running multiple harnesses at once is the most important skill in agent work right now... one harness plus its own subagents isn't enough here's a basic example: - from Claude Code, send plans to Codex and GPT-6 Astra for review - from Codex, writing goes to Fable... genuinely better than the GPT models - the background bulk goes to Deepseek, stupidly cheap for what it handles it takes real reps to find which harness + model combo is best at each task but once you find it, the speed and quality jump to

running multiple harnesses at once is the most important skill in agent work right now... one harness plus its own subagents isn't enough here's a basic example: - from Claude Code, send plans to Codex and GPT-6 Astra for review - from Codex, writing goes to Fable... genuinely better than the GPT models - the background bulk goes to Deepseek, stupidly cheap for what it handles it takes real reps to find which harness + model combo is best at each task but once you find it, the speed and quality jump to

Machina(@EXM7777)

同时运行多个harnesses是当前agent工作的核心技能，结合不同模型可显著提升效率和质量。

1. Claude Code适合发送计划至Codex和GPT-6 Astra进行审核 2. Fable在写作任务上优于GPT系列模型 3. Deepseek以超低成本处理后台批量任务

AI代理, 模型优化, 工程实践