What do you optimize for?

Draft note. I am putting the idea here before it is fully formed, so I can come back and improve it later.

When we train machine learning models, we usually define something like a loss function. We tell the system what to reduce, what to reward, and what kind of error matters. Then we let it optimize.

We may not know every detail of what happens inside the model. We may not understand every representation it builds or every internal shortcut it discovers. But we do know one thing: over time, it will become good at what we repeatedly optimize it for.

That feels obvious in machine learning. But I think it is also true for people.

If someone optimizes for looking smart, they slowly become good at looking smart. If they optimize for being liked, they become sensitive to approval. If they optimize for money, they start noticing opportunities, leverage, and incentives everywhere. If they optimize for learning, they become more willing to be confused, corrected, and occasionally embarrassed.

None of these are purely good or bad. The dangerous part is when the optimization function is unconscious. You think you are just making small daily choices, but those choices are training you.

The meetings you say yes to, the people you try to impress, the numbers you refresh, the discomfort you avoid, the work you repeat, the feedback you listen to, and the feedback you ignore all become part of the training loop.

After enough repetitions, you do not just get the outcome. You become the kind of person who is optimized for that outcome.

So maybe one of the most important questions is not only, "What do I want?" It is also, "What am I currently becoming good at?"