Currently, you can mention a ChatGPT conversation inside Codex, which is excellent for brainstorming with an instant cheap model while implementating in Codex.
However, there is a gap - as ChatGPT itself can not access Codex chats at all.
Although a complete sync and unification of ChatGPT and Codex chats would be favorable - and ‘switching’ to Codex would be possible mid-chat similar to switching a model (thus switching from a chat/instant model to the heavier code models),
A nice middle ground that also fixes the current asymmetry would be to make it possible to reference a Codex chat in ChatGPT, thus bridging the gap.
Otherwise, the current situation is like:
Architect: Here’s the blueprint.
Builder:builds a house
Architect: …
Builder: You don’t get to see the house, but you still gotta direct and guide the work.
If references of Codex inside ChatGPT is not acceptable/planned/desirable by the team at Codex/OpenAI for whatsoever reason, a light version would be to be able to fire a codex task from regular ChatGPT, like /codex, which would push a summary or response once done. Thus kinda making a ChatGPT conversation more like an orchestration chat.
Such a general change of having more syncs between ChatGPT and Codex would be kinda really great for context continuity, coherence and iteration speed, possibly even costs.