The human brain estimates the future linearly: two steps is twice one step. Yet the processes that matter most — wealth, technology, contagion, knowledge — are mostly exponential. This mismatch between intuition and reality makes us quit too early, when "nothing seems to be happening," and then leaves us blindsided by the steep blow-up at the tail. The real miracle of compounding isn't the arithmetic — it's that it lies entirely beyond the resolution of our senses.
Exponential growth is defined by this: the increment is proportional to the current stock. A growth rate acts on an ever-larger base, so the "added part" keeps growing too. A handy approximation is the Rule of 72 — divide 72 by the growth rate to get the doubling period: 7% doubles in about 10 periods, 10% in about 7. Its most counterintuitive trait: nearly all of the total change is crammed into the very last stretch. The long stretch of "nothing happening" is the engine quietly building up.
| Round | +100 each (Linear) | ×2 each (Exp., start 1) |
|---|---|---|
| Round 5 | 500 | 16 |
| Round 10 | 1,000 | 512 |
| Round 15 | 1,500 | 16,384 |
| Round 20 | 2,000 | 524,288 |
Imagine a lily pad whose area doubles every day, covering the whole lake on day 48. Question: on which day is it half covered? Day 47 — meaning that on the day the lake still looks half empty and wide open, it is only one day from total suffocation. From another angle: fold an ordinary sheet of paper 42 times and its theoretical thickness reaches the Moon. The force of both examples is that they strike at our utter blindness to how steep the tail of an exponential really is.
In biology it is the exponential reproduction of bacteria when nutrients are plentiful; in epidemiology it is the early outbreak driven by the basic reproduction number R0 — which is exactly why "cases are still few" is a dangerous illusion; in the history of technology it is the steady doubling of compute under Moore's Law; and cognitively it is the compounding of skill and knowledge — every book read and connection built becomes a higher starting point for the next round of learning.
As a technologist pursuing the "AI super-individual," your investments in tooling, a personal knowledge base, and automated workflows sit on that early plateau of the exponential curve where "no return is visible." The real risk isn't investing too little — it's quitting just before the inflection out of anxiety. Try to treat each day's small accumulation as "base," not as an isolated "increment."
For the thing you keep pursuing with no visible return yet — if it is in fact on the early stretch of an exponential curve, how much longer are you willing to persist before you draw conclusions?
Two centuries of mainstream economics rest on "diminishing marginal returns" — the more you invest, the smaller each unit's payoff, so markets tend toward equilibrium. But in a world dominated by information, networks, and knowledge, the rule often reverses: bigger means stronger, the leader pulls further ahead, and the winner takes all. Once you accept that "increasing returns" is also normal, your entire worldview — that fair competition leads to equilibrium — begins to loosen.
The engine of increasing returns is positive feedback and lock-in. Three forces drive it: the learning curve (the more you do, the more skilled and cheaper per unit), network effects (the more users, the more valuable the product to each), and economies of scale. They make a tiny initial lead self-amplify; the system no longer has a single stable endpoint but multiple possible equilibria — which one it lands on depends on small, accidental early differences. Once locked in, even a later, better alternative struggles to dislodge the incumbent arrangement.
The QWERTY keyboard we still type on came from an "anti-optimization" over a century ago, meant to slow typists down so mechanical typebars wouldn't jam. The later Dvorak layout is more efficient, yet almost no one switched — because the whole world is locked into QWERTY, and switching alone is never worth it for any individual. The technically superior loses to the first mover; this plays out again and again in history. The lesson: the winner is often not "the best" but "the first to reach critical mass."
In sociology it is the "Matthew effect" — those with reputation more easily attract new resources and citations; in ecology it is the founder effect and priority occupation, where the first species to arrive reshapes the whole niche structure; in neuroscience it is "use it or lose it," where repeatedly activated pathways grow stronger and easier to fire again; and in complex systems it is positive feedback itself, amplifying tiny fluctuations and creating path dependence.
When choosing a platform, a tech ecosystem, or the audience for a personal brand, the "increasing returns" logic means early direction often decides the long-run outcome more than effort does — which winner-take-all network are you accumulating assets in? It is also a warning: what is locked in need not be optimal. Periodically check whether you are merely trapped by path dependence while believing it was your own choice.
