "WTF is a loop?" Here's the question nobody's asking: what runs the loop? The AI discourse has converged on loops as a core primitive of agentic systems. Matt Van Horn (@mvanhorn)

中文介绍 AI领域共识「循环」是Agent系统核心原语,但关键在于「谁驱动循环」。博主拆解Agent Loop架构,探讨循环执行机制与设计思路。

The Self-Improving Loop: a 300-agent swarm on Kimi K2.6, verified by Opus 4.8

@0xMovez · 26.7K 粉丝 · 208.0K 阅 · 504 赞 · 59 转

A free open-source model is running 300 parallel agents across 4,000 coordinated steps from a single prompt, and it scores higher on real research tasks than models you pay 5x more for. Most people

中文介绍 开源模型Kimi K2.6实现300个Agent并行、4000步协同的「自改进循环」,单提示即可驱动,研究任务评分超过5倍价格的闭源模型,由Opus 4.8验证。

how to get Fable-level intelligence back:

@EXM7777 · 118.9K 粉丝 · 107.7K 阅 · 509 赞 · 44 转

for a few days, we had something that felt like AGI... Fable 5 showed up, effectively unlimited inside the plans, and the ceiling on what you could build lifted overnight but then Anthropic killed it,

中文介绍 Fable 5曾短暂展现接近AGI的能力,计划内几乎无限,但被Anthropic削弱。博主探讨如何恢复Fable级别的智能,引发对模型能力上限与监管的讨论。

ORACLE: Official AI Agents Trade on Polymarket

@OracleMindAI · 21.0K 粉丝 · 105.0K 阅 · 2.8K 赞 · 582 转

In 2026, autonomous AI agents have become one of the most effective strategies on prediction markets. Over 30% of all activity on Polymarket now comes from algorithmic and AI-powered wallets. We

中文介绍 2026年自主AI Agent已成为预测市场最有效策略之一,Polymarket超30%活动来自算法/AI钱包。ORACLE官方Agent在Polymarket进行交易,展示AI金融应用。

How GLM-5.2 Beat Fable 5 at Website Design

@Designarena · 13.9K 粉丝 · 80.4K 阅 · 518 赞 · 39 转

GLM 5.2 ranks 1st overall on Design Arena’s single-turn, HTML Web Design (Non-Agentic) evaluation, 5 places higher than its predecessor GLM-5.1. To do so, it beat Claude Fable 5, Opus 4.6, and Opus

中文介绍 GLM-5.2在Design Arena单轮HTML网页设计评测中排名第一,超越Claude Fable 5、Opus 4.6等强手,比前代GLM-5.1跃升5位,显示非Agentic场景下的设计能力。

How to Build a GTM Team on Claude Code You Can Run Alone

@nifinet · 10.4K 粉丝 · 62.8K 阅 · 519 赞 · 38 转

A GTM team looks like a sending operation. Most of its real work is judgment: which company is worth a message this week, what to say that proves you noticed, which no-show to chase, what actually

中文介绍 GTM团队本质是判断工作:目标公司筛选、个性化消息撰写、跟进无回应线索。博主教你用Claude Code构建单人GTM团队,自动化发送与决策流程。

The Stanford STORM Method: How to Make Claude Research Like a PhD in Minutes

@heynavtoor · 143.5K 粉丝 · 47.7K 阅 · 538 赞 · 70 转

Most people use Claude like a search box. Ask, answer, close tab. They are leaving the best feature locked. Save this :) Stanford built a research system called STORM. In peer reviewed testing it

中文介绍 斯坦福开发的STORM研究系统,能让Claude像博士一样在几分钟内完成深度研究。博主详解该方法,教你摆脱一问一答模式,解锁Claude的真正能力。

Introducing Design Crit: We taught AI to judge design like a designer.

