According to CoinMeta, OpenAI is preparing to optimize the super-long conversation experience for ChatGPT and Codex, with a expected launch next week. Internal tests have shown that the average opening time for a giant chat consisting of 741 rounds and 231MB of data has decreased from 27.62 seconds to 1.66 seconds, representing a nearly 94% improvement in loading speed. This optimization mainly targets the loading of chat records, rather than the speed of model responses. During the tests, the number of conversation entries that needed to be loaded was reduced from 15,529 to 64, and the number of requests decreased from 894 to 16. The overall memory usage of the application also decreased by 41.2%. For users who have been using ChatGPT or Codex for a long time, this will significantly reduce the waiting time when opening old conversations.