The ecosystem you invest in most (tech stack, platform, community) — is it on an increasing-returns track heading toward winner-take-all, or an already locked-in old order that isn't actually optimal?
No exponential growth lasts forever. In the real world, every burst of growth is ultimately S-shaped: a slow start, a rapid explosion, then a flattening as it approaches saturation. Knowing which segment of the S-curve you are on decides whether to go all-in now or start hunting for the next curve. Mistaking one segment of an S-curve for eternity is among the costliest judgment errors there is.
The S-curve arises when growth meets a boundary. Early growth is roughly exponential, but as the resource, market, or susceptible population is gradually exhausted, the growth rate falls as saturation rises, so the curve bends and flattens into an "S." In diffusion, adopters fall into categories by timing: a tiny group of innovators, then early adopters, the early majority, the late majority, and laggards. The "chasm" lying between early adopters and the early majority is exactly where most new things die — crossing it requires reaching a critical mass, after which diffusion accelerates on its own.
A disruptive new technology is, at birth, almost always "worse" than the mature incumbent — slower, pricier, less reliable — and so is dismissed alike by mainstream users and dominant incumbents. But it sits on the steep rise of its own S-curve, improving far faster than the near-saturated old technology, and at some crossover point the seemingly solid leader is suddenly overtaken. Kodak held early digital-camera patents yet, mired in the topped-out film curve, missed the pivot — the leader's death is often not from doing things wrong, but from mistaking a topped-out curve for eternity.
In epidemiology, the S-curve is the cumulative infection curve, flattening as the susceptible pool runs out; in ecology it is logistic growth, a population approaching carrying capacity; in physics it echoes the rapid switch of system states across a phase transition; and in linguistics, the spread of a new usage follows the same "slow–fast–saturate" diffusion path.
When assessing an AI technology, a product, or one of your own skills, ask first: which segment of the S-curve is it on? The rising segment deserves full commitment; near saturation, you should pre-position the "second curve." Stalled growth is often not a sign of insufficient effort but of the current curve topping out — where redoubling effort yields ever-thinner returns.
The core capability or business you stand on — is it on the rising or the saturated segment of its S-curve? If the latter, where is your "second curve," and how much have you invested in it?
Economics 101 teaches the law of demand: the higher the price, the lower the demand. But one class of goods does the opposite — the more expensive, the more people want it, because for them the high price is the core value being bought. This tears open a truth: much consumption isn't aimed at "using" something, but at "being seen by others to own" it.
The utility of such goods comes largely from scarcity and from being "expensive enough to be visible at a glance." Here price is not a cost but a signal — it announces the buyer's wealth and status to the world. So as the price rises, the signal grows stronger, and demand rises rather than falls; over this stretch the demand curve slopes upward, defying the norm. This differs from a "Giffen good": the latter is an inferior staple the poor are forced to buy more of when its price rises, while the former is a luxury symbol the affluent use to display.
What luxury brands fear most is a deep price cut — once discounted, sales often fall rather than rise, because it dissolves the scarcity signal carried by "expensive" and scares off the real buyers. More intriguing still is a neuroeconomics experiment: label the same wine with a higher price, and tasters not only judge it more delicious but, on fMRI, show stronger activation in pleasure-related brain regions — price genuinely changes the taste a person experiences.
In evolutionary biology it maps to the "costly signal" — the peacock's burdensome tail, precisely because affording it proves good genes, becomes a credible mating signal; in sociology it is class distinction and identity construction through consumption taste; and in psychology it shares a root with the placebo effect — price shapes expectation, and expectation in turn rewrites the actual subjective experience.
Whether pricing a product, building a personal brand, or designing signals in hiring, recognize that "expensive" and "hard to get" themselves transmit information. Conversely, as a consumer and investor, be wary of whether you are paying for a pure "price signal" or "narrative premium" rather than for real value.
Recall your most recent "the pricier, the more I wanted it" purchase — were you buying the thing's use, or the fact that others see you owning it?