Settling into how OpenAI’s dots are best used. My dot is a replacement for regular ChatGPT. It’s a single thread that knows everything and can be my go-to for any random question. For coding, I use Codex directly, but I think the future is multiple dots that delegate to Codex.

@manton I have already found myself wishing for a better threading interface for the dot; I suppose “multiple” dots (with different roles?) is a natural extension.

For now, it’s often nice to open up the actual task the dot kicks off and interact there in a single context.

@manton not had chance to play with dots yet, but I’m imagining one per project with just that context would be ideal, but maybe just more my need for mental separation