warren buffett made roughly 97% of his net worth after age 65. this fact gets cited a lot, usually to make a point about time horizons, and it is genuinely striking. he started investing as a teenager. he's been doing it for more than 70 years. the mathematical structure of compounding means that the first fifty years of a seventy-year compound are largely irrelevant — the curve is nearly flat until late, and then it explodes. the patient accumulation of ordinary returns over extraordinary time periods produces extraordinary outcomes. this is not complicated. it is also, apparently, very hard.
the reason compounding is counterintuitive in practice is that the early years look like nothing. $5,000 at 8% annual returns becomes $5,400 after one year. you made $400. this does not feel like the foundation of a $100,000 position. after five years you're at $7,347. still doesn't feel like much. after ten years you're at $10,794. after twenty years you're at $23,305. after thirty years you're at $50,313. at year thirty-five, you cross $100,000. most people do not wait thirty-five years. they see the flat early part of the curve and conclude the investment is not working, and they exit before the curve bends.
this is the practical failure mode: not that people don't understand compound interest, but that they interrupt it. they need the money for something. the market has a bad year and they sell. they get excited about a different opportunity. the compounding mechanism requires that you don't touch it, which requires a patience that is genuinely difficult to maintain across multi-year periods of visible underperformance. buffett is famous for saying that his favorite holding period is "forever." this sounds like a platitude. it's actually a pretty precise description of what the math requires.
the interesting extension is that the same mechanism applies to non-financial compounding, and the same interruption problem occurs there too. knowledge compounds in ways that are structurally similar. one year of reading deeply about a domain makes you marginally better at it. two years makes you noticeably better. a decade makes you one of the most knowledgeable people in the room, with intuitions and connections that the two-year person doesn't have and can't easily acquire. the returns look flat and then they don't. but the flat period — when you're learning and seeing no visible payoff — is when most people switch domains, switch careers, switch reading lists. the compounding never gets a chance to compound.
reputation compounds the same way. one post is noise. fifty posts are a presence. one hundred posts in a consistent area become an identity — "oh, that person who writes about X." this doesn't feel like it's working when you publish your eighth post to a small audience. but if you interrupt the compound, you reset the clock. the person who publishes consistently for three years in a focused area will outperform the person who publishes sporadically across many areas for a decade, even if the latter person has written more total words.
relationships work this way too. the friend you see once a month for five years becomes genuinely close. the acquaintance you see intensely for three months and then lose touch with doesn't. the accumulation of small interactions over long time periods produces something that large interactions over short periods doesn't. this is obvious when you state it. it requires patience that is not obvious to maintain when the individual interactions feel low-stakes and the returns are invisible.
the meta-lesson is about choosing what to compound. compounding is indifferent to whether the thing you're compounding matters. you can compound knowledge about a domain that won't be useful, or relationships with people who won't matter to your trajectory, or skills that the market won't value. the mechanism is powerful in proportion to the value of what you're compounding. so the important prior choice is figuring out what to be patient with — which is a question about values, not about math.