每日简报

2026-09-29

← 历史归档

debpalash/VoiceStudio

Python · ★ 44,269 · 🍴 5,108 · 📈 3,221 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 ElevenLabs 的开源本地替代方案,无需云端即可完成语音克隆、音色设计、视频配音、听写、转写和有声书制作,支持 646 种语言。适合对隐私和成本敏感的内容创作者、播客与有声书制作者,以及需要本地化配音的团队。

paperclipai/paperclip

TypeScript · ★ 92,903 · 🍴 15,902 · 📈 3,197 stars today

The open-source app everyone uses to manage agents at work

中文介绍 用于在工作场景中管理 AI agent 的开源应用。它把多个 agent 的配置、任务分发与运行状态集中到一个界面统一调度,解决 agent 分散在各处、难以协同和追踪的问题,适合在团队中落地 agent 工作流的开发者与运营人员。

vectorize-io/hindsight

Python · ★ 41,054 · 🍴 5,544 · 📈 4,561 stars today

Hindsight: Agent Memory That Learns

中文介绍 面向 AI agent 的记忆层,强调“会学习”的长期记忆能力:通过记录历史交互并从中提炼可复用经验,让 agent 在后续任务中持续改进,而非每次从零开始。适合构建多轮对话、长周期任务的开发者,用来解决上下文丢失与经验无法沉淀的问题。

NawfalMotii79/PLFM_RADAR

PLSQL · ★ 25,780 · 🍴 5,887 · 📈 158 stars today

Open-source, low-cost 10.5 GHz PLFM phased array RADAR system

中文介绍 开源低成本 10.5 GHz PLFM 相控阵雷达系统,公开硬件设计与信号处理方案,把通常昂贵的相控阵与脉冲压缩雷达带到个人和小团队可承受的成本。适合雷达、射频与嵌入式爱好者做测距测速实验,也便于高校教学和科研原型验证。

cs341-illinois/coursebook

TeX · ★ 2,537 · 🍴 239 · 📈 195 stars today

Open Source Introductory Systems Programming Textbook for the University of Illinois

中文介绍 伊利诺伊大学 CS341 课程的开源系统编程入门教材,内容涵盖 C 语言、内存管理、进程与并发等底层主题,配套课程讲义与练习。适合计算机专业学生自学系统编程,也可供教师作为课程教材或参考资料直接使用与改编。

byoungd/up

JavaScript · ★ 64,727 · 🍴 6,516 · 📈 327 stars today

An advanced guide which might benefit you a lot 🎉 . 韩先凯的人生进阶指南 人生进阶指南 离谱的人生 人生进阶 AI学习 AI指南 韩先凯的AI学习指南 英语学习指南/英语学习教程/英语学习/学英语

中文介绍 一份个人成长与技术学习的进阶指南合集,内容涵盖 AI 学习方法与工具、英语学习路径以及职业与人生规划建议。适合希望系统入门 AI、提升英语并规划学习路线的自学者,可当作长期参考的学习地图而非速成教程。

mvschwarz/openrig

TypeScript · ★ 1,756 · 🍴 139 · 📈 734 stars today

Multi-agent harness that runs Claude Code and Codex together as one system

中文介绍 多 agent 编排框架,把 Claude Code 与 Codex 两个编码 agent 当作同一系统中的协作角色来运行,实现任务分工、互相调用与结果汇总。适合需要同时利用多个编码模型、希望在本地构建自动化开发流水线的工程师。

dream-num/univer

TypeScript · ★ 21,295 · 🍴 1,805 · 📈 1,099 stars today

The Office Harness for AI Agents — Spreadsheets, Docs, Slides, Canvas, Relational Tables, and PDF in one runtime.

中文介绍 把电子表格、文档、幻灯片、画布、关系表和 PDF 统一在同一个运行时中的 Office 框架,专门为 AI agent 提供可读写的文档操作能力。适合需要让 agent 直接生成和编辑办公文件的开发者,用于自动化报表、文档生成等场景。

Your Transformer Can Hold Two Thoughts at Once: Evidence of Linear Superposition in LLMs

👍 80

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 研究发现,尽管 LLM 由高度非线性组件构成,却展现出基本线性性质:将来自不同文本流的输入线性组合后,模型输出恰为各自 next-token 分布的叠加,作者称之为 linear superposition,为理解 LLM 的表示与并行处理能力提供证据。

Disaggregated Quantization: Specializing LLM Prefill and Decode

👍 45

Prefill and decode reward different approaches to quantization: low-precision arithmetic accelerates prompt processing, while compact weights reduce memory traffic during generation. We propose "disaggregated quantization" (DQ), which specializes computation formats, weights and storage placement to

中文介绍 Prefill 与 decode 对量化的需求不同:低精度算术可加速 prompt 处理,紧凑权重则降低生成阶段的内存流量。本文提出 disaggregated quantization(DQ),为两个阶段分别特化计算格式、权重与存储位置,同时优化预填充与解码效率。

Rufus-Air: An Open LLM Post-Training Recipe

👍 21

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-A12B)上开源、可复现的后训练配方,由八个串行阶段组成:SFT、推理 RL、代码 RL、指令遵循 RL、通用 Agent、代码 Agent、搜索 Agent 与 RLHF,并公开数据、奖励设计与基础设施细节。