@contralabs_ai · 2.9K 粉丝 · 46.6K 阅 · 501 赞 · 37 转

Everyone keeps talking about taste. But you can't improve what you can't measure. So we measured it. Design Crit is a dataset of ten professional designers ranking four frontier image models across

中文介绍 Design Crit数据集由10位专业设计师对4个前沿图像模型进行评分,用可量化的标准衡量「品味」,帮助AI像设计师一样评价设计质量,提升审美判断可信度。

Arcads MCP + Claude Just Changed AI Ad Creation Forever

@Just_sharon7 · 44.1K 粉丝 · 32.5K 阅 · 502 赞 · 18 转

If you're still switching between 10 tools to make UGC ads, analyze performance, and iterate... your workflow is outdated. Arcads just dropped MCP support, and pairing it with Claude (especially Fable

中文介绍 Arcads推出MCP支持,与Claude(特别是Fable)结合,一站式完成UGC广告制作、性能分析和迭代,告别多工具切换,彻底改变AI广告创建流程。

World Models and Interpretability Are Two Sides of the Same Coin

@soniajoseph_ · 16.9K 粉丝 · 13.9K 阅 · 861 赞 · 21 转

There has been a lot of confusion around world models. Depending on who you ask, a world model is a generative model, a 3D reconstruction model, or a latent-space prediction model. I would like to

中文介绍 世界模型定义混乱:生成模型、3D重建或潜空间预测?博主指出世界模型与可解释性实为硬币两面,意图澄清概念并探讨二者统一关系。

10 Crazy Facts About AI That Will Blow Your Mind

@ActionModelAI · 57.8K 粉丝 · 5.7K 阅 · 501 赞 · 366 转

Most people think AI is still in its early stages. They're right. And that's exactly why these facts are so crazy. 1. AI could add $15.7 trillion to the global economy by 2030 According to PwC, AI

中文介绍 列举10个惊人AI事实,如2030年AI将为全球经济贡献15.7万亿美元、模型数量暴增等。虽为常见综述,但数据具体,适合快速了解AI影响力。

The Nervous System of the Autonomous Age

@OptimaiNetwork · 99.4K 粉丝 · 5.3K 阅 · 511 赞 · 360 转

The old internet was built for humans browsing pages, clicking ads, and buying subscriptions. The autonomous internet needs something else: live context, verified signals, permissioned execution, and

中文介绍 自主互联网需要全新底层架构:实时上下文、验证信号、授权执行。博主提出「自主时代神经系统」概念,描述面向AI Agent而非人类浏览的互联网基础设施。

ORACLE: Official AI Agents Trade on Polymarket

@Trade_OracleAI · 21.1K 粉丝 · 60 阅 · 2.8K 赞 · 584 转

In 2026, autonomous AI agents have become one of the most effective strategies on prediction markets. Over 30% of all activity on Polymarket now comes from algorithmic and AI-powered wallets. We

中文介绍 ORACLE官方AI Agent在Polymarket交易,利用自主策略获取预测市场收益。展示AI在金融预测中的实战表现,超30%市场活动已来自算法驱动。

A startup claims it broke through a bottleneck that’s holding back LLMs

Miami-based AI startup Subquadratic came out of stealth mode last month with a huge claim. It announced that it had solved a mathematical bottleneck that had been holding back large language models for almost a decade. The details were thin, and many people were unconvinced. But Subquadratic has sta

中文介绍 位于迈阿密的AI初创公司Subquadratic宣布走出隐身模式,声称已解决制约大型语言模型近十年的数学瓶颈。但细节有限,多数人对此表示怀疑。

The Professor of Outputmaxxing — Anjney Midha, AMP

We talk about how this legendary investor went from humble beginnings in Singapore to leading rounds in Anthropic, Mistral, Black Forest Labs, and Periodic Labs... and the AMP secret master plan!

中文介绍 与Anjney Midha对话:这位传奇投资人从新加坡白手起家,主导了对Anthropic、Mistral、Black Forest Labs及Periodic Labs的投资,并透露AMP秘密计划。

New usage analytics and updated spend controls for enterprises

OpenAI introduces new spend controls and usage analytics for ChatGPT Enterprise, helping organizations manage costs and scale AI with confidence.

