The transition from dictation (which was always clunky) to having an actual AI amanuensis is the key.

It’s not just about speed, it’s about having the model clean up the messiness of human speech into something actually usable.

It’s a total game changer