This was on the first release of ChatGPT so I guess GPT-3.5. Pretty much like WASDx describes. In my case it was even more meta because I was writing a script t

This was on the first release of ChatGPT so I guess GPT-3.5. Pretty much like WASDx describes. In my case it was even more meta because I was writing a script that was making ChatGPT API calls. It’s an I/O network call so it was fairly easy to rewrite as a multithreaded generator loop, which it got right on the first go. Nice speedup of about 10X, I imagine right about up to the API rate limit.