中文介绍 OpenAI为ChatGPT Enterprise推出新的用量分析和更新支出控制功能,帮助组织管理成本并自信扩展AI应用。

Improving health intelligence in ChatGPT

Learn how GPT-5.5 Instant improves ChatGPT’s health and wellness responses with stronger reasoning, better context, clearer communication, and physician-informed evaluations.

中文介绍 GPT-5.5 Instant版ChatGPT改善了健康和保健回复,具备更强推理、更佳语境、更清晰沟通,并参考了医生评估。

Using AI to help physicians diagnose rare genetic diseases affecting children

Researchers used an OpenAI reasoning model to help diagnose rare diseases, identifying 18 new diagnoses in previously unsolved cases.

中文介绍 研究人员利用OpenAI推理模型辅助医生诊断罕见儿童遗传病,在先前未确诊的病例中发现了18种新诊断。

🔬 The Self-Driving Lab — Joseph Krause, Radical AI

Radical AI's Joseph Krause on why the moat in materials is the lab, not the model

中文介绍 Radical AI的Joseph Krause表示,材料科学中真正的护城河在于实验室而非模型,强调实验自动化的重要性。

市场总览

加密市场当前极度恐慌(恐慌贪婪指数23),总市值2.26T,BTC主导率56.2%,ETH占9.1%。各资产类别的技术面分化明显:美股方面,QQQ、SPY、NVDA等科技ETF在近1日反弹(QQQ+2.51%,NVDA+2.95%),且多头排列,但MSFT(-2.8% pct5Day,RSI34.9)和META(空头排列)表现偏弱,整体中性偏多。加密资产全面空头:BTC、ETH、SOL均低于所有均线,RSI在37-43之间,短期无反转信号。中概股(BABA、PDD、JD、0700.HK)全线空头,其中BABA RSI 24.7超卖,但未见底背离。商品外汇方面,黄金期货(GC=F)跌破SMA20和SMA50,死叉形成,原油(CL=F)近5日暴跌9.83%,RSI31.2;而美元指数DXY强势上行,RSI70.9超买,接近52周高,反映避险情绪。宏观指标中,10Y美债收益率(^TNX)多头排列,VIX空头排列,表明市场波动预期下降。整体来看,权益类资产技术面修复中,但加密、中概和商品仍承压,美元走强对风险资产形成压制。

今日关注

BABA 阿里巴巴 (BABA)
偏下行

当前价格107.10,远低于SMA20(120.91)、SMA50(129.30)和SMA200(149.19),呈空头排列。RSI 14 为24.7,处于超卖区域,MACD柱状图-6.2971,信号线-4.8371,绿柱持续缩短,但尚未金叉。近5日下跌4.96%,整体技术面偏弱,超卖状态可能短期反弹,但趋势仍空头主导。

DX-Y.NYB 美元指数 DXY
偏上行

当前价格100.85,接近52周高点(仅差-0.27%),SMA20(99.68)>SMA50(98.96)>SMA200(98.70),多头排列明显。RSI 14 为70.9,处于超买区域,MACD 金叉(0.3995>0.3057),且近5日上涨1.10%,动量向上。超买提示回调风险,但短期趋势仍偏强。

QQQ Nasdaq 100 ETF
偏上行

当前价格740.62,距52周高仅-1.07%,SMA20(726.88)>SMA50(693.18)>SMA200(628.63),多头排列。RSI 14 为59.1,处于正常偏强区间,MACD 柱状图9.5234,信号线11.1108,虽未金叉但绿柱为正。近1日上涨2.51%,近5日上涨3.28%,短期动能强劲。

CL=F WTI 原油期货
偏下行

当前价格76.54,远低于SMA20(87.49)和SMA50(93.89),但高于SMA200(73.73),均线空头排列。RSI 14 为31.2,接近超卖,MACD 死叉(-5.1711< -3.7021),绿柱放大。近5日暴跌9.83%,空头动量强劲,近期价格可能测试200日均线支撑。

