Against Optimizing Everything

Optimization has a way of expanding until it covers your whole life. It starts with something reasonable, like sleep or spending, and ends with a spreadsheet for friendship.

The trouble is that optimizing requires a metric, and metrics are always simplifications. This is fine when the simplification is close enough. It is a disaster when the thing you care about has no honest number attached. You end up improving the measurement and quietly degrading the thing.

You can watch this happen in real time with reading. Someone decides to read more, so they count books. Counting books rewards short books, skimmed books, and books finished out of obligation. Within a year they have read forty things and remember four. The metric went up. The reading got worse.

There is also a cost to the optimizing itself. Every domain you bring under management is a domain that now requires attention. A life with twelve tracked variables is not twelve times better than a life with two; it is a part-time job you did not apply for.

I have come to think the right approach is to optimize hard in a small number of places and to deliberately leave the rest untended. Pick the two or three things where compounding is real and the measurement is honest. Everywhere else, aim for adequate and stop looking.

The untended parts are not waste. They are where the surprises come from.