jevVIDEO

Adding a Jev dynamic reasoning throttle to GPT-6 in Codex

Jev adjusts reasoning depth at millisecond scale along the execution path, thinking less when things flow and more when stuck, with tests showing about 50% lower cost and no loss of existing cache.

AYi

@AYi_AInotesArchived 2026.09.23Originally posted 2026.09.22

AYi @AYi_AInotesOriginally posted 2026.09.22

Open original on X

兄弟们，这是我近期见过最实用的 Jev 用例之一，给 Codex 里的 GPT-6 装上动态油门，顺畅时少费脑，卡住时才拼命思考，老哥真是个天才哈哈，实测不仅运行更快、没丢掉原有缓存，连整体成本都直接打了对折（-50%），比起在任务前盲目挑模型，这种在执行链路上毫秒级自适应调控推理深度的做法，才是接下来降本提速的正确打开方式啊兄弟们对于每天重度跑 Agent 或长任务的开发者来说，这种把算力动态按需分配的工程解法，才是真正把账单打下来的硬核思路 https://x.com/miu21590/status/2101857866378362926/video/1

This post quotes another post: x.com/miu21590/status/2101857866378362926

x.com/ayi_ainotes/status/2102358409530192181Archived 2026.09.23 10:17