Block Sparse Attention with Log-Linear Complexity

👍 20

Scaling language models to long contexts is limited by the quadratic cost of self-attention. Block sparse attention offers an efficient alternative, but selecting the retained blocks remains a bottleneck. Conventional block selection requires scoring all query-block pairs and therefore remains quadr

中文介绍 长上下文自注意力的二次开销可由块稀疏注意力缓解,但传统块选择需为所有 query-block 对打分,仍为二次复杂度并成为瓶颈。本文提出对数线性复杂度的块稀疏注意力方法,降低选块代价,使长上下文建模更具可扩展性。

Learning to Discover Interesting Mathematics

👍 13

Recently, Large Language Models (LLMs) have been increasingly able to solve advanced mathematical problems, including many that have been open for decades. This opens the door to expansion of mathematical knowledge at unprecedented scale. Yet, while LLMs may be able to conjecture and prove more and

中文介绍 LLM 已能求解许多悬置数十年的数学难题,使数学知识的大规模扩展成为可能,但模型能否提出有价值的新猜想仍不清楚。本文研究如何让模型「学习发现有趣的数学」,以生成与筛选更具研究价值的数学内容。

Coding Agents for Generalized Task and Motion Planning Problems

👍 11

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、动力学约束紧耦合,即使完全可观测、状态以物体为中心也依然困难;广义 TAMP 试图利用跨问题实例的规律性。本文提出用 coding agents 求解广义 TAMP 问题。

PUBG Ally: A Conversational Embodied Agent as an AI Teammate

👍 11

We introduce PUBG Ally, an embodied agent for PUBG: BATTLEGROUNDS that can reason, act autonomously, and play alongside players as a voice-enabled teammate. Building such a teammate requires combining two difficult capabilities: it must perceive and respond to a constantly changing game world under

中文介绍 提出 PUBG Ally,一个面向《绝地求生》的具身智能体,能以语音队友身份与玩家并肩作战,自主推理并行动。其难点在于同时具备两项能力:在持续变化的游戏世界中感知与响应,以及与人类进行实时语音交互。

Just Ask Jev: Reinforcement Learning for Calibrated Decisions as a Zero-Shot Detector of AI Alignment Failures

👍 10

Detectors of alignment failures screen deployed language models and score alignment benchmarks. Most are generative judges that spend a decoding pass on every criterion, and classifiers that read token probabilities, such as Llama Guard, still score one fixed label per call. Jev, a model trained wit

中文介绍 现有对齐失败检测器多为生成式评判器,每条标准都要消耗一次解码,或如 Llama Guard 每次只输出固定标签。本文提出 Jev,用强化学习训练校准决策模型,以零样本方式检测对齐失败,支持选项、二值判断与打分输出,成本更低。

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 LLM agent 分解复杂查询、检索证据并综合有依据的答案,但现有 ReAct 式 agent 存在角色耦合(单一策略兼顾规划、证据使用与合成)与上下文持续累积两大问题。本文提出角色解耦的迭代合成方法 IterSynth 加以解决。

Jev in the Wild: A Data-Driven Analysis of the Jev Model's Functionality, Applications and Ecosystem

👍 8

Jev is a fast, low-cost decision model that answers natural-language questions with choices, binary judgments, and scores. As its public ecosystem grows rapidly, it remains unclear how Jev is used across applications and how public attention relates to project distribution. To answer these questions

中文介绍 Jev 是一个快速、低成本的决策模型,以选项、二值判断和打分形式回答自然语言问题。随着其公开生态迅速扩张,其实际使用方式以及公众关注度与项目分布之间的关系仍不清晰。本文对 Jev 的功能、应用与生态系统进行数据驱动分析。

RGBD20K: A Large-Scale Benchmark for RGB-D Semantic Segmentation

👍 8

In this paper, we propose RGBD20K, a novel dataset for facilitating the development of more robust and general RGB-D semantic segmentation by encompassing abundant categories and high-quality annotations. RGBD20K possesses several attractive properties: (1) Expanded Semantic Space. In particular, it

中文介绍 提出 RGBD20K,一个用于推动更鲁棒、更通用的 RGB-D 语义分割的大规模数据集,具备类别丰富、标注质量高等特点,并显著扩展语义空间,为该任务的泛化性与鲁棒性研究提供新基准。

Neural Spectral Capacity: Measuring and Designing Architectures from Network Specification Alone

👍 8

Modern Transformer design and compression both reduce to allocating capacity under a budget. The standard scalars for these decisions, #Params and #FLOPs, capture size and compute but not architectural structure: two architectures with identical parameter budgets but different depth-width, head, or

中文介绍 Transformer 设计与压缩本质是在预算下分配容量,但常用的 #Params 与 #FLOPs 只反映规模与计算,无法刻画架构结构:参数预算相同而深度宽度、head 配置不同的架构难以区分。本文提出 Neural Spectral Capacity,仅从网络规格衡量并指导架构设计。

TimeEvo: Failure-Driven Self-Evolution of a Time Series Agent

👍 7

Time series agents answer analytical questions by calling external tools, and which tools they carry is decided by people before the agent runs. However, we identify two failures in this setup. Human-Agent Tool Misalignment: a library of 21 expert-curated tools helps on some tasks and hurts on other

中文介绍 时序 agent 依靠调用外部工具回答分析问题,但工具集由人工预先设定,作者指出两类失败,其中人-代理工具错配表现为 21 个专家精选工具在部分任务上有帮助、在另一些任务上反而有害。TimeEvo 通过失败驱动的自演化让 agent 自行调整所需工具。