BTC-USD Bitcoin
偏下行

当前价格63420.39,低于所有短期均线(SMA20 64991.51,SMA50 72921.13,SMA200 77119.63),空头排列。RSI 14 为37.2,偏弱,MACD 绿柱-2489.10,信号线-2960.60,但MACD线略高于信号线(死叉有收敛迹象)。整体受加密恐慌贪婪指数23(极度恐慌)影响,市场情绪低迷,趋势仍偏空。

全部资产

^VIX

VIX 恐慌指数

$16.78 +2.32%
5 日
-5.09%
距 52w 高
-52.5%
RSI(14)
47.1
趋势
空头
SMA 20 / 50 / 200
17.41 / 17.74 / 18.57
MACD / 信号
-0.173 / -0.077
死叉(SMA50↓SMA200) (7 天前)MACD 死叉 (1 天前)空头排列

^TNX

10Y 美债收益率 (%)

$4.45 -0.27%
5 日
-1.87%
距 52w 高
-10.9%
RSI(14)
50.8
趋势
多头
SMA 20 / 50 / 200
4.52 / 4.42 / 4.21
MACD / 信号
0.020 / 0.029
多头排列

DX-Y.NYB

美元指数 DXY

$100.85 -0.00%
5 日
+1.10%
距 52w 高
-0.3%
RSI(14)
70.9
趋势
多头
SMA 20 / 50 / 200
99.68 / 98.96 / 98.70
MACD / 信号
0.400 / 0.306
MACD 金叉 (2 天前)RSI 超买接近 52 周高多头排列

SPY

S&P 500 ETF

$746.74 +0.78%
5 日
+1.22%
距 52w 高
-1.8%
RSI(14)
54.1
趋势
多头
SMA 20 / 50 / 200
747.08 / 729.66 / 688.36
MACD / 信号
3.930 / 5.524
接近 52 周高多头排列

QQQ

Nasdaq 100 ETF

$740.62 +2.51%
5 日
+3.28%
距 52w 高
-1.1%
RSI(14)
59.1
趋势
多头
SMA 20 / 50 / 200
726.88 / 693.18 / 628.63
MACD / 信号
9.523 / 11.111
接近 52 周高多头排列

AAPL

Apple

$298.01 +0.70%
5 日
+0.81%
距 52w 高
-6.1%
RSI(14)
50.9
趋势
多头
SMA 20 / 50 / 200
303.40 / 288.74 / 268.19
MACD / 信号
1.180 / 3.216
多头排列

MSFT

Microsoft

$379.40 +0.13%
5 日
-2.80%
距 52w 高
-31.7%
RSI(14)
34.9
趋势
空头
SMA 20 / 50 / 200
413.15 / 412.98 / 451.35
MACD / 信号
-8.571 / -3.544
空头排列

NVDA

Nvidia

$210.69 +2.95%
5 日
+2.84%
距 52w 高
-10.9%
RSI(14)
50.4
趋势
多头
SMA 20 / 50 / 200
211.79 / 209.31 / 189.90
MACD / 信号
-1.067 / -0.113
多头排列

GOOGL

Alphabet

$368.03 +1.17%
5 日
+2.87%
距 52w 高
-9.9%
RSI(14)
49.1
趋势
多头
SMA 20 / 50 / 200
371.63 / 367.37 / 311.10
MACD / 信号
-1.850 / -0.734
多头排列

TSLA

Tesla

$400.49 +1.04%
5 日
+0.34%
距 52w 高
-19.7%
RSI(14)
47.0
趋势
空头
SMA 20 / 50 / 200
413.70 / 402.49 / 416.96
MACD / 信号
-2.850 / -0.505
空头排列

