How to Use AI Well
Read AI's next move while keeping your own moves open. Using AI well means expanding your options, not becoming dependent on one answer.
How to Use AI Well
People ask how to use AI well.
They say you need good prompts.
Enough context.
A role, examples, and a specified output format.
All true, of course.
But after using AI for a long time, I reached a slightly different conclusion.
Someone who uses AI well is not someone who gets good answers from it.
It is someone who reads AI’s next move while keeping their own next moves open.
Go and chess have a similar essence.
Reduce the opponent’s moves and increase your own.
Your opponent may appear to have twenty options, but if nineteen are bad, they effectively have one. Conversely, even if you have only five options, you are free if all five are good.
A strong player is not the one who captures the most pieces.
It is the one who narrows the opponent’s future while expanding their own.
AI is the same.
After using it a few times, you begin to see response patterns.
How it reacts to particular questions,
where it becomes conservative,
what it overlooks,
and what context changes the direction of its thinking become predictable.
Beginners adapt themselves to AI at this point.
They ask questions the way AI prefers, reshaping their thoughts into forms it understands easily.
At a higher level, however, the reverse begins.
I anticipate AI’s responses without making myself too easy for AI to predict.
I do not wait for the first model’s answer to decide my second question.
I think two or three moves ahead.
If I ask this, AI will answer that way.
Then I can extract another piece of information here.
Next, I change perspective and examine the same problem again.
I give different models the same problem in different ways.
When needed, I deliberately set up an opposing hypothesis.
AI becomes a player, rather than merely an answerer.
The user designs the game around it.
So the most important ability for using AI well may not be prompt writing.
It may be prediction and the management of options.
You need to anticipate AI’s next response.
At the same time, you need multiple paths so you do not become dependent on one answer.
A good AI user does not become dependent on AI.
Using AI increases their options instead.
They can investigate more broadly,
test more hypotheses,
build more products,
and execute faster.
If using AI reduces your options, you are using it poorly.
If you cannot judge anything without AI,
simply select one of its proposed answers,
and cannot conceive of directions it did not consider,
you have entered the tool’s policy rather than used the tool.
The important question in the AI era
is not “How do I get better answers from AI?”
That is not the point.
The larger question is:
“How can I use AI to increase my possible moves?”
On a chessboard, the side with greater freedom is stronger.
The same holds for companies, people, and the way we use AI.
Using AI well is surprisingly simple.
Read AI’s moves.
But keep as many of your own moves open as possible.