AgentWorld: Benchmarking Long-Horizon Collaboration of Multi-agent LLMs

👍 7

Existing multi-agent benchmarks primarily test in competitive settings, short-horizon interactions under 20 steps, or simply aggregate individual performance, failing to isolate and highlight genuine collaboration capabilities of LLM-based agents. We introduce AgentWorld, a benchmark of 100 human-an

中文介绍 现有多智能体基准多聚焦竞争场景、20 步以内的短时程交互,或仅聚合个体表现,难以隔离并凸显 LLM agent 的真正协作能力。本文提出 AgentWorld,一个包含 100 项人工设计任务、面向长时程协作的多智能体基准。

SLCA-GRPO: Resolving Cross-Segment Credit Misattribution in Tool-Calling RL

👍 7

Tool-calling agents produce heterogeneous outputs, interleaving structured tool invocations with user-facing natural language summaries. This output heterogeneity presents a structural failure mode in standard on-policy Reinforcement Learning (RL): algorithms like GRPO indiscriminately broadcast a h

中文介绍 工具调用 agent 的输出高度异构,结构化工具调用与面向用户的自然语言总结交错出现;GRPO 等标准 on-policy RL 会把同一优势广播到所有 token,造成跨段信用误分配。本文提出 SLCA-GRPO 以解决这一问题。

AgentKernel: The Trust-Native Agentic Operating System

👍 7

Modern AI agents routinely cross trust boundaries: they ingest untrusted content, combine it with privileged instructions, persist intermediate beliefs in long-term memory, and invoke privileged tools. This creates an attack surface in which malicious payloads can enter through model inputs and caus

中文介绍 现代 AI agent 频繁跨越信任边界:摄入不可信内容、与特权指令混合、把中间信念写入长期记忆并调用特权工具,形成可被恶意载荷利用的攻击面。本文提出 AgentKernel,一个以信任为原生的 agent 操作系统,在系统层面提供隔离与管控。

SAGE: Mitigating Long-Horizon Reasoning Biases via Topological Guidance

👍 6

Long-horizon reasoning remains a central challenge for large language models (LLMs) under sparse-reward regimes. We argue that this brittleness arises from two biases induced by complex reasoning spaces: an exploration bias, where models are drawn toward locally plausible but structurally unstable b

中文介绍 长时程推理在稀疏奖励下仍是 LLM 的核心难题。作者认为其脆弱性源于复杂推理空间带来的两种偏差:探索偏差使模型偏向局部看似合理但结构不稳定的分支。SAGE 通过拓扑引导缓解这些偏差,提升长时程推理的稳定性。

DeltaWAM: Delta World Action Models for Bimanual Manipulation

👍 6

World-action models (WAMs) transfer visual and motion priors from pretrained video generators to robot control by jointly modeling visual dynamics and actions. Existing WAMs, however, predict dense future frames during training, repeatedly modeling largely unchanged content and coupling action-condi

中文介绍 World-action model 通过联合建模视觉动态与动作,把预训练视频生成器的视觉与运动先验迁移到机器人控制;但现有方法训练时预测稠密未来帧,反复建模几乎不变的内容并使动作条件相互耦合。本文提出 DeltaWAM,面向双臂操作只建模帧间变化。

IndicBankBench: Evaluating Safety and Reliability of Language Model Assistants in Indian Retail Banking

👍 4

Banking assistants must use account-specific information to answer requests and, in many cases, take actions through tools. Evaluating only the final response misses important errors. An assistant may ask for information it already has, rely on stale context, select the wrong account, or write an in

中文介绍 银行助手须结合账户专属信息作答,并常需通过工具执行操作,仅评测最终回复会漏掉关键错误,例如索要已知信息、依赖过期上下文、选错账户或写出不当内容。本文提出 IndicBankBench,评测印度零售银行场景下语言模型助手的安全性与可靠性。

CARD: Cluster-level Adaptation with Reward-guided Decoding for Personalized Text Generation

👍 4

Adapting large language models to individual users remains challenging due to the tension between fine-grained personalization and scalable deployment. We present CARD, a hierarchical framework that achieves effective personalization through progressive refinement. CARD first clusters users accordin

中文介绍 将大模型适配到单个用户面临细粒度个性化与可扩展部署之间的张力。CARD 采用层次化框架,先按用户聚类,再通过奖励引导解码逐步细化,实现高效且可扩展的个性化文本生成。

Meet your Team Bots

@n2parko · 5.7K 粉丝 · 274.2K 阅 · 533 赞 · 38 转

Today we're introducing Team Bots, and it's likely the biggest change to Grok Bot since launch. I wanted to share a bit about why we built it and how these new digital colleages might change the way

中文介绍 Grok Bot 推出 Team Bots,官方称这是该产品发布以来最大的一次改动。作者解释为何要做这些「数字同事」,以及多智能体组队协作可能怎样改变工作方式。属产品发布加设计思路说明,非操作教程。

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 作者宣布开源 company brain 项目的多人协作 harness 设计。团队两周前刚停掉该产品、把大量用户迁移到其他产品,在客户和外界持续建议下,决定把整套架构公开。偏工程复盘与架构分享。

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 Michael Saylor 的数字经济观点文:AI 将让个人和公司的产出能力远超今天,因此自由创造、融资、拥有与交换资产的权利变得更加重要。属宏观叙事与产权主张,不涉及具体工具或方法。