META

Meta

$577.22 +1.70%
5 日
+1.55%
距 52w 高
-27.5%
RSI(14)
42.8
趋势
空头
SMA 20 / 50 / 200
599.48 / 621.90 / 654.92
MACD / 信号
-11.455 / -9.861
空头排列
加密恐慌贪婪
23
极度恐慌
加密总市值
$2.26 T
+0.30% / 24h
BTC 主导率
56.2%
ETH 9.1%
24h 成交量
$60.2 B
活跃币 17,423

BTC-USD

Bitcoin

$63,420.39 +0.83%
5 日
-1.55%
距 52w 高
-49.7%
RSI(14)
37.2
趋势
空头
SMA 20 / 50 / 200
64,991.51 / 72,921.13 / 77,119.63
MACD / 信号
-2,489.099 / -2,960.601
空头排列

ETH-USD

Ethereum

$1,707.22 -0.14%
5 日
+1.61%
距 52w 高
-65.5%
RSI(14)
38.7
趋势
空头
SMA 20 / 50 / 200
1,749.72 / 2,024.37 / 2,382.75
MACD / 信号
-86.101 / -107.682
MACD 金叉 (4 天前)空头排列

SOL-USD

Solana

$69.75 +0.18%
5 日
-1.99%
距 52w 高
-72.5%
RSI(14)
42.5
趋势
空头
SMA 20 / 50 / 200
69.85 / 80.19 / 98.10
MACD / 信号
-2.644 / -3.613
空头排列

BABA

阿里巴巴 (BABA)

$107.10 -0.32%
5 日
-4.96%
距 52w 高
-44.4%
RSI(14)
24.7
趋势
空头
SMA 20 / 50 / 200
120.91 / 129.30 / 149.19
MACD / 信号
-6.297 / -4.837
RSI 超卖空头排列

PDD

拼多多 (PDD)

$79.56 -0.38%
5 日
-2.14%
距 52w 高
-42.9%
RSI(14)
32.0
趋势
空头
SMA 20 / 50 / 200
85.43 / 93.76 / 110.59
MACD / 信号
-4.046 / -3.933
接近 52 周低空头排列

JD

京东 (JD)

$27.57 -1.22%
5 日
-1.75%
距 52w 高
-25.2%
RSI(14)
35.6
趋势
空头
SMA 20 / 50 / 200
29.07 / 30.07 / 30.25
MACD / 信号
-0.650 / -0.511
空头排列

0700.HK

腾讯控股 (0700.HK)

HK$440.20 -1.17%
5 日
-3.72%
距 52w 高
-35.5%
RSI(14)
42.4
趋势
空头
SMA 20 / 50 / 200
449.04 / 469.34 / 566.02
MACD / 信号
-4.590 / -5.040
空头排列

GC=F

黄金期货

$4,172.90 -1.21%
5 日
-1.00%
距 52w 高
-25.3%
RSI(14)
36.5
趋势
中性
SMA 20 / 50 / 200
4,358.34 / 4,545.92 / 4,439.32
MACD / 信号
-94.957 / -92.742
MACD 死叉 (今天)

CL=F

WTI 原油期货

$76.54 -0.08%
5 日
-9.83%
距 52w 高
-35.9%
RSI(14)
31.2
趋势
中性
SMA 20 / 50 / 200
87.49 / 93.89 / 73.73
MACD / 信号
-5.171 / -3.702

USDCNY=X

美元 / 人民币

¥6.77 -0.01%
5 日
+0.03%
距 52w 高
-6.1%
RSI(14)
42.5
趋势
空头
SMA 20 / 50 / 200
6.77 / 6.80 / 6.96
MACD / 信号
-0.011 / -0.012
接近 52 周低空头排列
风险提示

请注意,过去走势不代表未来表现,本报告仅基于公开行情数据计算的技术指标进行客观解读,不构成任何投资建议。技术分析存在滞后性和局限性,投资者应结合自身风险承受能力审慎决策。仅供技术指标解读参考。

Iran war live: Tehran says US must ensure Israel ends attacks on Lebanon

Iranian deputy foreign minister says Iran 'ready to move forward' on diplomacy with US, but war must end on all fronts.

