2026-06-07（内容获取于 06/07 19:13）

mvanhorn/last30days-skill

Python · ★ 29,439 · 🍴 2,496 · 📈 439 stars today

AI agent skill that researches any topic across Reddit, X, YouTube, HN, Polymarket, and the web - then synthesizes a grounded summary

中文介绍 AI agent 技能，可跨 Reddit、X、YouTube、HN、Polymarket 等平台搜索任意主题，综合生成有根据的摘要。适合需要快速获取近期多源信息观点的用户，如市场调研、舆情分析。

CopilotKit/CopilotKit

TypeScript · ★ 33,513 · 🍴 4,263 · 📈 631 stars today

The Frontend Stack for Agents & Generative UI. React, Angular, Mobile, Slack, and more. Makers of the AG-UI Protocol

中文介绍 面向 Agent 与生成式 UI 的前端工具集，支持 React、Angular、Mobile、Slack 等多平台。提供 AG-UI 协议，让开发者轻松构建集成 AI 功能的交互界面，加速应用交付。

MemPalace/mempalace

Python · ★ 54,524 · 🍴 7,127 · 📈 446 stars today

The best-benchmarked open-source AI memory system. And it's free.

中文介绍 开源 AI 记忆系统，基准测试表现最佳且免费。为 AI 应用提供持久化、可检索的记忆能力，解决大模型上下文窗口有限的问题。适用于构建长期记忆的对话系统或个人助手。

danielmiessler/Personal_AI_Infrastructure

TypeScript · ★ 15,147 · 🍴 2,135 · 📈 70 stars today

Agentic AI Infrastructure for magnifying HUMAN capabilities.

中文介绍 用于放大人类能力的 AI 基础设施，属于 Agentic AI 范畴。旨在构建个人化、自主化的 AI 系统，帮助用户高效执行复杂任务，适合追求生产力提升的技术从业者。

openai/plugins

JavaScript · ★ 1,886 · 🍴 263 · 📈 213 stars today

OpenAI Plugins

中文介绍 OpenAI 官方插件仓库，扩展 ChatGPT 等 AI 模型的能力边界。允许第三方开发者编写插件，让 AI 接入实时数据、执行操作，推动人机交互的新范式。

Panniantong/Agent-Reach

Python · ★ 22,864 · 🍴 1,938 · 📈 683 stars today

Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.

中文介绍 零 API 费用的 CLI 工具，赋予 AI Agent 读取 Twitter、Reddit、YouTube、GitHub、B站、小红书等全平台内容的能力。适合数据采集、舆情分析、竞品监控等场景。

sveltejs/svelte

JavaScript · ★ 87,093 · 🍴 4,940 · 📈 25 stars today

web development for the rest of us

中文介绍 激进的前端框架，编译时将声明式组件转为高效原生 JavaScript。无需虚拟 DOM，体积小、性能高。适合追求极简开发体验和极致运行时性能的 Web 应用。

nginx/nginx

C · ★ 30,748 · 🍴 7,959 · 📈 20 stars today

The official NGINX Open Source repository.

中文介绍 高性能 HTTP 服务器与反向代理，官方开源仓库。以事件驱动架构处理高并发连接，广泛用于 Web 服务、负载均衡、API 网关。互联网基础设施核心组件。

aquasecurity/trivy

Go · ★ 36,065 · 🍴 458 · 📈 159 stars today

Find vulnerabilities, misconfigurations, secrets, SBOM in containers, Kubernetes, code repositories, clouds and more

中文介绍 全方位安全扫描器，检测容器、Kubernetes、代码仓库、云环境中的漏洞、错误配置、密钥泄露和 SBOM。集成 CI/CD 流水线，保障应用供应链安全。

golang/go

Go · ★ 134,575 · 🍴 19,087 · 📈 30 stars today

The Go programming language

中文介绍 Go 编程语言官方仓库。以简洁语法、原生并发和高性能著称，广泛用于云原生、微服务、网络工具等领域。编译快、部署简单，备受 DevOps 和系统开发者青睐。

lfnovo/open-notebook

TypeScript · ★ 26,855 · 🍴 3,057 · 📈 794 stars today

An Open Source implementation of Notebook LM with more flexibility and features

中文介绍 Notebook LM 的开源替代，提供更灵活的功能特性。支持多模态笔记、知识库化，可整合 AI 能力进行内容摘要与问答。适合研究者、知识工作者管理信息。

obra/superpowers

Shell · ★ 220,031 · 🍴 19,575 · 📈 700 stars today

An agentic skills framework & software development methodology that works.

中文介绍 Agent 技能框架与软件开发方法论，强调实用性与可操作性。提供结构化的 AI Agent 开发方法，帮助团队高效构建自主系统，提升交付质量。

santifer/career-ops

JavaScript · ★ 49,687 · 🍴 10,242 · 📈 193 stars today

AI-powered job search system built on Claude Code. 14 skill modes, Go dashboard, PDF generation, batch processing.

