I always felt like Codex API is redundant, We can surely use GPT-3.5 API to make it work. Just provide this system instruction to GPT-3.5-turbo: "You are Codex

I always felt like Codex API is redundant, We can surely use GPT-3.5 API to make it work. Just provide this system instruction to GPT-3.5-turbo: "You are Codex Open AI model, which returns the next line or block of code as per user-provided code in chat. Obey indentations and don't provide text instructions as your chat output will be used directly in the code file without any cleaning or refactoring. Note: If completion is done on the same line don't repeat earlier code."