fofr@fofrAI

2026-06-19 17:14·13天前

AI 摘要

设置一个编排器智能体，由它配置Gemini Managed Agents或modal cpu实例，在独立环境中启动子智能体。先执行深度研究任务，探索管理智能体团队的最佳实践（角色、技能、跨通信与规划），再将最佳推荐应用于各智能体。重复该循环（可逐步增加智能体数量）。之后给团队一个空代码仓库，挑战其构建产品、建立最佳实践，让智能体观察问题、提出修复建议并快速迭代，从而暴露竞争条件、完善规划方法。

How to make a team of co-ordinated AI agents：

set up your first preferred agent （it's your orchestrator）
ask it to configure Gemini Managed Agents or something like modal cpu instances to spin up sub agents in their own environment
kick off a deep research task to investigate best practices for managing a team of agents： best roles and skills for those roles， as well as managing the team and cross-communication and planning
have the agents apply best recommendations from the research
repeat this process in a loop （with increasing agent numbers if needed）

With your team：

give them an empty repo， challenge them to make something， establish best practices， have agents observe problems， suggest fixes for agent management and rapidly iterate （this fleshes out race conditions， planning approaches， and so on）

智能体教程/实践

在 X 查看原推导出 Markdown

fofr@fofrAI · X

55导出 Markdown

2026-06-19 17:14·13天前

在 X 看原推· x.com

AI 摘要

How to make a team of co-ordinated AI agents：

set up your first preferred agent （it's your orchestrator）
ask it to configure Gemini Managed Agents or something like modal cpu instances to spin up sub agents in their own environment
kick off a deep research task to investigate best practices for managing a team of agents： best roles and skills for those roles， as well as managing the team and cross-communication and planning
have the agents apply best recommendations from the research
repeat this process in a loop （with increasing agent numbers if needed）

With your team：

give them an empty repo， challenge them to make something， establish best practices， have agents observe problems， suggest fixes for agent management and rapidly iterate （this fleshes out race conditions， planning approaches， and so on）