[OpenAI Codex Product Lead: Multiple Astra Quality Issues Resolved, Reset Scheduled Before Midnight] According to a report by Jinse Finance, on September 12, OpenAI Codex Product Lead Tibo released an update regarding recent quality issues reported by Astra users. Tibo stated that the team has identified and resolved several defects and will perform a reset before midnight local time (12:00 PM Beijing time). Tibo mentioned that after collaborating with some users, the team confirmed and addressed the following issues: 1) Abnormal triggering of skills in older models: Certain skills written for previous models were triggered too frequently or caused the model to fail in checking its own work. 2) Context management experiment disabled: An optional context management experiment could lead to premature halts or responses to outdated messages. This experiment has now been disabled. It is estimated that approximately 4,000 to 5,000 users were affected by this experiment. 3) Removal of improperly configured engines: Some engines were improperly configured, resulting in measurable quality degradation in tail-end traffic flowing through them. These engines have now been removed. Additionally, the team has implemented several minor improvements, stating that the overall experience should see significant enhancements, including more consistent execution, more accurate tracking of users' latest messages, and better verification of results during task progression. (Jin10)