中文摘要 伊朗副外长称伊朗「愿与美国推进外交」,但要求美国确保以色列停止对黎巴嫩的攻击,战争必须在所有战线结束。

US to stop funding HIV programmes in South Africa

More than eight million South Africans are living with HIV – the highest number of any country in the world.

中文摘要 美国将停止资助南非的HIV项目。南非有超过800万人感染HIV,为全球最高。

Lebanon Emerges as Weak Link in U.S.-Iran Deal to End War

The conflict between Israel and Hezbollah, once seen as a secondary front to the American-Israeli war on Iran, has become one of the main obstacles to ending it.

中文摘要 黎巴嫩成为美伊终止战争协议中的薄弱环节。以色列与真主党的冲突原被视为美以对伊战争的次要战线,现成为主要障碍之一。

Thirty dead at DRC displacement camp as Ebola threat grows

At least 30 people have died since May in the Kigonze displacement camp in the Democratic Republic of Congo.

中文摘要 刚果(金)Kigonze流离失所营地自5月以来至少30人死亡,埃博拉威胁加剧。

Zelensky stripped of highest Polish honour over WW2 name of army unit

Ukraine has denounced the move, calling it a "strategic mistake" and "disrespectful".

中文摘要 乌克兰总统泽连斯基被剥夺波兰最高荣誉,因其部队使用二战名称。乌克兰谴责此举为「战略错误」和「不尊重」。

Here’s the latest.

中文摘要 美伊协议谈判最新动态:目前暂无更多细节披露。

Does Trump have to submit the Iran memorandum of understanding to Congress?

A 2015 law requiring congressional approval for any Iran nuclear deal is in the spotlight after this week's memo.

中文摘要 本周美伊谅解备忘录签署后,2015年一项要求国会批准任何伊朗核协议的法律受到关注。特朗普是否需要将该备忘录提交国会?

US announces new round of Israel-Lebanon talks in Washington next week

Talks come as Hezbollah, Israel say new ceasefire has been reached in wake of US-Iran memorandum to end war.

中文摘要 美国宣布下周在华盛顿举行新一轮以色列-黎巴嫩谈判。此前真主党与以色列表示已达成新停火协议,美伊签署结束战争备忘录。

Israel kills ‘a child a day’ on average in Gaza despite ceasefire, UN says

'No ceasefire can be considered meaningful while children continue to be killed' says UNICEF.

中文摘要 联合国称,尽管停火,以色列在加沙平均每天仍杀死一名儿童。联合国儿童基金会表示,「只要儿童继续被杀,任何停火都不能视为有意义」。

Italy's Meloni, once Trump's closest ally in Europe, says he made up a story about her

"Italy and I do not beg," Meloni said in a video rebuke posted on social media Friday. Italy's top diplomat, meanwhile, said he was cancelling a visit to the U.S because of the alleged remarks.

中文摘要 意大利总理梅洛尼发表视频驳斥特朗普,称「意大利和我不会乞求」。意最高外交官因特朗普言论取消访美计划。

The UK: No. 10 Revolving Door

UK PM Keir Starmer faces growing pressure after Andy Burnham’s by-election win.

中文摘要 英国首相斯塔默面临压力,因安迪·伯纳姆在补选中获胜,唐宁街10号「旋转门」现象引发关注。

One person dead and 89 injured after UK train crash

Collision on the Midland Main Line is the first fatal rail incident in nearly two years

中文摘要 英国米德兰主线发生火车相撞事故,造成1人死亡、89人受伤,是近两年来首起致命铁路事故。

Trump accepts Qatar jet as new Air Force One despite ethics concerns

The $400mn gift has raised charges of conflicts between personal interests and official duties