How to build motion design studio with Opus 5.5 ( Full-course )

@0xMovez · 36.4K 粉丝 · 78.9K 阅 · 737 赞 · 46 转

Most people who try motion design with Opus 5.5 end up with the same video: centered text on a gradient, everything fading in, a logo at the end. They don't give it a reference, don't give it a

中文介绍 用 Opus 5.5 做 motion design 的完整课程。作者指出大多数人只能得到同一类视频:居中文字、渐变背景、全部淡入、结尾挂 logo,根源是没给模型参考素材。教程围绕如何提供参照、规范流程展开。

Its not just the f*cking sandbox

@joedaroo · 12.2K 粉丝 · 41.2K 阅 · 517 赞 · 76 转

A lot of the perspective on all the AI incidents has been shared from the outside in, and little has been said from the inside looking out, through the lens of a security person living through it.

中文介绍 一名安全从业者从「内部视角」复盘 AI 安全事件,认为现有讨论多来自外部、停留在 sandbox 层面,忽略了亲历者在真实运营环境中的处境。属经验反思类内容,不是技术教程或工具推荐。

i made 10+ motion videos with opus 5.5 in 3 days. here's the whole pipeline (open source)

@RaphaelAubryy · 393 粉丝 · 40.9K 阅 · 507 赞 · 42 转

Most people trying motion design with Opus 5.5 hit the same wall: ↳ the first render is "mid": centered text, a gradient, everything fading in ↳ the music and the picture live separate lives ↳ one

中文介绍 作者 3 天内用 Opus 5.5 做出 10 多条 motion 视频,并开源整套 pipeline。他点出常见瓶颈:首版渲染平庸(居中文字、渐变、全场淡入)、音乐与画面脱节,流程正是针对这些环节做改进。

How I Make $10K Launch Videos for $0 with Opus 5.5 (Full Course)

@twoclipping · 7.5K 粉丝 · 40.5K 阅 · 510 赞 · 35 转

im open sourcing my entire workflow for making the motion designs that studios charge $5,000 to $15,000 for this uses 0 mcp, 0 external tools and im not selling you any subscription you dont need

中文介绍 作者开源自己制作高端 launch video 的全流程:用 Opus 5.5 复刻工作室报价 5000 至 15000 美元的 motion design 效果,0 个 MCP、0 个外部工具、不需要任何订阅。主打零成本与可复现。

You Don't Hate AI Art. You Hate Admitting You Like It

@lilykeilani011 · 3.0K 粉丝 · 10.7K 阅 · 521 赞 · 71 转

The Problem Isn't Always the AI It's no secret that rejection of AI-generated art is highly visible on social media. Open any post related to AI art and you'll find hundreds of negative comments.

中文介绍 针对社交媒体上对 AI 生成艺术的强烈抵制,作者提出:问题往往不在 AI 本身,很多人只是不愿承认自己其实喜欢这类作品。属观点争论贴,没有工具或操作方法。

Claude Code’s Next Era — Thariq Shihipar, Anthropic

Shipping Opus/Sonnet 5.5, Mods, Plugins, Projects, Tag while Pacing the Frontier

中文介绍 Anthropic 的 Thariq Shihipar 在 Latent Space 播客中谈 Claude Code 的下一阶段,涉及 Opus 与 Sonnet 5.5 的发布,以及 Mods、Plugins、Projects、Tag 等功能,并讨论如何在推进前沿模型的同时把握节奏。

How we will do better for Australia

OpenAI apologizes for incidents involving Australian government websites and outlines stronger safeguards and support to strengthen Australia’s cyber defences.

中文介绍 OpenAI 就涉及澳大利亚政府网站的相关事件致歉,并说明将强化安全保障措施,同时为澳大利亚提升网络防御能力提供支持。

Roundtables: The Deadly Failures of The Virtual Border Wall

Listen to the session or watch below The US has spent billions building a “virtual wall” of surveillance towers along its southern border over the past 25 years, promising they will help detect and apprehend border crossers and save lives. But a groundbreaking investigation by MIT Technology Review

中文介绍 美国过去 25 年投入数十亿美元,在南部边境修建由监控塔组成的「虚拟墙」,承诺有助发现和抓捕越境者并挽救生命。一项调查揭示了该系统的致命缺陷,相关讨论以圆桌形式呈现。

When can we say AI made a scientific discovery?

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 Wednesday, Anthropic announced that earlier this year it had launched a molecular biology lab, where Claude agents read and conjecture about hard biology pro

中文介绍 Anthropic 称其今年早前设立了一个分子生物学实验室,由 Claude 智能体阅读文献并提出假设。文章由此探讨:在什么条件下才能说 AI 真的做出了科学发现。

Who’s liable when AI agents go rogue?

MIT Technology Review Explains: Let our writers untangle the complex, messy world of technology to help you understand what’s coming next. You can read more from the series here. Over the past few months, a cascade of cyberattacks by AI agents has stunned the world. In July, OpenAI disclosed that a

中文介绍 MIT 科技评论解释性文章探讨 AI 智能体失控时由谁承担责任。近几个月多起由 AI 智能体发起的网络攻击引发关注,文章梳理相关事件并分析现有的责任归属框架。

The Lenfest Institute grows landmark program with expanded OpenAI support

OpenAI is expanding the Lenfest AI Collaborative and Fellowship Program with $5 million in funding and up to $5 million in software credits and engineering support.

