Exo 架构：AI 智能体运行时自我改进
What if AI agents could safely improve their own harness at runtime? UC Berkeley PhD candidate and senior researcher @AlexKrentsel from @GoogleResearch recently
Google Research 高级研究员、UC Berkeley 博士生 Alex Krentsel 在 Arena 分享了自改进 AI 智能体架构 Exo，它借鉴操作系统 exokernel 思路，把智能体拆成受保护的机制层（状态、历史、密钥与“guardian”）和可完全演化的策略层。在 benchmark 与开放世界 RuneScape 任务上，Exo 比 Codex 和 Pi 便宜 31-62%。
正文
What if AI agents could safely improve their own harness at runtime? UC Berkeley PhD candidate and senior researcher @AlexKrentsel from @GoogleResearch recently presented his work on self-improving AI agents at Arena HQ.
This architecture, Exo, borrows the exokernel idea from operating systems, separating an agent into a protected mechanism layer (state, history, secrets, and a "guardian") and a fully evolvable policy layer.
Alex walks through the design, results on benchmarks and open-world RuneScape tasks, where Exo was 31-62% cheaper than Codex and Pi, and more.
Big thanks to @AlexKrentsel for joining us to present his work with @istoica05, @martin_casado, @ankrgyl, Sylvia Ratnasamy, @BerkeleySky, and @netsyslab.
https://t.co/LRNbInTzSn