中文介绍 基于 Claude Code 的 AI 求职系统，内置 14 种技能模式，支持 Go 仪表盘、PDF 生成和批量处理。自动化岗位搜索、简历优化等流程，提升求职效率。

openai/whisper

Python · ★ 101,998 · 🍴 12,450 · 📈 150 stars today

Robust Speech Recognition via Large-Scale Weak Supervision

中文介绍 OpenAI 开源语音识别模型，通过大规模弱监督训练实现多语言高精度转录。支持数十种语言，适用于语音转文字、字幕生成、会议记录等场景。

vitejs/vite

TypeScript · ★ 81,270 · 🍴 8,280 · 📈 25 stars today

Next generation frontend tooling. It's fast!

中文介绍 下一代前端开发与构建工具，依赖原生 ES 模块实现闪电般冷启动与热更新。开箱即用，支持 Vue、React、Svelte 等框架。迅速成为最受欢迎的前端构建器。

microsoft/mxc

Rust · ★ 670 · 🍴 26 · 📈 64 stars today

Policy-driven, layered isolation and containment

中文介绍 微软开源的策略驱动分层隔离与容器化项目。通过精细化策略控制进程间隔离，增强系统安全性。适用于多租户、边缘计算、高安全需求场景。

PaddlePaddle/PaddleOCR

Python · ★ 81,160 · 🍴 10,671 · 📈 433 stars today

Turn any PDF or image document into structured data for your AI. A powerful, lightweight OCR toolkit that bridges the gap between images/PDFs and LLMs. Supports 100+ languages.

中文介绍 百度开源的轻量级 OCR 工具包，支持 100+ 语言，可提取 PDF 和图片中的结构化文本。为 LLM 准备数据，适合文档数字化、票据识别、多语言文本分析。

microsoft/VibeVoice

Python · ★ 48,612 · 🍴 5,400 · 📈 216 stars today

Open-Source Frontier Voice AI

中文介绍 微软开源的前沿语音 AI 项目，专注于语音交互与生成。可能集成最新语音合成与识别技术，为开发者构建语音助手、播客工具等提供基础能力。

Regret Minimization with Adaptive Opponents in Repeated Games

👍 1

06/04 08:00

In this paper, we study regret minimization in repeated games with adaptive opponents who can respond based on histories of play. The standard metric of external regret in online learning is known to fail to capture such adaptivity. To account for players' counterfactual reasoning, we introduce {\tt

中文介绍 研究重复博弈中面对自适应对手的遗憾最小化问题，指出传统外部遗憾指标未能捕捉对手的适应性，需考虑反事实推理。

Learning Geometric Representations from Videos for Spatial Intelligent Multimodal Large Language Models

👍 3

06/04 08:00

Multimodal Large Language Models (MLLMs) excel at 2D semantic understanding but lack intrinsic 3D awareness, resulting in representations that fail to maintain geometric and spatial consistency across video frames. Given the scarcity of large-scale 3D data, we present GeoVR, a novel framework that l

中文介绍 提出GeoVR，从视频中学习几何表示，赋予多模态大模型3D感知能力，以解决其缺乏空间一致性的问题。

AffordanceVLA: A Vision-Language-Action Model Empowering Action Generation through Affordance-Aware Understanding

👍 9

06/04 08:00

Vision-Language-Action (VLA) models leverage the rich world knowledge of pretrained vision-language models (VLMs) to enable instruction-following robotic manipulation. However, the structural mismatch between VLM semantic spaces and embodied control policies often hinders the learning of precise per

中文介绍 提出AffordanceVLA，一种视觉-语言-动作模型，通过功能感知理解增强机器人操作中的指令跟随能力，克服语义与控制的失配。

Code2LoRA: Hypernetwork-Generated Adapters for Code Language Models under Software Evolution

👍 65

06/04 08:00

Code language models need repository-level context to resolve imports, APIs, and project conventions. Existing methods inject this knowledge as long inputs (retrieved through RAG or dependency analysis) or through per-repository fine-tuning and LoRA -- costly at repository scale and brittle to evolv

中文介绍 提出Code2LoRA，使用超网络生成适配器，为代码语言模型在软件演化中高效注入仓库级上下文，避免传统微调的高成本。

AURA: Intent-Directed Probing for Implicit-Need Surfacing in Situated LLM Agents

👍 1

06/04 08:00

A situated query like "where is Lin Wei?" often encodes more than its literal content: the user may also want to know whether Lin Wei is free, in a good mood, or worth interrupting now. Standard tool-use agents answer the literal question and stop. AURA inserts an inference step between scene percep

中文介绍 提出AURA，通过意图导向探测，让语言模型代理在情境查询中推断用户的隐含需求（如某人是否空闲），超越字面回答。

Benchmark Everything Everywhere All at Once

👍 2

06/04 08:00

Benchmarks are fundamental for evaluating and advancing LLMs and MLLMs by providing standardized and explicit measures of performance. However, their construction is labor-intensive and hard to reuse, raising concerns about sustainability and scalability. Moreover, existing benchmarks often quickly