中文摘要 特朗普接受价值4亿美元的卡塔尔喷气式飞机作为“空军一号”,此举被指利益冲突,引发道德争议。

Value Investing Legend Seth Klarman: Masters in Business

Barry sits down with Munger disciple Seth Klarman, CEO of Baupost Group, a Boston-based investment manager with a multi-strategy approach. They discuss Seth's start as a 25 year-old and journey to CEO. They also discuss his approach to risk, IPOs, and sectors along with his sports passions including

中文摘要 价值投资大师、Baupost Group首席执行官Seth Klarman接受访谈,分享其从25岁起步到成为CEO的历程,以及风险、IPO和行业投资观点。

Starbucks Cuts UK, Hong Kong Office Jobs in Restructuring Effort

Starbucks Corp. laid off corporate workers in the London and Hong Kong hubs that oversee parts of its international business, as the coffee chain gives third-party licensees greater latitude to run its stores outside of North America.

中文摘要 星巴克在伦敦和香港的企业办公室裁员,作为重组举措,将北美以外门店运营权更多交给第三方授权商。

Canada Imposes Temporary 10% Tariff on Canned Vegetables

Canada applied a 10% import tax on imports of canned vegetables to protect domestic growers and food processors, the Department of Finance said in a statement Friday.

中文摘要 加拿大对进口罐装蔬菜征收10%临时关税,以保护国内种植商和食品加工商。

Germany’s financial watchdog removes three Berenberg bosses

BaFin’s move comes as country’s oldest lender reports possible breaches of corporate governance

中文摘要 德国金融监管机构BaFin以可能存在公司治理违规为由,撤换贝伦贝格银行三位高管。

Brookfield Is Said to Lead Bidding for Drahi’s XpFibre Business

Brookfield Asset Management Ltd. is emerging as the frontrunner to acquire a controlling stake in telecom tycoon Patrick Drahi’s French fiber optic company XpFibre, according to people with knowledge of the matter.

中文摘要 Brookfield资产管理公司成为收购电信大亨Patrick Drahi旗下法国光纤公司XpFibre控股权的领先竞标者。

Gold coin salesmen orchestrated plot to set up rival, court finds

Ex-employees of Hattons of London took confidential customer data amid a booming market for bullion

中文摘要 法院裁定,Hattons of London前雇员在黄金市场繁荣期间窃取客户数据,策划设立竞争公司。

First Quantum Deemed Broadly Compliant at Panama Copper Mine

Panama moved closer to deciding whether First Quantum Minerals Ltd. can reopen a giant copper mine after receiving a long-awaited audit that found the project met most of its environmental, legal, fiscal and operational obligations.

中文摘要 巴拿马审计显示First Quantum旗下巨型铜矿在环境、法律、财政义务上基本合规,矿山重启进程接近决定。

【CHY公益站】继续补充额度

本帖使用社区公益推广,符合推广要求。我申明并遵循社区要求的以下内容: 我的项目是免费使用的,无收费(变相收费、赞助)部分: 是 我的帖子已经打上 公益推广 标签: 是 我的项目属于个人项目,与公司或商业机构无关: 是 我的项目不存在QQ、TG等群组引流: 是 我的项目不存在非运营必要的网站引流: 是 我的项目不存在为他人推广、AFF: 是 我的项目无关联的商业项目: 是 我的站点存在登录,并已接入 LINUX DO Connect: 是 我帖子内的项目介绍,AI生成、润色内容部分已截图发出: 是 以上选择我承诺是永久有效的,接受社区和佬友监督: 是 以下为项目介绍正文内容,AI生成、润色内容已

Anthropic CEO 达里奥明确表示:在百度的经历并未塑造我对中国的看法

Emily Chang (主持人): 你在百度的所见所闻,是否塑造了你对中国的看法? Dario Amodei (Anthropic CEO): 其实并没有。我在那里工作了一年。我想我学到更多的可能是关于语音、语音识别以及诸如此类的技术… Inside the Mind of Anthropic CEO Dario Amodei | The Circuit | Extended Interview 21 个帖子 - 18 位参与者 阅读完整话题

