New Playground layout makes it hard to quickly iterate on and test prompts

As a commercial user who works as a chatbot developer and has a personal project also involving the API, I always found the Playground to be my first choice for tweaking prompts before introducing any changes to API calls.

A key part of prompt engineering is being able to quickly identify, tweak and expand specific parts of existing prompts (both the System and User roles), and to be able to quickly compare these changes on the same set of examples. For some tasks, reliable prompts can get quite long, especially when they include injected data.

The layout change introduced alongside GPT-4 makes this workflow very, very difficult.

You can no longer see the System prompt and the first User prompt at once in the same space to make and test quick edits (or at least I found no way to do that), every change requires scrolling up and down the “conversation”, and when the prompts or responses are long, it’s sometimes hard to even track down where messages of a certain role begin. The Clear button doesn’t help, since it also removes User-role messages.

I know some people like the new interface, but I would love if OpenAI could add the option to switch back to the old school two-sided interface for PC users, with the System prompt in a separate area (that sort of interface for editing the System role is still being used for Assistants).