中文介绍 讨论基准测试构建的劳动密集和不可复用问题，提出“一次性基准测试一切”，以提升可持续性和可扩展性。

Revising Context, Shifting Simulated Stance: Auditing LLM-Based Stance Simulation in Online Discussions

👍 2

06/04 08:00

Large language models are increasingly used to simulate social media users and infer how individuals may respond to online discussions. However, it remains unclear whether these simulations reflect precise user-specific beliefs or whether they are highly sensitive to semantically independent changes

中文介绍 审计基于大模型的立场模拟对语义扰动的高度敏感性，质疑其能否准确反映用户特定信念。

ForeSci: Evaluating LLM Agents for Forward-Looking AI Research Judgment

👍 0

06/04 08:00

AI research often requires decisions before future evidence exists: which bottleneck to attack, which direction to pursue, or where a project should be positioned. We introduce ForeSci, a temporally controlled benchmark for evaluating whether LLM agents can make such forward-looking research judgeme

中文介绍 引入ForeSci，一个时间控制基准，评估大模型代理在缺乏未来证据时做出前瞻性AI研究判断的能力。

Dream.exe: Can Video Generation Models Dream Executable Robot Manipulation?

👍 15

06/04 08:00

Video generation models have made impressive strides in synthesizing visually compelling content, yet their outputs remain confined to the virtual domain. A natural question follows: how well do these models reflect the physical world when their generated videos leave the screen and enter reality? W

中文介绍 探究视频生成模型（如Dream.exe）生成的虚拟内容能否直接用于真实机器人操作，测试其物理世界反映能力。

Towards One-to-Many Temporal Grounding

👍 4

06/04 08:00

Temporal Grounding (TG) aims to localize video segments corresponding to a textual query. Prior research predominantly focuses on single-segment retrieval. Real-world scenarios, however, often require localizing multiple disjoint segments for a single query -- a setting we term One-to-Many Temporal

中文介绍 提出“一对多时间定位”任务，要求根据一个查询文本定位视频中多个不连续的片段，扩展传统单片段定位。

LLMs Can Leak Training Data But Do They Want To? A Propensity-Aware Evaluation of Memorization in LLMs

👍 7

06/04 08:00

Large language models can reproduce training data, but existing memorization evaluations mostly measure whether models can be forced to do so, rather than whether they do so under ordinary use. We introduce PropMe, a propensity-aware framework for memorization evaluation that contrasts prefix-based

中文介绍 提出PropMe，一种倾向感知框架，评估大模型在正常使用下而非被迫情况下泄露训练数据的倾向。

AdaPlanBench: Evaluating Adaptive Planning in Large Language Model Agents under World and User Constraints

👍 36

06/04 08:00

Planning for real-world problems by language models often involves both world and user constraints, which may not be fully specified upfront and are progressively disclosed through interaction. However, existing benchmarks still underexplore adaptive planning under such progressively revealed dual c

中文介绍 提出AdaPlanBench，评估大模型代理在交互中逐步揭示的世界与用户约束下的适应性规划能力。

ArcANE: Do Role-Playing Language Agents Stay in Character at the Right Time?

👍 45

06/04 08:00

Role-playing language agents (RPLAs) should play characters whose values and behavior evolve as the story progresses, not maintain a fixed persona. Existing benchmarks measure factual recall at a given chapter, not whether responses align with the character's psychological trajectory, especially in

Reinforcement Learning Elicits Contextual Learning of Unseen Language Translation

👍 23

06/04 08:00

Prior work has shown that large language models (LLMs) can translate unseen or low-resource languages by undergoing continued training or even by encoding a grammar book in their context. However, both methods typically overfit specific languages, with limited zero-shot transfer at test time. To tra

TIDE: Proactive Multi-Problem Discovery via Template-Guided Iteration

👍 38

06/03 08:00

Agents are widely deployed as assistants over documents, tools, and code. However, they typically act only on explicit user requests, which surface only the problems the user has noticed, while many other important problems coexist, hidden in plain sight, within the broader user context, with their

MAOAM: Unified Object and Material Selection with Vision-Language Models

👍 9

06/02 08:00

Selection is a core operation in interactive image editing. To be practical, a user should be able to specify and disambiguate the desired selection region through either text or click-based interactions, and the system should support selecting not only objects but also other criteria, such as mater

SEAOTTER: Sensor Embedded Autoencoding with One-Time Transcode for Efficient Reconstruction

👍 4

06/02 08:00

In robotics systems, vast amounts of visual data are easily captured at high resolution using low-cost, low-power hardware. Yet, limited bandwidth and on-device compute resources prevent full utilization when transmitted via conventional codecs like JPEG/MPEG. Newer codecs, like AV1/AVIF, improve th

RobotValues: Evaluating Household Robots When Human Values Conflict

👍 24

06/02 08:00

While household robots are often evaluated based on task completion, everyday domestic environments involve value-conflicting situations in which robots are expected to choose actions that prioritize other values than task success, such as human autonomy, efficiency, or social appropriateness. Yet,

EvoDS: Self-Evolving Autonomous Data Science Agent with Skill Learning and Context Management

