Tools & thinking
How to use AI tools well
Every model release makes half the specific advice stale within months. These ten habits have transferred across every tool I’ve used, which is why they’re the ones worth learning.
-
Pick one and go deep.
Fluency in one tool beats a shallow tour of nine. The compounding is in knowing exactly where a model is strong and where it makes things up, and that only comes from hours.
-
Context is the game.
The gap between a useless answer and a good one is almost always what you gave it, not which model you picked. Paste the real document, the real error, the real constraints.
-
Give it a role and a standard.
“You are reviewing this as the person who has to maintain it in two years” produces a different and better answer than “review this”. Tell it what good looks like.
-
Show it one good example.
One example of the output you want is worth three paragraphs describing the output you want. This is the cheapest quality improvement available and most people never does it.
-
Ask for the reasoning, not just the answer.
You cannot check a conclusion you can’t see the working for. Asking it to explain how it got there is also how you catch the confident wrong answer.
-
Iterate in the same thread.
Three rounds of “closer, but shorter and drop the second point” beats one perfect prompt. Treat it as a conversation with a fast colleague, not a search box.
-
Ask it to argue against itself.
“What’s the strongest case that this is wrong?” Models are agreeable by default, and agreeableness is the failure mode that costs you the most.
-
Never ship what you can’t check.
Use it hardest where you can verify the output -- code that runs, sums that reconcile, drafts you know the subject of. Its confidence is uncorrelated with its accuracy, which is the whole problem.
-
Keep the prompts that worked.
A text file of the ten prompts you reuse is a genuine personal asset. Most people rewrite the same instructions from scratch every week.
-
Use it where you’re weakest.
It flattens your worst skill faster than it sharpens your best one. The leverage is in the thing you avoid doing, not the thing you’re already good at.