On September 12, in response to recent quality issues reported by Astra users, Tibo, the product lead for OpenAI Codex, announced that the team has identified and fixed multiple defects, with plans to perform a system reset before midnight local time (12:00 PM Beijing Time) today. Tibo stated that after collaborating with some users, the team confirmed and resolved the following issues: First, abnormal triggering of skills from older models, where skills written for previous models were being triggered too frequently, affecting the model's self-checking capabilities; Second, the context management experiment was disabled, as it could cause responses to stop prematurely or repeat old messages, impacting approximately 4,000 to 5,000 users; Third, improperly configured engines were removed, as they were degrading traffic quality. Additionally, the team made minor improvements, such as more consistently and accurately tracking users' latest messages and more thoroughly checking results during work progress, which will significantly enhance the overall user experience.