how much effort would it be to train a good truly multilingual speech to text model?
Also it seems like sota models have a fundamental flaw of trying to use words they have seen instead of transcribing what they hear as best as they can.
@thsottiaux Sometimes the window just gets stuck - specifically, the text input field is not responsive until you close and reopen (not quit so not fatal) - it's not very smooth.
Same with chatgpt app cmd + +/- the magnification is slow
@jxnlco Loving codex. Please:
- Pause queue (sometimes you have a queue but you need to intervene and have a back and forth with the model...)
- Diff should auto recognize when codex agent is roaming in a different worktree...
- List of branches to point the diff to does not fresh, bug?