what should my AI know about me?
The most useful thing your AI can know about you is not your name or your job title. It is how you think when something goes wrong.
Why Preferences Miss the Point
Most people tell their AI what they like: short answers, bullet points, a certain tone. Those things matter a little, but they are surface. What actually shapes whether an AI helps you or frustrates you is your reasoning style under pressure. If you tend to go quiet and analytical when a problem gets hard, you need an AI that gives you information and then stops talking. If you tend to spiral outward, talking through options to find the one that feels right, you need something different. Knowing your preference for font size does not tell the AI either of those things.
The Part You Probably Skip
People rarely think to tell an AI what they are actually trying to protect. Not goals in the abstract, but the specific thing you will not compromise on, even when it would be easier to let it go. A father who works for himself needs to know that the AI understands his availability has a hard floor, not a soft suggestion. A person in recovery needs the AI to know that certain framings of discipline and failure land differently for them than they do for most people. That context changes everything about what a useful answer looks like.
How You Handle Being Wrong
This one is underrated. Some people want an AI that will push back directly when they are off base. Others find that kind of correction deflating in a way that shuts down the conversation entirely, and they do better with a question that helps them see the gap themselves. Neither response is a character flaw. They are just different cognitive styles, and an AI that does not know yours will default to something that may be the opposite of what you need. Think about the last time someone changed your mind. How did they do it? That is worth writing down somewhere.
What You Can Actually Tell It Right Now
Start with three things: how you prefer to receive hard information, what you are working on that matters most to you in the next 90 days, and one thing that has consistently made you distrust advice in the past. That last one is the most useful. If you have always felt patronized by people who explain too much, say so. If you have burned out chasing others' definitions of success, that is context the AI can use to flag when a suggestion might be pulling you in that direction again. Concrete context beats personality labels almost every time.
When this runs your life, it usually traces to one underlying pattern. For this, it is most often the Mirror rhythm, the thing under the behavior.