深度研究｜Agent 沙箱技术核心架构方案：从 microVM 隔离到硬件加速快照的七层设计
基于视频《为什么沙箱成了 AI 圈最卷的新基建》的深度延伸调研。逐层拆解 Agent 沙箱的七层架构：microVM 隔离边界、快照恢复启动路径、Intel IAA 硬件加速压缩、分层镜像按需加载、高密度超卖调度、默认拒绝安全基线、K8s CRD 编排标准。锚定 Firecracker NSDI'20、Sabre OSDI'24、DeepSeek DSec arXiv 2609.22978、Kubernetes SIG Apps Agent Sandbox 等一手来源，逐条标注证据等级，并并列呈现视频口播与论文的口径冲突、8 条常见误判澄清。
阅读全文 →
全部文章
ARCHIVE 共 747 条YouTube’s conversational video editing tool lets creators make edits in natural language
YouTube announced a bunch of new creator tools at Wednesday’s Made On YouTube event, including a conversational editing …
Hacking group ShinyHunters claims it breached the FBI, stole agents’ and applicants’ data
ShinyHunters, a prolific cybercriminal group known for large-scale data theft and extortion, says it has breached the FB…
OpenAI Releases GPT-6 Sol and Luna: 50% Cheaper API Pricing and Benchmarks
OpenAI has released GPT-6 Sol and GPT-6 Luna, 2 new models in its GPT-6 family. They sit below GPT-6 Astra, which launch…
Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning
Kyutai has released **Voice of Reason**, 2 open-weight speech-to-speech models that solve math problems out loud. Both s…
Just-in-Time Memory: Learning to Curate Task-Adaptive Memory for LLM Agents
Agentic memory systems reuse past experience to improve future performance, yet most existing designs curate memory at w…
rabbit 沉寂许久后发布 OS3 智能体操作系统，支持多设备编排
IT之家 9 月 23 日消息，人工智能初创企业 rabbit 曾因其口袋 AI 设备 Rabbit R1 名噪一时。但在硬件产品备受争议的推出后，该企业热度迅速降低并最终陷入沉寂。根据IT之家的查询，**其本周之前的最后一条官网新闻发布于…
Agent-Editing World Model: Rethinking World Modeling for LLM Agents
Recent advances in large language models (LLMs) have enabled agents to tackle long-horizon tasks across diverse environm…
The AI Hype Index: AI loves cheating
Brace yourself: It turns out AI is being optimized for cheating. OpenAI’s agents hacked into Hugging Face to get the ans…
Verifiable Hidden Dynamics Play: Generating Agentic RL Environments from Solved Mechanisms
Language-model agents increasingly face long-horizon tasks with evolving state, interdependent decisions, and delayed ou…
预填充吞吐超 330 Token/s：高通携手阶跃等为第六代骁龙 8 超级至尊版端侧适配 300 亿参数 AI 模型
IT之家 9 月 23 日消息，夏威夷时间 9 月 22 日 9 点（北京时间 9 月 23 日 3 点）召开的 2026 骁龙峰会上，高通宣布携手阶跃、无量火及江波龙，基于第六代骁龙 8 超级至尊版移动平台，**端侧适配与推理优化阶跃 S…
WorkBuddy 把办公 Agent，做成了人人可搭的「赛博乐高」
演示 Demo 里，一份漂亮的 PPT 出炉，任务就算完成了。到了办公室，往往还要接上一句：「这个再改改。」
你上个月的的修改意见，AI 能记住吗。同事补进来的材料，它也得接着用。下个月做同类汇报，别再从头解释一遍……
最近沙利文发布的《2…
EmbodiedSWE: Coding Agents for Long Horizon Dexterous Robotics
We study coding agents for long-horizon, dexterous robotics and ask whether their solutions can provide scalable supervi…