智能体生态观察
智能体生态动态速递
追踪 AI 智能体（Agent）生态的模型发布、框架开源、行业落地与安全治理动态。
Recurrent Harness for Agentic Continual Learning
Language model agents increasingly operate over long interaction streams, requiring online adaptation to distribution sh…
Meta’s Muse launches on iPad just a month after its mobile debut
If there’s any doubt about how seriously Meta is taking AI, here’s a new signal: The company on Wednesday announced that…
Opera: A Verbal Critic Framework for Long-horizon Coding Agents
Long-horizon coding agents need timely corrections, yet feedback can be ineffective or even harmful when it misjudges on…
AI breakthroughs in robotics won’t change your life any time soon
The story is a collaboration between MIT Technology Review and Aventine, a non-profit research foundation that creates a…
谷歌云发布 Gemini Agent，定位“通用工作智能体”
谷歌云在 Gemini at Work 2026 发布会上推出面向企业客户的 Gemini Agent，支持 Gemini Enterprise、Workspace 及第三方服务，用户给出目标即可完成任务，还可作为团队成员担任 PM、财务分析师等角色。
Nous Research confirms it hit $1.5B valuation, launches AI agents for business users
Nous Research, the startup developing the open source Hermes Agent, has raised a $90 million Series B at a $1.5 billion …
SWE-Duel: Towards a Self-Scaling Arena for Coding Agents
Static coding benchmarks are fixed at release. Over time their tasks leak into training data, and stronger agents eventu…
Mental-Models for Multi-Agent Systems
Large foundation models have accelerated progress toward general-purpose agents that interact with humans and other agen…
Pumpire: Unified Benchmark for Metric Distance Estimation
We present Pumpire, a unified benchmark for evaluating metric point-pair distance estimation capability of both image- a…
SoftwareGym: Scaling Computer-Use Agent Training with Diverse Synthesized Software Environments
Computer-use agents (CUAs) are rapidly improving, yet synthesizing diverse and reliable software environments at scale r…
Memento 3: Model-Based Recursive Self-Improvement through Reflective Rulebooks
Learning to act in unfamiliar environments requires agents to infer how the world works and revise that understanding as…
Accurate but Not Humble: Evaluating Epistemic Humility in LLM Agents under Knowledge Conflict
When retrieved evidence contradicts an agent's prior beliefs, does it revise its answer, acknowledge uncertainty, or per…
BrickBench: Evaluating Agentic Brick Design
We propose BrickBench, a benchmark for agentic text-conditioned LEGO-set design. Given a prompt, an agent is tasked with…
微软发布 VS Code v1.141：清理“僵尸”Agent 会话等
IT之家 10 月 8 日消息，微软昨日（10 月 7 日）更新推出 Visual Studio Code 1.141，本次更新聚焦 AI 智能体（Agent）工作流管理，**重点改进会话存储清理、跨平台沙盒安全、多会话并排布局，以及多 G…
Robot data startup Mecka AI nabs $60M from Sequoia
Mecka AI, a startup that collects and analyzes human motion data to train humanoid robots and other kinds of robots, ann…
微软描绘 Autopilot 全新蓝图：让 Windows AI 从“回答问题”走向“执行任务”
IT之家 10 月 8 日消息，微软今天召开 Surface 新品发布会，现场展示 Autopilot 未来方向。根据其设想，这款智能体将让 Windows AI 进一步实现自主化执行任务。
在微软的设想下，Autopilot 不再只是根据…
微软重点介绍 MXC SDK，为 AI 智能体提供受控执行环境
IT之家 10 月 8 日消息，据科技媒体 The Verge 报道，微软今天在 Windows 与 Surface 发布会重点介绍 MXC（ Microsoft Execution Containers） SDK，将其称为 Windows…
Teen's AI-guided mountain hike ends with a helicopter rescue and a lesson in common sense
Manuel Uth
Oct 8, 2026
A 16-year-old had to be airlifted off a dangerous rock face on Crown Mountain near Vancouver afte…
办公 AI 的深水区：Copilot 三千万席位之后，微软开始卖「管理层」
纳德拉把 Agent 365、Microsoft IQ 与 Copilot 并列为微软 AI 的三大支点。但比发布会更值得拆的是商业模式的变化：Agent 365 不生产智能体，它管智能体——观察、治理、保护，按用户每月 15 美元收费。本文梳理 Agent 365 从 Ignite 2025 发布到 2026 年 GA 的时间线、Copilot 三千万付费席位的财报口径，把微软、Google、Salesforce 与钉钉飞书四条定价路线放进同一张表，并用三组 RCT 与 MIT「95% 无回报」报告的正反证据，说清办公 AI 为什么正从「卖生产力」转向「卖管理层」。
MCP 协议深度拆解：规范演进、安全攻击面与 2026 生态格局
发布不到两年，MCP 从 Anthropic 的一个开放标准长成了 AI 应用连接层的事实标准——又在 2026-07-28 的规范里把自己重构了一遍：会话没了、握手没了、sampling 和 roots 被废弃。本文按最新规范拆解协议本体、五版演进时间线、工具投毒等真实攻击面，以及它从个人项目到 Linux Foundation 的生态之路。