👍 2

06/02 08:00

Recent progress in Large Language Model (LLM) agents has enabled promising advances in automated data science. However, existing approaches remain fundamentally limited by their static action sets and lack of principled long-horizon context management, hindering their ability to accumulate reusable

LLM Anonymization Against Agentic Re-Identification

👍 1

06/01 08:00

Agentic LLMs with web search change the threat model for text anonymization: weak contextual cues can become cross-referenceable evidence for re-identification, yet those same details also carry downstream analytic value of the text. Existing defenses either remove explicit identifiers, perturb text

Absorbing Complexity: An Interaction-Native Knowledge Harness for Financial LLM Agents

👍 2

06/01 08:00

Financial AI agents often fail for a simple reason: they make users carry the complexity. A user must repeatedly restate goals, risk preferences, portfolio context, past judgments, and shifting market assumptions, while the agent answers, retrieves, acts, and forgets. In finance, this is not just in

AdaCodec: A Predictive Visual Code for Video MLLMs

👍 4

06/01 08:00

Video is temporally redundant: adjacent frames usually share most objects, background, and layout. Yet existing video multimodal large language models (video MLLMs) usually encode each sampled frame as an independent RGB image, causing visual tokens to repeat content already present in earlier frame

SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces

👍 0

05/31 08:00

Large language models are increasingly deployed as coding agents, shifting safety from individual responses to action sequences. Existing benchmarks, however, primarily assess whether models refuse unsafe prompts, leaving impacts on stateful workspaces largely unexamined. We present SABER, a benchma

The Shape of Addition: Geometric Structures of Arithmetic in Large Language Models

👍 3

05/29 08:00

Large Language Models exhibit paradoxical fragility in fundamental arithmetic, implying a disconnect between internal computation and discrete output. By analyzing the residual stream geometry during multi-operand addition, we identify the Iso-Raw-Sum Trajectory (IRST), a geometric structure where r

MechVQA: Benchmarking and Enhancing Multimodal LLMs on Comprehensive Mechanical Drawing Understanding

👍 3

05/29 08:00

Multimodal Large Language Models (MLLMs) have demonstrated significant achievements in general visual question answering (VQA) tasks. However, they remain brittle on mechanical engineering drawings, where high annotation density and weak domain knowledge, compounded by unreliable spatial relation re

Combinatorial Synthesis: Scaling Code RLVR via Atomic Decomposition and Recombination

👍 1

05/29 08:00

Reinforcement Learning with Verifiable Rewards (RLVR) has recently emerged as the cornerstone for shaping the remarkable coding abilities of Large Language Models (LLMs). However, the scalability of RLVR is severely constrained by the scarcity of sufficiently challenging verifiable code tasks that t

Multimodal Music Recommendation System using LLMs

👍 1

05/28 08:00

Music recommendation systems typically treat songs as opaque tokens, relying on collaborative interaction histories which overlooks semantic or acoustic content. Prior work has explored LLM-augmented, multimodal, and text-enhanced approaches to sequential recommendation, and while some methods parti

Meta-Cognitive Memory Policy Optimization for Long-Horizon LLM Agents

👍 6

05/28 08:00

Memory-augmented LLM agents tackle complex long-horizon tasks by recursively summarizing interaction trajectories into compact memory. However, existing approaches typically train these memory policies using outcome-based reinforcement learning, failing to localize where intermediate memory quality

Trust Region Q Adjoint Matching

👍 3

05/26 08:00

Off-policy reinforcement learning of pretrained flow policies remains challenging due to the instability of optimization arising from the multi-step sampling process. Recently, Q-learning with Adjoint Matching (QAM) addressed this issue by reformulating into a memoryless stochastic optimal control (

Is This Edit Correct? A Multi-Dimensional Benchmark for Reasoning-Aware Image Editing

👍 1

04/16 08:00

Diffusion-based image editing has achieved strong visual fidelity under natural language instructions, yet most existing systems still operate at the level of surface instruction following, without reasoning about the implicit contextual constraints embedded in real user requests. This often leads t

Context as Topology: Why Your Agent's Memory Forgets, and How Structure Escapes It

@elpresidank · 116 粉丝 · 2.9M 阅 · 543 赞 · 35 转

06/02 19:14

Most AI agent memory is built on embeddings. And there's now a proof that this entire class of system is going to forget what you stored in it — and confidently make up things you never stored at all.

中文介绍 AI 智能体的记忆系统基于嵌入向量存在根本缺陷：这类系统注定会遗忘存储内容，并自信地编造未曾存储的信息。从拓扑学视角解释记忆失效原因，挑战当前主流架构。

SpaceX IPOs in 7 days. I Fed the S1 Doc Into Claude. Here Is What It Found Buried in 300 Pages.

