Agentic AI is here—and that means AI systems are starting to make decisions and take action autonomously. To deliver on AI’s value, leaders need to reckon with new risks.

layout: post title: “AI 正在成为知识守门人？Google DeepMind 新论文警告：信任危机迫在眉睫” description: “当 AI 从工具变成’知识代理人’，我们该如何信任它们？Google DeepMind 提出信任框架，警告认知退化和’认识论漂移’风险” date: 2026-03-05 08:30:00 +0800 categories: [AI...

Notion引入开源模型MiniMax M2.5，作为Custom Agents的实验性功能，使用成本低于闭源模型，标志着其在国际生产力工具中的应用。

How the 6-person team at Stably ships AI testing agents faster with Vercel—moving from weeks to hours. Their shift highlights how Vercel's platform eliminates infrastructure anxiety, boosting...

AgentAlpha 团队提出了 Idea2Story，这是一种预计算框架，通过从同行评审论文中构建方法论知识图谱，将模糊的研究想法转化为结构化、可复用的模式，从而减少大语言模型的上下文限制与幻觉，同时在无需运行时重新处理文献的前提下实现高效、新颖的科学发现。亚马逊研究人员提出了 Insight...

By Chester Curme and Mason DaughertyAs the addressable task length of AI agents continues to grow, effective context management becomes critical to prevent context rot and to manage LLMs’ finite...

075 失控的移民执法如何撕裂美国 When Federal Agents Kill American Citizens

Introducing the Agent Builder Template Library: a collection of ready-to-deploy agents for common tasks, equipped with the tools you already use.

Breaking down complex tasks across specialized agents is one of the most effective approaches to building capable AI systems. In this post, we'll show you how to build multi-agent systems with Deep Agents.

在Keenotes的开发中，王福强分享了使用AI/Coding Agents的经验。他从“挑选使用”转向“全权委托”，虽然提高了效率，但失去了对细节的掌控。验收仍然至关重要，明确目标和路径能更好地指导AI，避免时间浪费。他强调关注目标而非工具，AI应服务于目标实现。