good instinct. the useful bit is not “be a generalist” as a personality brand.

it starts with what kind of world you’re learning in.

some environments are kind. some are wicked.

kind means stable rules, repeatable patterns, and fast accurate feedback.

chess is the classic example. golf too. free throws, scales, stuff where reps map cleanly to improvement.

yes. very good.

not everywhere. that’s the part people flatten.

if the environment is kind, drilling the real pattern works because feedback is honest.

hidden rules, changing rules, delayed feedback, noisy feedback, or feedback that teaches you the wrong lesson.

startups, hiring, markets, policy, forecasting, creative work, relationships. medicine can be wicked too, depending on the problem.

yeah. you make a decision, it works once, and your brain goes “i am a genius.” maybe timing was lucky.

exactly. in wicked worlds, expertise can overfit. familiar patterns start feeling true just because they’re familiar.

yep. or hiring someone because they remind you of the last person who worked out, even though the role changed.

not trivia-hoarding. range gives you more pattern libraries.

you can borrow analogies, notice when this problem is not the same as the last one, and avoid locking onto the first familiar story.

exactly. one detailed map is amazing in a kind world. in a wicked world, one map can get you lost with confidence.

classify the environment first.

if rules are stable and feedback is quick and accurate, specialize, drill, tighten the loop.

if rules are hidden or changing, vary practice, seek outside analogies, slow down confidence, update your model often.

treat early wins like clues, not proof.

ask: did this work because my model is good, or because the world was friendly this time?

that’s most good advice, unfortunately.

pretty good. kind domains want reps. wicked domains want sampling, comparison, and revisability.

yep. that’s the useful version.