@DamiDefi · 96.5K 粉丝 · 2.3M 阅 · 584 赞 · 80 转

06/04 17:31

The number that stopped me was not the $2 trillion valuation. It was $791 million. That is what SpaceX made in net income in 2024. A profitable, growing aerospace company with a genuine moat in launch

中文介绍 用 Claude 分析 SpaceX IPO 文件（S1），发现关键数字是 7.91 亿美元净利润而非 2 万亿美元估值。公司盈利且具备护城河，提供深度解读。

Range and Depth on Demand

@1salman · 363 粉丝 · 2.0M 阅 · 682 赞 · 45 转

06/01 09:45

Everyone keeps asking whether AI favors specialists or generalists. I think that is the wrong question. AI does not pick a side. It changes the tradeoff. The old world forced a choice. You could go

中文介绍 AI 并非偏袒专才或通才，而是改变了这个权衡：传统需要选择深度或广度，现在 AI 让两者兼得——按需获取范围和深度。

How To Become An AI Engineer in 2026 (Without a CS Degree)

@sairahul1 · 110.7K 粉丝 · 710.8K 阅 · 509 赞 · 97 转

06/05 16:10

How To Become An AI Engineer in 2026. Without a CS degree. Without a bootcamp. Without knowing what a transformer is today. Here's what nobody tells you: The companies hiring right now don't need

中文介绍 2026 年成为 AI 工程师的路线图：不需要 CS 学位、不需要训练营、不需要懂 Transformer。当前招聘公司不看重这些，更注重实际能力。

How to master Dynamic Workflows in Claude Code: 6 patterns and 14 steps Anthropic engineers actually

@0xCodez · 3.3K 粉丝 · 637.2K 阅 · 510 赞 · 59 转

06/03 19:00

Most Claude Code users still write their workflows by hand. They chain prompts, copy outputs, paste them into the next prompt, fix what went wrong, repeat. 9 out of 10 builders haven’t tried Dynamic

中文介绍 90% 的 Claude Code 用户仍手动编写工作流。介绍 Anthropic 工程师实际使用的 6 种动态工作流模式和 14 个步骤，提升自动化效率。

What an Enterprise Context Layer Actually Is

@prukalpa · 23.1K 粉丝 · 583.2K 阅 · 506 赞 · 80 转

06/02 02:17

A field guide to what it is, what it is not, and where it fits in your AI architecture. I have had some version of the same conversation with a CIO almost every day this year. Their team has read

中文介绍 深入解读企业上下文层的定义、不是啥以及如何在 AI 架构中定位。基于与 CIO 的日常交流，澄清常见误解并提供指导。

hacking pewdiepie's AI agent harness using an evil cocomelon website (then helping protect it)

@theonejvo · 22.1K 粉丝 · 504.3K 阅 · 861 赞 · 1 转

06/02 02:01

Over the past year, @pewdiepie, has been turning into one of the most visible champions of private, self-hosted computing, and it has been a genuine pleasure to watch. What began in late 2025 as an

中文介绍 通过恶意 Cocomelon 网站入侵 PewDiePie 的 AI 智能体框架，然后帮助加固防御。展示了自托管计算环境的安全漏洞与防护方案。

Generative UI Is the New Frontend

@Saboo_Shubham_ · 116.2K 粉丝 · 263.3K 阅 · 517 赞 · 74 转

06/04 01:12

The frontend used to be a fixed thing. Designers drew it. Engineers built it. Users got what shipped. That's over. The interfaces shipping in 2026 are drawn partly by the agent itself, in real time,

中文介绍 前端不再固定：2026 年界面由智能体实时生成。设计师和工程师的静态工作方式终结，生成式 UI 成为新范式。

Claude Code + NotebookLM + Obsidian: Research Monster That Gets Smarter Every Time You Use It

@monokern · 1.2K 粉丝 · 263.1K 阅 · 505 赞 · 72 转

05/31 19:16

Most people treat research as a manual task. You open 10 tabs. You watch videos. You read articles. You take notes somewhere. An hour later you have a pile of information you're not sure what to do

中文介绍 将 Claude Code、NotebookLM 和 Obsidian 组合成自动研究系统，每次使用都会变得更智能。替代手动打开 10 个标签页的低效方式。

How to get 100k YouTube subscribers in 3 hours (The Complete Guide)

@maubaron · 16.9K 粉丝 · 233.8K 阅 · 506 赞 · 19 转

06/06 03:22

Our YouTube channel has 125k subscribers and we've never made or uploaded a single video ourselves. This is a completely automated system. It is this very same strategy that made us the first app

中文介绍 125k 订阅者、0 个自创视频：完全自动化系统在 3 小时内获取 10 万 YouTube 订阅者。揭秘全自动策略。

Stop building Foxconn factories for your agents

@garrytan · 853.3K 粉丝 · 180.6K 阅 · 503 赞 · 43 转

06/01 22:26

In January I got back into coding and I built Garry's List. Over five hundred thousand lines of Rails and the tests to police it. I was proud of it. I shouldn't have been. The thing worth being proud

