When using the iOS gpt app to connect to codex, when you run a task it hits rate limits when the codex app itself does not. I’m not sure if this is layering token usage or what but I definitely hit rate limiting more on the gpt app controlling the codex app remotely as opposed to just using the codex Mac app