让 Astra 先复述你的提示词再动手，减少跑偏输出

virgilxbt @virgilxbt原推发布于 2026.09.23

在 X 上打开原推

This is f*cking insane.. Chinese researchers just showed that most bad outputs from GPT-6 Astra start in the prompt, long before the model does anything. So stop writing prompts -> Make it rewrite yours first. It restates what you literally asked, then what you actually meant. The gap between those two is where every bad output comes from. Then it stops and waits for your approval before doing any work. Send this to GPT-6 Astra before you type anything else today, full prompt below ↓

这条是引用着另一条写的：x.com/slash1sol/status/2098395296845729871