GPT-6 prompt cache survives reasoning-effort changes; Codex adapts effort mid-CoT
daniel_mac8 · x · 2026-09-22
danielmac8 confirms a practical detail: for GPT-6 family models, changing the reasoning effort setting no longer invalidates the prompt cache (OpenAI docs linked), making it far cheaper to switch effort levels mid-session.
He also praises Codex's adaptive reasoning as "utter genius": powered by Jev from miu21590, it adjusts reasoning effort mid-CoT based on task difficulty instead of locking in one level per request.
Related event: Codex Gains Adaptive Reasoning Mid-Chain-of-Thought(2 posts)→
More from coding & agent
- GitHub Copilot app adds Sentry canvas: from crash report to fix PR in one app — mariorod1 · 2026-09-22
- Giving an autonomous agent homeostatic sleep: fatigue from real token cost, plus REM dreaming — Dzikula · 2026-09-22
- Dev on AI coding's biggest pain point: models are still too stupid, slow and expensive — remilouf · 2026-09-22
- MCP Is Speedrunning the Web 2.0 Story: Platforms Will Tighten Integration Gates — dbreunig · 2026-09-22
- Firebase Tutorial: Build Apps That Talk, Laugh and Whisper with Gemini TTS — jggomezt · 2026-09-22
- Debate: RL Environment Synthesis Is Valuable but Not Novel — stochasticchasm · 2026-09-22