中文介绍 作者反思自己写了 50 万行 Rails 代码，认为不该骄傲。核心观点：不要为智能体搭建「富士康工厂」式复杂基础设施，简单才是关键。

Building cloud agent infrastructure: what's different, and what we learned

@intuitiveml · 6.4K 粉丝 · 171.3K 阅 · 524 赞 · 70 转

06/05 08:55

Most agent frameworks today assume a desktop. One user, one machine, one process. The agent runs while the laptop is open, writes to a local filesystem, holds API keys in environment variables, and

中文介绍 当前智能体框架多假设桌面场景（单用户、单机、单进程）。构建云端智能体基础设施面临不同挑战：持久化、权限、多用户等。分享实战经验。

A guide to /goal 🥅

@dkundel · 19.3K 粉丝 · 116.9K 阅 · 523 赞 · 40 转

06/05 05:39

We launched the goal mode (or /goal) as a way to help you have Codex drive towards a concrete outcome. When you set a goal Codex will continue to work until the goal is achieved, whether that takes

中文介绍 Codex 的 /goal 模式指南：设置一个具体目标后，Codex 会持续工作直到达成，无论需要多久或多少步骤。推动 Agent 从对话转向成果导向。

🥇Top AI Papers of the Week

@dair_ai · 124.6K 粉丝 · 84.0K 阅 · 504 赞 · 83 转

05/31 23:14

1. SkillOpt Microsoft Research treats a compact natural-language skill document as the trainable state of a frozen agent, then learns that document through rollouts, reflection, and bounded edits

中文介绍 本周 AI 论文精选：Microsoft Research 的 SkillOpt 将自然语言技能文档作为冻结智能体的可训练状态；还有其他前沿研究综述。

State of Memory in Agent Harness

@mem0ai · 17.6K 粉丝 · 82.8K 阅 · 520 赞 · 60 转

06/02 22:49

Agent harnesses are where AI software actually runs. Cursor, Devin, Claude Code, Codex: these environments handle context, orchestrate tools, coordinate agents, and increasingly, manage memory. The

中文介绍 梳理 Agent 框架（Cursor、Devin、Claude Code、Codex）中的记忆管理现状：这些环境处理上下文、编排工具、协调 Agent，并越来越注重记忆能力的实现。

A harness for every task: dynamic workflows in Claude Code

@trq212 · 263.1K 粉丝 · 75.7K 阅 · 542 赞 · 36 转

06/03 04:26

Last week, we released dynamic workflows in Claude Code. Claude can now write its own harness on the fly, custom-built for the task at hand. While the default Claude Code harness is built for coding,

中文介绍 Claude Code 新动态工作流功能：Claude 能即时为手头任务自建框架，超越默认的编码场景，适应更广泛的自动化需求。

A Functional Taxonomy of World Models

@drfeifei · 738.0K 粉丝 · 72.2K 阅 · 699 赞 · 144 转

06/04 02:57

“The world is everything that is the case.” — Ludwig Wittgenstein, Tractatus Logico-Philosophicus, 1921 The world is not made of words. In an earlier essay, we argued that spatial intelligence is AI’s

中文介绍 从功能角度对世界模型进行分类，论证空间智能是 AI 的下一个前沿。引用维特根斯坦，强调世界并非由文字构成。

How to Build a Custom Agent Harness

@sydneyrunkle · 7.5K 粉丝 · 69.5K 阅 · 511 赞 · 74 转

06/04 00:57

Building useful agents is largely about customization: connecting your agent to the right context, data, and environment(s) for the task at hand. At its core, an agent is a model calling tools in a

中文介绍 构建自定义智能体框架的教程：关键在于将 Agent 连接到正确的上下文、数据和环境。核心是模型在循环中调用工具。

some notes on getting into frontier ai labs

@itsreallyvivek · 3.6K 粉丝 · 65.8K 阅 · 521 赞 · 28 转

06/05 23:48

A few days ago I wrote that getting into a frontier AI lab mostly comes down to two things: proven research and trench engineering. The more I think about it, the less these feel like separate skills.

中文介绍 进入前沿 AI 实验室的两个关键：已验证的研究能力和工程实战能力。这两者并非分离技能，而是相辅相成。

Hermes Agent NEW Super-App and DeepSeek v4 Catches Up To Opus 4.8?

06/07 04:10

中文介绍 Hermes Agent 推出新版超级应用，DeepSeek v4 性能接近 Opus 4.8 水平。

ChatGPT And Codex Are Merging (This Changes Everything)

06/04 06:17

中文介绍 ChatGPT 与 Codex 即将合并，这将彻底改变编程和AI交互方式。

The Truth about Opus 4.8 and other Huge agent News

06/01 06:27

Browsers Are Dead. Claude Just Replaced Them.

05/28 22:56

Every NEW Claude Feature Explained

05/24 06:29

Codex: Build Your Full AI Marketing Team (Agents + Skills)

05/18 23:31

OpenAI just released Codex Mobile

05/16 02:58

Codex Mobile App Released (Complete Setup Guide)

05/15 09:35

Vibe Coding for Beginners (Full Course 2026)

05/08 03:28

Codex is The NEW Best AI Coding Tool (Here's Why)

05/03 04:35

Her · हेर — a detective for your Claude Code sessions

Hugging Face Blog · 06/07 18:13

中文介绍 Hugging Face 发布文章介绍「Her · हेर」工具，作为 Claude Code 会话的侦探，协助调试与分析。

[AINews] not much happened today

Latent Space · 06/06 12:34

a quiet day of RSI.

