thoughts on using voice to prompt

like everyone else, i've been primarily using my voice to prompt ai agents. the speed between feedback and execution, as well as my ability to give the agents context has never been faster.

lately however, i've been running into this issue where the direction of where i'm going is still unclear, even after i've spent considerable time giving the agent a brain dump of contextual information.

below are some self-reflections of what i think contribute to this issue. i'm not against the use of voice, but i think we should we more intentionally deliberate with choosing it as a tool, not the default medium.

oftentimes i have these floating, disconnected nodes of ideas. when i'm prompting ai with my voice, i'm essentially outsourcing the work of connecting and understanding the relationships between them.

i may have some vague sense of how they fit together, but i don't entirely know what the map looks like yet. i'm hoping it can organize my thoughts and draw the connections i haven't quite figured out myself.

ultimately, im gambling with the fact that whatever shape it gives them ends up being what i was trying to convey. when this happens, the result i end up getting is slop... the average of the ai's training data.

another thing i've noticed is that when i'm using voice, it feels like i'm filling up a context meter. there's this assumption that the more i fill it up, the better. but what usually happens is i start introducing a bunch of unanswered questions into my rant, and those questions end up pulling the ai in directions i haven't fully thought through.

i think in general, we have to be a lot more careful in being self-aware of whether we're sub-consciously outsourcing our thinking and making sure that we are the ones driving the direction.