不敢睡,号池可用账号没30+真的不敢睡

很抱歉昨天起床晚了,中午14:00号池榨干拉闸一段时间 这位国际友人的扫码次数已经被我榨干了 夜猫福利,今天签到88.88刀额度最后15分钟(1:30关闭签到) @jiven @FoundZiGu 最近提供的帮助 公益推广申明 状态 我的项目是免费使用的,无收费(变相收费、赞助)部分 是 我的帖子已经打上 公益推广 标签 是 我的项目属于个人项目,与公司或商业机构无关 是 我的项目不存在 QQ、TG 等群组引流 是 我的项目不存在非运营必要的网站引流 是 我的项目不存在为他人推广、AFF 是 我的项目无关联的商业项目 是 我的站点存在登录,并已接入 LINUX DO Connect 是 我帖子

一个Free号池的Image2生图公益站

地址:https://img2.design 可直接使用linuxDO登录,以 令牌桶 方式限流不限额,后续更高并发我以CDK形式发放兑换码吧 目前实现: AI提示词优化 原图编辑 图片智能增强(在不改变主题的情况下进行图片效果优化增强) 图片反推提示词 多图上传合并 SVG像素生成 API 同/异 调用 高并发出图 预设多组比例格式/多种预设风格 免费 gpt号池,没有高质量图。 下面是高并发令牌桶兑换码 (点击了解更多详细信息) 上面的码如果用完了,就CDK领吧,20LDC cdk.linux.do LINUX DO CDK Linux Do 社区 CDK 快速分享平台 - 让分享变得更简

世界癫了还是我癫了

想了很久还是发出来,昨晚吧,一件事情彻底的颠覆了我的三观,虽然我本身已经很渣了,可以说是扎到骨子里的那种,我端午节随机约见了一位在我眼里能打的上8分的女生(满分10分),一起去吃饭、买盲盒、去极地海洋公园啥的,晚上一起去了酒吧,然后自然而然的最后就去了维也纳,该发生的也都发生了,离谱的事情来了,那啥完她去洗澡了,迷迷糊糊的,有一个电话打过来,我以为是我的手机响了,我就直接接了,结果是一个男生,我还懵了,当时脑子没转过来,我直接朝着浴室喊了一句XX你的电话,然后那个男生就开始崩溃了应该是,我本以为他会说很多难听的话啥的,结果离谱的是还在求,最后我才知道那个男生应该是个舔狗,舔了半年了,差不多每次

阿里云300块的券别急着买服务器,还有GLM5.2!

先放个链接,去领300的券,已经用来买服务器的佬别着急,还能退的(没带aff不能杀我吧 university.aliyun.com 云工开物_阿里云高校用云支持计划,支持高校AI人才培养和科研创新-阿里云 “云工开物”是由阿里云发起的高校用云支持计划,支持AI时代高校人才培养和科研创新。与高校合作共建人工智能通识课,为大学生送免费算力资源。 通过验证变成立即使用就可以了,然后创建一个apikey bailian.console.aliyun.com 大模型服务平台百炼控制台 百炼控制台是阿里云大模型服务平台,提供AI模型训练、部署、推理一站式服务,支持多种大模型框架,助力企业快速构建AI应用。

我毕业啦!!

我这个臭本科还是民办的肯定比不上站里各位92​,还是在家里县城上找了一个小工作,大富大贵不会有,自己一人肯定够活了。接下来 8 月就要迈入牛马生活了,还有人生最后一个暑假过! 33 个帖子 - 32 位参与者 阅读完整话题

【CHY公益站】大概就是模型恢复了吧

懒得处理前端的一大堆冲突了,套的New-API前端,换了个自写后端,应该没什么问题了 如果请求遇到空回或429说明号池的号临时负载满了,等一会再试即可 17 个帖子 - 14 位参与者 阅读完整话题

害!训练模型!

关羽过江 我也过江! 40 个帖子 - 38 位参与者 阅读完整话题