中文介绍 今日AI领域相对平静，主要话题涉及RSI（递归自我改进）相关动态。

How to Stop Shipping Low-Quality RL Environments (with Examples)

Latent Space · 06/06 02:49

Your broken harness is actively making the model worse. Here's what I keep seeing after years of eyeballing trajectories, and what you need to fix.

中文介绍 文章讨论如何避免发布低质量强化学习环境，指出有缺陷的测试工具会使模型表现恶化，并提供修正建议。

The Meta hack shows there’s more to AI security than Mythos

MIT Tech Review AI · 06/05 17:00

On June 5, 404 Media reported that attackers had been using Meta’s AI customer support agent to steal Instagram accounts. Their approach was simple: They asked the agent to link the accounts to email addresses that they controlled, and the agent complied. One attacker broke into the dormant Obama Wh

中文介绍 2026年6月5日，404 Media报道攻击者利用Meta AI客服助手窃取Instagram账户，通过简单请求将账户链接到攻击者控制的邮箱即可成功。事件凸显AI安全问题的复杂性。

[AINews] not much happened today

Latent Space · 06/05 14:44

a quiet day

中文介绍 今日AI领域消息较少，整体较为平静。

not much happened today

Smol AI News · 06/05 13:44

**Anthropic's Mythos/Opus cycle** sparked mixed reactions with praise for **Claude Mythos**'s one-shot workflows and concerns over **Opus 4.8** benchmark regressions. **Opus 4.7** showed strong chemistry task performance, "making Claude a chemist." **Sakana AI** launched an **RSI Lab** focusing on r

中文介绍 Anthropic的Mythos/Opus周期引发热议，Claude Mythos的一次性工作流受好评，Opus 4.8基准出现退化；Opus 4.7在化学任务上表现优秀。Sakana AI推出RSI相关项目。

How one Anthropic seller rebuilt his team's workflows with Claude Code

Claude Blog · 06/05 08:00

How one Anthropic seller rebuilt his team's workflows with Claude Code

中文介绍 Anthropic一位销售员利用Claude Code重新构建团队工作流程，提升效率。

The Claude Cowork product guide

Claude Blog · 06/05 08:00

The Claude Cowork product guide

中文介绍 Claude Blog发布Claude Cowork产品指南，介绍其功能和使用方法。

Jun 5, 2026ScienceMaking Claude a chemist

Anthropic Research · 06/05 08:00

Jun 5, 2026ScienceMaking Claude a chemist

中文介绍 Anthropic研究团队于2026年6月5日发布成果，使Claude具备化学家能力，在化学任务上表现优异。

Anthropic Oceanus leaks 🤖, ChatGPT Dreaming 💭, recursive self improvement 🚀

TLDR AI · 06/05 08:00

中文介绍 今日AI动态：Anthropic Oceanus模型泄露、ChatGPT出现拟人化Dreaming现象、递归自我改进进展持续。

Reality: The Final Eval — Lukas Petersson and Axel Backlund of Andon Labs

Latent Space · 06/05 04:39

We talk with the VendingBench authors on evaling Claudes from Haiku to Mythos, and how they build leading, and lasting, frontier evals from scratch.

中文介绍 与Andon Labs的Lukas Petersson和Axel Backlund讨论VendingBench评估方法，涵盖从Haiku到Mythos的Claude模型评测及前沿评估构建。

Nemotron 3.5 Content Safety: Customizable Multimodal Safety for Global Enterprise AI

Hugging Face Blog · 06/05 02:57

中文介绍 NVIDIA发布Nemotron 3.5内容安全模型，提供可定制的多模态安全能力，适用于全球企业AI部署。

EVA-Bench Data 2.0: 3 Domains, 121 Tools, 213 Scenarios

Hugging Face Blog · 06/04 20:24

中文介绍 ServiceNow AI发布EVA-Bench Data 2.0，覆盖3个领域、121个工具和213个场景，用于评估AI代理能力。

How Endava is redesigning software delivery around AI agents

OpenAI News · 06/04 20:00

Learn how Endava is using AI agents, ChatGPT Enterprise, and Codex to accelerate software delivery, automate workflows, and build an AI-native culture across the enterprise.

