voice ui is deceptive

the voice computer use demos are interesting but there is something weird about them. they feel good for like basic stuff (open app, play music or something which some pre llm things did already) but try it out with some finer things.

like imagine someone reciting "open the doc, scroll to the third paragraph, bold the heading".. i mean they'll have to perform a script. which is a strange af. the entire appeal of a gui was that pointing collapsed the distance between intent & action. ironically voice reexpands it cuz you have to serialize a spatial, parallel, & glanceable task into a linear utterance, which is strictly more cognitive load for most direct manipulation work.

it's the difference between grabbing a cup & describing the cup to someone who then grabs it. we are early but very fun to see how far we can push these boundaries.