OpenAI has announced that it has fixed an issue that was causing some users of its Codex AI programming agent to reach usage limits faster than usual.
OpenAI has announced that it has fixed an issue that was causing some users of its Codex AI programming agent to reach usage limits faster than usual.
Thibaut Sottio, head of development at OpenAI Codex, explained that the code writing tool was doing more “background” work than intended, which was consuming too much computing power, Business Insider reports .
Over the weekend, Codex users began to complain en masse on X (Twitter) that limits were being exhausted much faster than when performing similar tasks a week ago. In response, Sottio said that OpenAI had called an emergency team meeting on Sunday to investigate the incident, and also completely reset limits for all users.
Sottio explained that features like auto-review, which checks lines of code without human intervention, and helper “subagents” sometimes ran more often than necessary. They could fire twice or retry requests too aggressively after errors occurred.
He also added that the Codex dashboard was erroneously displaying activity for which users were not actually charged.
“All fixes are now deployed, and we have added more granular monitoring to more quickly detect background usage disruptions. We continue to closely monitor the results,” Sottio noted, adding that OpenAI has once again “completely reset” usage limits for all users.