中文介绍 Endava利用AI代理、ChatGPT Enterprise和Codex加速软件交付、自动化工作流，构建AI原生企业文化。

How courts are coping with a flood of AI-generated lawsuits

MIT Tech Review AI · 06/04 18:50

Most days in her chambers, Judge Maritza Braswell, a federal magistrate judge in Colorado, sifts through stacks of documents written by people without a lawyer. Many of them can’t afford to hire a lawyer, and others have cases too weak or too small to interest one. She reads each one carefully, mind

中文介绍 美国法院面临大量由AI生成的诉讼文件涌入，法官Maritza Braswell每日需处理许多无律师代理的案件，系统应对挑战日益严峻。

阿里的 Qwen3.7-Max 比我想象中差

06/07 16:29

5 回复 · 程序员 节点

用 gpt-image-2 做设计图，用 claude opus4.8 扣了一个通宵才复原一张图的 98%

06/07 12:30

8 回复 · 程序员 节点

汇编修改 BIOS 驱动安装包

06/07 12:26

6 回复 · 程序员 节点

一个疯狂的想法：用本体论抽象一个现实

06/07 01:24

7 回复 · 程序员 节点

写 README/技术方案时被 Ai 生成的 PlantUml/Mermaid 折腾烦了，做了个 预览小工具

06/06 23:08

8 回复 · 程序员 节点

谈下 ios 端的 surge 相比 loon 不足的一些地方

06/06 22:52

36 回复 · Apple 节点

iPhone air 和 iPhone17 同价，都是 5000 元，你会选哪一个

06/06 18:16

46 回复 · Apple 节点

09 年 apple 分配的 @mac.com 邮箱，因旧 Apple ID 改绑/换区失踪了，求助如何找回

06/06 17:22

14 回复 · Apple 节点

『ArchLinux』简明安装指南(Laptop+Nvidia+Cachyos Kernel +xfs+KDE)

06/05 23:26

7 回复 · Linux 节点

Surge macOS beta 支持 Tailscale

06/05 23:13

6 回复 · Apple 节点

Scientists ejected from diabetes conference for distributing journal reprints

06/07 18:10

33 points · 14 comments

The 29th International Obfuscated C Code Contest (IOCCC) 2025 Winners

06/07 13:47

187 points · 49 comments

I design with Claude more than Figma now

06/07 13:04

175 points · 149 comments

Valve P2P networking broken for more than 2 months

06/07 11:21

181 points · 86 comments

Field of clones: How horse replicas came to dominate polo

06/07 10:46

97 points · 44 comments

How Liminalism Became the Defining Aesthetic of Our Time

06/07 10:12

77 points · 42 comments

Show HN: Oproxy – inspect and modify network traffic from the browser

06/07 10:11

50 points · 7 comments

Tokenomics: Quantifying Where Tokens Are Used in Agentic Software Engineering

06/07 09:37

110 points · 41 comments

Public Domain Image Archive

06/07 08:22

135 points · 21 comments

Sem: New primitive for code understanding – not LSPs, but entities on top of Git

06/07 04:03

125 points · 48 comments

Ntsc-rs – open-source video emulation of analog TV and VHS artifacts

06/07 03:17

351 points · 94 comments

Meta confirms 1000s of Instagram accounts were hacked by abusing its AI chatbot

06/07 02:35

609 points · 219 comments

Zeroserve: A zero-config web server you can script with eBPF

06/06 22:59

238 points · 56 comments

Motorola effectively bricked its entire line of WiFi routers without explanation

06/06 22:43

161 points · 80 comments

Moving beyond fork() + exec()

06/06 22:34

306 points · 295 comments

Nvidia is proposing a beast of a CPU system for Windows PCs

06/06 20:52

289 points · 480 comments

Pokemon Emerald Ported to WebAssembly (100k FPS)

06/06 19:12

324 points · 94 comments

Google to pay SpaceX $920M a month for compute capacity at xAI data centers

06/06 04:06

254 points · 841 comments

Harness engineering: Leveraging Codex in an agent-first world

06/06 02:20

208 points · 132 comments

Symbolica 2.0: Programmable Symbols for Python and Rust

06/06 01:13

91 points · 7 comments

Win16 Memory Management

06/05 19:16

33 points · 9 comments

Arithmetic Without Numbers – How LLMs Do Math

06/05 18:19

33 points · 10 comments

Ask HN: What was your "oh shit" moment with GenAI?

06/05 07:42

Most of us were amused when DALL-E and its peers went mainstream, and we were quick to point out the obvious flaws.Then ChatGPT hit the scene and again, many of us dismissed it as a parlor trick that would never amount to much.Using LLMs for coding initially was a only small step up from basic code

Introducing Boron Buckyballs: Theory that B80 cages can’t be made is disproved

06/05 06:58

86 points · 22 comments

Games Between Programs: The Ruliology of Competition

06/05 01:54

11 points · 0 comments

Speculative KV coding: losslessly compressing KV cache by up to ~4×

06/04 23:29

54 points · 9 comments

Enforcing the First as in BGP AS_PATHs

06/04 21:04

5 points · 0 comments

Show HN: Free animated icon library for Vue

06/04 16:07

Hi everyone!Tim here, maintainer of the lucide-motion-vue library. I build this as a way to use nice animated icons in my webapps. We were already on lucide, and found animate-ui animated icons as a great collection, unfortunately React only or made to be used with shadcn.So I ported the library to

Biohub releases a world model of protein biology

06/04 11:12

87 points · 8 comments

My Software North Star

06/04 04:16

91 points · 42 comments