Could you bring back the old speech-to-text behavior where the text appeared in the box instead of auto-submitting? It let us edit or add more before sending, which was super helpful to build layered messages before submitting. The timer was a great touch too. Also, the new version sometimes fails to submit due to a ‘network error,’ and unlike before, there’s no retry option.. So everything just gets lost, especially after speaking for a minute or more.