中文介绍 OpenAI 扩大对 Lenfest Institute 的 AI 协作与奖学金项目的支持,提供 500 万美元资金,以及最高 500 万美元的软件额度与工程支持。

Are you a Codex Original?

We’re collecting real stories of builders, tinkerers, researchers, and creators who are using Codex to do incredible things. If you want to be a part of the next chapter of the Codex Originals program, tell us more about your story and project below.

中文介绍 OpenAI 公开征集使用 Codex 的开发者、研究者、创业者与创作者的真实故事和项目,作为 Codex Originals 项目下一阶段的内容来源。

Basis completes a tax workbook 2x faster with GPT-6 Astra

GPT-6 Astra completed a 50-tab tax workbook twice as fast as GPT-5.6 Sol, and its stronger understanding of user intent gives Basis more confidence in real-world use.

中文介绍 Basis 使用 GPT-6 Astra 完成一份含 50 个工作表的税务工作簿,速度是 GPT-5.6 Sol 的两倍;对用户意图更强的理解让 Basis 在真实业务场景中更有信心。

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 Latent Space 播客对话 OpenRouter 的 Alex Atallah 与 AMP 的 Anjney Midha,回顾其从种子轮走到被 Stripe 以 70 亿美元收购的历程。2023 年多数人还怀疑前沿模型实验室最多只能有一两家,如今已有数十家。

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 Proaction 借助 Codex、GPT-Live-1 与 GPT-6 Astra,更快地构建、运营并销售现代车队管理系统,销售额提升 60%,并节省超过 75 小时工时。

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 美国国防部预算请求显示,政府计划未来五年投入 3030 万美元改进测谎技术。该项目名为 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 AINews 内容,借当日新闻较为清淡之机介绍其幕后工作,并宣布现已开放合作。

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 介绍 Runway 的 WorldPrompt 与实时世界生成工程。GWM Worlds 2 通过持久上下文与定时动作来引导世界模型,实时生成视频与音频。

市场总览

美股方面,SPY 报 765.61,距 52 周高仅 -1.77%,位于 sma20 的 765.16、sma50 的 762.01 与 sma200 的 718.86 之上形成多头排列,RSI14 50.5 处中性区,MACD 1.3768 高于信号 1.0656 维持金叉;QQQ 736.53、RSI14 58.5 亦为多头排列,但内部明显分化——META 单日 -4.79%、TSLA 单日 -3.94%,VIX 报 16.07 且单日、5 日均 +8.07%,仍处 sma200 的 18.07 下方。加密方面,恐慌贪婪指数 73 位于「贪婪」区间,总市值 2.85T USD、24 小时 -3.57%,BTC 主导率 58.3%、ETH 11.4%;BTC 报 82997.57 呈多头排列但距 52 周高 -34.23%,ETH 的 MACD 84.7229 低于信号 88.4836 出现死叉,SOL RSI14 60.7,市值回落与情绪读数形成对照。中概整体偏弱:BABA RSI14 40.5 为空头排列,PDD 虽 MACD 金叉但 RSI14 40.9 仍处空头排列,JD RSI14 35.1,0700.HK 报 434.6、RSI14 47.2 亦为空头排列。商品与外汇方面,黄金 4163 RSI14 32.7 且空头排列、MACD -49.5659 弱于信号;WTI 原油 93.93 虽多头排列但 MACD 1.7069 低于信号 2.9267;美元指数 101.23 RSI14 67.8 接近 52 周高,10Y 美债收益率 5.24 的 RSI14 78.6 已超买,美元与利率端读数偏强。

今日关注

NVDA 英伟达 (NVDA)
偏上行

现价 228.86,站上 sma20 的 222.32、sma50 的 216.50 与 sma200 的 199.69,构成多头排列;RSI14 为 58.6 未触及超买;MACD 2.5922 高于信号线 2.0508,维持红柱并触发金叉;单日 +1.68%、5 日 +0.65%,距 52 周高仅 -3.25%,动量与趋势指标同向。

^TNX 10Y 美债收益率 (%)
偏上行

现价 5.24,单日 +1.08%、5 日 +5.58%,距 52 周高仅 -0.64%;RSI14 达 78.6 进入超买区,为全表最高读数;MACD 0.1179 高于信号 0.0935,并与 sma20 的 4.95、sma50 的 4.78、sma200 的 4.43 构成多头排列。趋势强度显著,但超买读数本身提示短线动能存在透支特征。

GOOGL Alphabet (GOOGL)
中性

现价 342.75,RSI14 48.8 贴近中轴;MACD -0.0470 已上穿信号 -0.3033,柱状体转正,但 trend 字段仍标注中性且信号列表为空;价格在 sma20 的 341.98 与 sma50 的 344.21 之间纠缠,sma200 的 338.34 位于下方;5 日 -3.44%、距 52 周高 -16.12%,方向尚未明朗。

TSLA 特斯拉 (TSLA)
偏下行

现价 357.45,单日 -3.94%、5 日 -4.76%,为美股样本中跌幅居前;MACD 4.9043 下穿信号 5.21 形成死叉;价格低于 sma200 的 395.18 与 sma20 的 366.12,仅略高于 sma50 的 347.78;RSI14 47.1 位于中性偏下,距 52 周高 -28.34%,技术结构偏弱。

