[jstandard]

Your cautious optimism is well-founded. Having worked in the space for a few years, I've found voice tech in the car isn't quite the promised land I was hoping it would be.

The experience is still uneven. There are times when I'm amazed the recognition and language models could pick out proper names of friends and other times when I'm red in the face shouting "I SAID 'PLEASE GET CELERY', NOT 'PLEASE SELL TO ME!'"

For the time being, we still need analog button backups to handle the cases where speech rec fails or routinely misinterprets. The long tail of error cases can actually be even more dangerous for the driver than a simple button press, since they can take your mind off the road for several seconds.