@YashHustle_22 Beautiful design…or even aesthetically pleasing design. This is one area where Claude is ahead. Integration of something like Claude Design but openAI style would be nice with ability to tweak and play with design in a design-style interface UI
@YashHustle_22 It would be nice to see it not forget what it was doing when I steer. I need to manually create a subagent to ensure that jobs are actually finished. It should do this automatically, maybe making a list of what it’s working on and validating actual completion
@thsottiaux can we get the ability to call GPT Pro from within Codex as a subagents of sorts? Currently especially with reasoning tasks will take stuff manually out of codex and have Pro think on it or process it and then send the output back to codex.
@kr0der For Claude I literally have to have a subagent that it has to check everythitn. The task for that agent is to make sure that the task was followed without shortcuts or deception. It catches a lot
@thsottiaux From remote in iPhone app it didn’t show pending permissions regularly. So when it seems to have stopped working I have to force quit and reconnect to see permission dialog.
@sandislonjsak I actually interact with both and regularly will someone have a 1M token Claude code task as a codex subagent. They play nicely with communication