GC=F 黄金期货 (GC=F)
偏下行

现价 4163,5 日 -4.88%,同时低于 sma20 的 4373.47、sma50 的 4360.48 与 sma200 的 4555.41,呈空头排列;RSI14 32.7 为全表最低读数之一,已接近超卖;MACD -49.5659 明显低于信号 -18.9390,绿柱扩张;距 52 周高 -25.48%,动量与趋势指标一致偏空。

全部资产

^VIX

VIX 恐慌指数

$16.07 +8.07%
5 日
+8.07%
距 52w 高
-54.5%
RSI(14)
51.9
趋势
中性
SMA 20 / 50 / 200
15.73 / 15.92 / 18.07
MACD / 信号
-0.111 / -0.073

^TNX

10Y 美债收益率 (%)

$5.24 +1.08%
5 日
+5.58%
距 52w 高
-0.6%
RSI(14)
78.6
趋势
多头
SMA 20 / 50 / 200
4.95 / 4.78 / 4.43
MACD / 信号
0.118 / 0.094
RSI 超买接近 52 周高多头排列

DX-Y.NYB

美元指数 DXY

$101.23 +0.03%
5 日
+0.62%
距 52w 高
-0.6%
RSI(14)
67.8
趋势
多头
SMA 20 / 50 / 200
99.99 / 99.96 / 99.22
MACD / 信号
0.428 / 0.242
接近 52 周高多头排列

SPY

S&P 500 ETF

$765.61 -0.74%
5 日
-1.02%
距 52w 高
-1.8%
RSI(14)
50.5
趋势
多头
SMA 20 / 50 / 200
765.16 / 762.01 / 718.86
MACD / 信号
1.377 / 1.066
MACD 金叉 (4 天前)接近 52 周高多头排列

QQQ

Nasdaq 100 ETF

$736.53 -1.07%
5 日
-0.67%
距 52w 高
-1.6%
RSI(14)
58.5
趋势
多头
SMA 20 / 50 / 200
721.88 / 713.48 / 666.14
MACD / 信号
7.414 / 4.693
接近 52 周高多头排列

AAPL

Apple

$338.40 -0.78%
5 日
-0.17%
距 52w 高
-2.0%
RSI(14)
61.8
趋势
多头
SMA 20 / 50 / 200
330.33 / 321.93 / 288.07
MACD / 信号
6.079 / 5.542
接近 52 周高多头排列

MSFT

Microsoft

$509.22 -1.35%
5 日
+1.52%
距 52w 高
-8.0%
RSI(14)
58.6
趋势
多头
SMA 20 / 50 / 200
499.82 / 478.12 / 432.15
MACD / 信号
6.972 / 7.435
多头排列

NVDA

Nvidia

$228.86 +1.68%
5 日
+0.65%
距 52w 高
-3.2%
RSI(14)
58.6
趋势
多头
SMA 20 / 50 / 200
222.32 / 216.50 / 199.69
MACD / 信号
2.592 / 2.051
MACD 金叉 (4 天前)多头排列

GOOGL

Alphabet

$342.75 -0.34%
5 日
-3.44%
距 52w 高
-16.1%
RSI(14)
48.8
趋势
中性
SMA 20 / 50 / 200
341.98 / 344.21 / 338.34
MACD / 信号
-0.047 / -0.303

TSLA

Tesla

$357.45 -3.94%
5 日
-4.76%
距 52w 高
-28.3%
RSI(14)
47.1
趋势
中性
SMA 20 / 50 / 200
366.12 / 347.78 / 395.18
MACD / 信号
4.904 / 5.210
MACD 死叉 (今天)

META

Meta

$715.62 -4.79%
5 日
-3.46%
距 52w 高
-8.2%
RSI(14)
60.9
趋势
中性
SMA 20 / 50 / 200
667.74 / 617.14 / 626.77
MACD / 信号
40.390 / 32.997
加密恐慌贪婪
73
贪婪
加密总市值
$2.85 T
-3.57% / 24h
BTC 主导率
58.3%
ETH 11.4%
24h 成交量
$153.7 B
活跃币 21,668

BTC-USD

Bitcoin

$82,997.57 -1.73%
5 日
-1.64%
距 52w 高
-34.2%
RSI(14)
58.9
趋势
多头
SMA 20 / 50 / 200
80,657.61 / 76,467.61 / 71,135.25
MACD / 信号
2,268.779 / 2,226.966
多头排列

ETH-USD

Ethereum

$2,664.09 -0.85%
5 日
-0.77%
距 52w 高
-44.0%
RSI(14)
60.4
趋势
多头
SMA 20 / 50 / 200
2,586.13 / 2,403.08 / 2,103.76
MACD / 信号
84.723 / 88.484
MACD 死叉 (今天)多头排列

SOL-USD

Solana

$117.06 -4.10%
5 日
+1.81%
距 52w 高
-50.7%
RSI(14)
60.7
趋势
多头
SMA 20 / 50 / 200
109.49 / 99.44 / 85.11
MACD / 信号
6.102 / 5.510
多头排列

BABA

阿里巴巴 (BABA)

$108.76 -0.89%
5 日
-6.04%
距 52w 高
-43.6%
RSI(14)
40.5
趋势
空头
SMA 20 / 50 / 200
111.16 / 117.68 / 131.90
MACD / 信号
-1.940 / -2.090
空头排列

PDD

拼多多 (PDD)

$78.49 +1.19%
5 日
-1.02%
距 52w 高
-43.7%
RSI(14)
40.9
趋势
空头
SMA 20 / 50 / 200
79.66 / 84.38 / 95.22
MACD / 信号
-1.781 / -1.908
MACD 金叉 (4 天前)空头排列

