It would seem better to do as many hard, high investment (i.e. time and energy), high risk of failure but high reward things as young as possible while still ha

It would seem better to do as many hard, high investment (i.e. time and energy), high risk of failure but high reward things as young as possible while still having a societal protection blanket of "that's impressive for a high schooler". No one expects much from a high schooler so if your "exploration" (startup, project, etc.) hits a bump - you can recover quite easily and just re-join the mainstream. "Explorations" after graduation can look like unemployment to potential employer. Also with today's plethora of online education available there's unlimited amount of options for exploration. In fact going deep application domain wise before college can be a great way to motivate deeper study of theory once in college. Imagine being a high schooler today mucking around with tensor flow apis. Once you get to college all those black magic random functions (i.e. predict(), fit()) you were calling in the Google tutorials will all of sudden make sense.