Event ·Multi-source

为 GPT-6 Astra 重新设计 Skills 和提示词

旧模型辛辛苦苦积累的指令，正在成为新模型的负担。GPT-6 Astra 能力提升后，过去用来“补偿模型弱点”的提示词约束过度约束了它；从“教它怎么做”转向“告诉它边界在哪、完成标准是什么”，把方法论的自主权还给模型。

Current read

OpenAI 建议开发者为 GPT-6 Astra 重新设计 Skills 和提示词，别再“硬教”模型做事，要告诉它边界和完成标准。

11 sources78° AI heatLast updated 2026/9/12 00:51:59

Evidence

11 sources

Primary sourceshao__meng

2026/9/12 00:51:59View source

Related source 1小互

2026/9/14 00:32:03View source

Related source 2Decoder

2026/9/12 13:10:11View source

Related source 3@OpenAIDevs

2026/9/11 18:34:05View source

Related source 4Geek

2026/9/10 07:51:57View source

Related source 5IT之家

2026/9/10 09:11:03View source

Related source 6lmarena.ai

2026/9/10 15:43:04View source

Related source 7向阳乔木

2026/9/10 18:28:41View source

Related source 8OpenAI

2026/9/10 18:35:32View source

Related source 9Greg Brockman

2026/9/10 21:12:15View source

Related source 10AI Will

2026/9/11 01:32:24View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story