JD

京东 (JD)

$26.61 +0.38%
5 日
-2.46%
距 52w 高
-27.8%
RSI(14)
35.1
趋势
中性
SMA 20 / 50 / 200
27.26 / 29.33 / 28.95
MACD / 信号
-0.645 / -0.682

0700.HK

腾讯控股 (0700.HK)

HK$434.60 -1.18%
5 日
-3.76%
距 52w 高
-36.4%
RSI(14)
47.2
趋势
空头
SMA 20 / 50 / 200
434.78 / 448.24 / 500.71
MACD / 信号
-2.700 / -4.113
空头排列

GC=F

黄金期货

$4,163.00 -0.13%
5 日
-4.88%
距 52w 高
-25.5%
RSI(14)
32.7
趋势
空头
SMA 20 / 50 / 200
4,373.47 / 4,360.48 / 4,555.41
MACD / 信号
-49.566 / -18.939
空头排列

CL=F

WTI 原油期货

$93.93 +1.44%
5 日
-0.70%
距 52w 高
-21.4%
RSI(14)
51.6
趋势
多头
SMA 20 / 50 / 200
96.18 / 88.46 / 81.89
MACD / 信号
1.707 / 2.927
多头排列

USDCNY=X

美元 / 人民币

¥6.71 -0.08%
5 日
+0.18%
距 52w 高
-6.7%
RSI(14)
41.4
趋势
空头
SMA 20 / 50 / 200
6.71 / 6.73 / 6.83
MACD / 信号
-0.007 / -0.009
MACD 金叉 (4 天前)接近 52 周低空头排列
风险提示

本报告仅为技术指标读数的客观复述,不构成任何投资建议;过去走势不代表未来表现,指标信号可能滞后或失效,仅供技术指标解读参考。

OpenAI scraps rollout of new model over safety concerns

OpenAI also issued an update on incidents in which its models accessed Australian government systems.

中文摘要 OpenAI 以安全担忧为由取消一款新模型的发布,同时就其模型访问澳大利亚政府系统的事件发布最新说明。

Australia news live: OpenAI apologises for Medicare hack, calling it a ‘new kind of cyber incident’

Follow the day’s news live Get our breaking news email, free app or daily news podcast New South Wales police say they are investigating the suspicious deaths of two men in the state’s north. Police said they found the first man – who has since been identified as a 79-year-old – dead at a property o

中文摘要 OpenAI 就 Medicare 系统遭黑客攻击致歉,称之为「新型网络事件」;新南威尔士州警方正调查该州北部两名男子的可疑死亡。

‘Something suspicious’: greenkeeper tells Beau Lamarre-Condon double murder trial about finding bloody belongings

Former Sydney police officer, 31, has pleaded not guilty to murdering AFL umpire Jesse Baird and flight attendant Luke Davies Follow our Australia news live blog for latest updates Get our breaking news email, free app or daily news podcast Brand new AFL shorts, a Qantas card and other bloodied belo

中文摘要 悉尼前警员、31岁的 Beau Lamarre-Condon 被控谋杀 AFL 裁判 Jesse Baird 和空乘 Luke Davies,他作无罪抗辩;一名球场管理员出庭讲述发现带血物品的经过。

Netanyahu Visits Emirati Leader After Reports on Warning About Hamas Attack

Official statements said bilateral relations were discussed. But the question of who bears responsibility for failing to prevent the Oct. 7 attack is a major election issue in Israel.

中文摘要 以色列总理内塔尼亚胡访问阿联酋领导人,此前有报道称以方曾收到关于哈马斯袭击的预警。官方称会谈涉及双边关系,但谁应为未能阻止10月7日袭击负责已成为以色列大选议题。

Vietnamese government accused of abducting Melbourne human rights activist

Tran Hiep, a 50-year-old Vietnamese Australian anti-human trafficking campaigner, not heard from since 18 September Vietnamese authorities have detained three activists, including an Australian, on what have been described as “fabricated” charges. The Vietnamese government said the three confessed t

中文摘要 越南裔澳大利亚人、50岁反人口贩运活动人士 Tran Hiep 自9月18日起失联,越南政府被指实施绑架;越南当局以被指「捏造」的罪名拘留三名活动人士,其中包括一名澳大利亚人。

US prosecutors reopen case of alleged gang rape at Cornell University

It comes after the woman filed a civil lawsuit alleging she was drugged and raped at a fraternity house in 2024.

中文摘要 美国检方重启康奈尔大学涉嫌轮奸案。此前一名女性提起民事诉讼,称2024年在该校兄弟会会所被下药并遭强奸。

OpenAI scraps release of latest AI model over safety concerns

AI giant says GPT-6.1 Astra failed to meet alignment standards during internal testing.

中文摘要 OpenAI 因安全顾虑放弃发布最新 AI 模型 GPT-6.1 Astra,称其在内部测试中未达到对齐标准。

Africa’s space ambitions are moving from policy to practice

As the Africa Space Expo closes in Abidjan, African innovators are showing what the continent can build.

中文摘要 非洲航天博览会在科特迪瓦阿比让闭幕,非洲创新者展示本大陆的航天能力,显示非洲航天雄心正从政策走向实践。

Inside Al Jazeera’s UNGA coverage

Here’s a look at what it was like behind the scenes of the coverage you saw online and on TV.

