OpenAI scrambles to fix Codex as coding agent usage blows past its own capacity models
OpenAI confirmed an active investigation after Codex Pro plan users began exhausting weekly quotas in one to two days. Engineering lead Thibault Sottiaux attributed the problem to abuse-detection systems overflagging ordinary usage, triggering universal quota resets. The incident reveals that AI coding agent adoption is outpacing OpenAI's capacity models.