OpenAI's official GPT-6 Astra prompting guide: five changes to stop prompting it like GPT-5
goyalshaliniuk · x · 2026-09-09
OpenAI released an official prompting guide for GPT-6 Astra, noting most people still prompt it like GPT-5. Key points: audit your AGENTS.md for conflicting stale instructions; define autonomy explicitly; keep skill descriptions to one line; stop micromanaging tests (repetition causes overtesting); define clear stop conditions. Core takeaway: give Astra ownership instead of controlling every step.
More from coding & agent
- Nous Research's Hermes Agent adds Perplexity Search API with 400B URL index — max_paperclips · 2026-09-09
- AGENTSCOPE: Microsoft & Tsinghua's neuro-symbolic method pinpoints LLM agent failure steps and types — rohanpaul_ai · 2026-09-09
- Microsoft & Tsinghua: structured run views lift GPT-5.1 agent failure localization from 3.6% to 31.4% — rohanpaul_ai · 2026-09-09
- Meta's Muse Spark 1.3 lands in Cursor, touted as cheaper than GPT 5.6 Sol — acbuller · 2026-09-09
- Tencent open-sources teamai-cli: share AI skills across team via Git review — aigclink · 2026-09-09
- Dual-model loop: using GPT Pro and Codex to cross-check math papers — jessi_cata · 2026-09-09