中文摘要 半岛电视台发布其联合国大会报道的幕后内容,展示网络与电视端报道的制作过程。

Iran war live: Trump says he did not offer Tehran sanctions relief

US President Trump rejects a news report claiming his administration offered Iran sanctions relief and frozen funds.

中文摘要 美国总统特朗普否认有关报道,称其政府未向伊朗提出制裁减免,也未解冻相关资金。

Hurricane Polo unleashes threatening winds as it approaches Mexico

The category-three storm is expected to make landfall of north of Cabo San Lucas, a popular tourist area, on the Mexican peninsula of Baja California Sur.

中文摘要 三级飓风 Polo 携危险强风逼近墨西哥,预计在下加利福尼亚半岛旅游胜地 Cabo San Lucas 以北登陆。

OpenAI scraps rollout of new model over safety concerns

OpenAI also issued an update on incidents in which its models accessed Australian government systems.

中文摘要 OpenAI 以安全担忧为由取消一款新模型的发布,并同时就其模型访问澳大利亚政府系统的事件发布情况更新。

Will Adani get an airline?

Also in this newsletter, insurance regulator proposes fees reforms

中文摘要 英国《金融时报》简讯讨论阿达尼集团是否将获得一家航空公司,文中另提及保险监管机构提出的费用改革建议。

Anthropic warns of ‘existential risks to humanity’ in IPO prospectus

Long-awaited S1 filing reports Claude maker lost $8bn last year on $4.6bn in revenue

中文摘要 Anthropic 提交外界期待已久的 S1 上市文件,披露去年营收 46 亿美元、亏损 80 亿美元,并在文件中警告人工智能可能给人类带来「生存性风险」。

Singapore’s bet on chipmaking

City-state’s economy is getting an AI uplift but costs and labour are constraints to its industrial growth

中文摘要 新加坡押注芯片制造,人工智能为该国经济带来提振,但成本与劳动力成为其工业增长的制约因素。

Burnham to unveil public body to invest in electricity grid

The prime minister will say he wants to reduce energy costs in his first conference speech as Labour leader.

中文摘要 伯纳姆将公布一家投资电力电网的公共机构;工党领袖在其首次党会议演讲中表示希望降低能源成本。

Why Earnings Growth Remain a Key Buffer Against Yield Risks

Capital.com Senior Market Analyst Kyle Rodda argues that strong corporate earnings continue to buffer equities against rising bond yields and AI valuation risks. He also cautions that investor tolerance could fade quickly if profit growth expectations begin to fade. (Source: Bloomberg)

中文摘要 Capital.com 高级市场分析师凯尔·罗达认为,强劲的企业盈利继续为股市提供缓冲,抵御债券收益率上升与 AI 估值风险,但若利润增长预期转弱,投资者容忍度可能迅速下降。

Gold Trades Near Seven-Week Low as Rate-Hike Pressure Mounts

Gold held a sharp decline as the deadlock between the US and Iran over the Strait of Hormuz continued to keep energy costs elevated and maintained pressure on the Federal Reserve to hike interest rates.

中文摘要 金价维持大幅下跌、逼近七周低点;美伊在霍尔木兹海峡问题上的僵局持续推高能源成本,并令美联储面临加息压力。

Flashlight Capital Urges Secom to Raise Stake in Korea’s S-1

Activist investor Flashlight Capital Partners urged Japan’s Secom Co. to roughly double its interest in South Korea’s S-1 Corp. and Taiwan Secom Co., roping in a new party after its earlier bid to acquire an S-1 stake was rejected.

中文摘要 激进投资者 Flashlight Capital Partners 呼吁日本 Secom 将其在韩国 S-1 及台湾 Secom 的持股大约翻倍,此前其收购 S-1 股份的提议遭到拒绝。

Pope Leo criticises Nvidia’s Jensen Huang over AI safety

Rare intervention from head of Catholic Church comes as top executives are due to meet at White House on Tuesday

中文摘要 教皇利奥就人工智能安全批评英伟达首席执行官黄仁勋,这是天主教会领袖的罕见介入;相关企业高管将于周二在白宫会面。

SpaceX’s Starship rocket reaches orbit for the first time despite engine failure

Milestone for Musk because the system is pivotal to winning more Nasa contracts and to transport humans to space

中文摘要 SpaceX 星舰首次进入轨道,尽管发生发动机故障;该进展对马斯克是里程碑,因星舰对赢得更多 NASA 合同及实现载人航天至关重要。

Samsung to Invest $1 Billion in KKR’s AI Infrastructure Company

Samsung Group companies including Samsung Electronics Co. agreed to invest a combined $1 billion in Helix Digital Infrastructure, a KKR & Co.-backed US company, deepening the South Korean conglomerate’s push into the rapidly expanding artificial-intelligence infrastructure market.

中文摘要 三星集团旗下包括三星电子在内的公司同意合计向 KKR 支持的美国公司 Helix Digital Infrastructure 投资 10 亿美元,加深其在人工智能基础设施领域的布局。

Warning more homes will be uninsurable due to flood risk

Aviva boss Amanda Blanc said England was "for sure" building homes that might be uninsurable at some point in the future.

中文摘要 Aviva 首席执行官阿曼达·布兰克表示,英格兰「必然」正在建造未来某个时点可能无法投保的房屋,洪水风险正令更多住宅面临不可保问题。

iPad mimi8

17 回复 · Apple 节点

该源今日无内容。