ask for the shape of the answer

a guest who has thought about personal ai for years says most people underuse it because they never tell it what shape the answer should take. here's his template.

share x linkedin facebook

a small table with two chairs and a lit candle on a warm meadow, with a boxy robot holding a lantern standing at the far side of the table under soft clouds

on friday i taped an episode with a guest who has been talking about personal ai systems for years, long before it was cool. he's kept meeting recordings and notes going back to the nineties because he figured one day ai would make them useful. that day showed up.

his main claim is simple. most business people underuse these tools because they never shape the conversation. they type a question and take whatever comes back. then they spend four more turns asking for what they wanted in the first place.

his fix is to tell the model what shape the answer should take before it writes a word.

he shared his standard instruction on air. if the response is going to run over a thousand words, start with a short summary. answer three specific questions he defines. show the sources it's relying on. and end with a list of the decisions he needs to make, with a recommendation for each.

that last part is the gold. most ai answers end with information, and his template forces the answer to end with decisions.

i tried it the same afternoon. it collapsed what would have been a back and forth of four or five messages into one answer i could act on. if you take one thing from this post, steal that template.

the bigger idea behind it is what he calls a personal intelligence system. a setup where the ai knows who you are, what you're trying to achieve, what you're working on, what comes next and what you've already done. so the answers fit your life instead of being generic.

this is close to my heart, because i've been building my own version. i call it a life operating system. it holds the picture of where i am today and where i'm trying to go, and it works on closing the gap. i told him on air that's the piece i care about most. a smart assistant that knows your present is useful. one that also knows your future is a different tool altogether.

we disagreed a little on agent swarms. he's excited about groups of ai agents with different personas debating a problem, the way a room of brokers debates a stock after earnings and lands on a weighted guess. i like the idea. i pushed him to find a wedge, one specific use case nobody else is doing, because a great concept and a good domain name aren't a business yet.

he also made a prediction i agree with. most people won't build custom ai tools for themselves. that's fine. a lot of the payoff comes from learning better conversation patterns with the tools you already have.

long term, i think the gap between people who use ai well and people who don't will look like the gap between people who could type and people who couldn't. the skill is learnable, and it starts with telling the machine what you want back.

more on how i use ai in my own day is in the ai book.

so look at the last thing you asked an ai... did you tell it what a useful answer looks like?

share x linkedin facebook
the newsletter

the machine economy brief

one email when it matters: bitcoin, ai, robotics, and what founders should do about it. unsubscribe anytime.

no spam, no list-selling. your email goes to sunny, nowhere else.

this is the thesis i'm building on.

i write the machine economy brief: bitcoin x ai x robotics, where it's all going, once a week.

join the machine economy brief →