The Lie We Were Told About Effort
Most of us absorbed a simple model of the world before we were ten. Work hard, get a proportional reward. Study twice as long, learn twice as much. Put in the hours and the payoff follows in a straight line.
Paul Graham published "Superlinear Returns" in October 2023, and it starts by taking that model apart. "'You get out,' I heard a thousand times, 'what you put in.' They meant well," he writes, "but this is rarely true." The people who told him that weren't lying on purpose. They were describing the world of chores and hourly jobs, where an hour of effort does buy roughly an hour of pay. That world is real. It's just not where the biggest outcomes happen.
In most things that matter, the curve bends. Performance that's a little bit better than average earns a lot more than a little bit more. Performance that's a lot better than average earns amounts that look absurd from the bottom of the curve. Graham calls this a superlinear return: output that grows faster than the input you feed it.
The reason this matters isn't philosophical. If you believe returns are linear when they're actually superlinear, you'll make bad decisions your whole life. You'll spread effort evenly instead of concentrating it. You'll optimize for looking busy instead of being exceptional. You'll quit right before the curve starts to bend, because the early part of a superlinear path feels flat and thankless. Understanding the shape of the curve changes what you choose to do with your time.
The Two Engines Behind Outsized Results
The essay's most useful move is reducing a messy phenomenon to something you can actually reason about. Superlinear returns show up in sports, politics, art, music, science, startups, and investing. On the surface those look unrelated. Graham argues they aren't.
"It may seem as if there are a lot of different situations with superlinear returns," he writes, "but as far as I can tell they reduce to two fundamental causes: exponential growth and thresholds."
Exponential growth is the virtuous cycle. Doing well in one round makes the next round easier, so success compounds on itself. A growing audience recruits more audience. A brand that people trust earns more trust. Knowledge you already have makes new knowledge faster to acquire.
Thresholds are step functions. Cross a line and the payoff jumps, while everyone who fell short of it gets little or nothing. The team that scores one more goal takes the whole trophy. The proof that finally closes gets the credit. The product that's marginally better wins the customer, and the runner-up wins nothing.
These two engines often run together, which is what makes the biggest outcomes so extreme. Crossing a threshold can kick off exponential growth, and exponential growth can carry you across the next threshold.
| Engine | How it works | Real example | What it rewards |
|---|---|---|---|
| Exponential growth | Success in one cycle makes the next cycle easier, so gains compound | An audience that grows because existing fans bring new ones | Consistency and staying in the game long enough to compound |
| Thresholds | Crossing a performance line unlocks a jump; missing it pays almost nothing | The startup whose product is just good enough to win the customer | Being the best at a specific thing, not merely good |
Once you see the world through these two lenses, a lot of career advice starts to look naive. "Work a little harder than the next person" assumes a linear curve. On a superlinear curve, a little harder is often the difference between everything and nothing.
Exponential Growth: When Winning Compounds
The clearest modern illustration of exponential growth is the technology most people now use every day. ChatGPT launched at the end of November 2022. It reached one million users in five days and roughly 100 million monthly users within two months, which made it the fastest-growing consumer application on record at the time. UBS analysts, trying to find a comparison, wrote that in twenty years of following the internet space they couldn't recall a faster ramp in a consumer app.
Put that against earlier products and the compounding is obvious.
| Product | Approximate time to 100 million users |
|---|---|
| ChatGPT | About 2 months |
| TikTok | About 9 months |
| About 2.5 years |
Each of those was itself considered a fast grower in its day. The curve keeps getting steeper because each generation of product compounds on the infrastructure, habits, and networks the previous one built.
The same compounding shows up on the business side. NVIDIA spent decades building expertise in parallel computing that most of the market ignored. When demand for AI training exploded, that accumulated advantage paid off all at once. NVIDIA became the first chipmaker to reach a one trillion dollar market value, powered by the AI wave, in May 2023, and crossed four trillion dollars in July 2025, the first company ever to do so. The company didn't suddenly get good in 2023. It had compounded quietly for years and then crossed a threshold where the compounding became visible.
Graham's point is that compounding feels slow and then feels sudden. The early phase looks like failure. Bacteria, to use his example, grow exponentially, but a barely visible smear and a full plate are the same process a few doublings apart. Most people abandon compounding projects during the flat-looking early phase, which is exactly when persistence is worth the most. This is why work that teaches you, or builds an asset that grows on its own, beats work that merely pays by the hour. For a deeper look at how this applies to knowledge specifically, see intellectual compound interest.
Thresholds: Why So Many Fields Are Winner-Take-All
Thresholds explain the fields where the gap between first and second place is savage. In a footrace, the runner who wins by a hundredth of a second gets the medal and the endorsements. The one who loses by that same margin gets a handshake. The inputs were nearly identical. The outputs weren't close.
Graham's practical takeaway from thresholds is one of the most quoted lines in the essay: "At the far end of the curve, incremental effort is a bargain." At the top of a steep curve, a small improvement in performance produces a large jump in reward, and almost nobody is competing up there because most people gave up lower down. The far end is underpriced. You get a disproportionate return for the marginal work precisely because so few are willing to do it.
Thresholds also explain fame and audience, where the two engines combine. Recruiting fans is exponential, because fans bring more fans, but the number of "A-list" slots in any field is limited, so there's a threshold you have to cross to matter at all. Kevin Kelly's famous counter-argument is that you don't actually need to cross into mass fame. You need 1,000 true fans, a smaller threshold that's reachable by ordinary creators and still compounds. Both ideas are threshold arguments. They just disagree about where the meaningful line sits.
The lesson isn't that thresholds are fair. They aren't. The lesson is strategic: figure out where the real threshold in your field is, aim to clear it rather than to be generically "good," and remember that the space just past it is less crowded than it looks.
Do Things That Don't Scale: Crossing the First Threshold
If superlinear returns wait on the far side of a threshold, the obvious question is how you cross the first one when you have nothing yet. Graham's answer connects this essay to his earlier and more famous one. You do things that don't scale.
The canonical case is Airbnb. In 2008 the founders were nearly out of money, so they designed novelty cereal boxes, "Obama O's" and "Cap'n McCain's," hand-folded and glued them in their apartment, sold them as election collectors' items for around forty dollars a box, and raised roughly thirty thousand dollars to keep the company alive. That bought them time. Then they noticed their New York listings weren't converting, so the founders flew to New York, rented an expensive camera, and went door to door photographing hosts' apartments themselves. Those listings started getting two to three times as many bookings, and the company's weekly revenue in New York roughly doubled. None of that scaled. All of it got them across the threshold where the network could start compounding on its own.
Stripe did a version of the same thing. When someone agreed to try the product, the Collison brothers wouldn't email a signup link and hope. They'd say, in effect, hand me your laptop, and install it on the spot. Y Combinator named the move the "Collison installation." It was manual, unscalable, and completely counter to how software is "supposed" to be sold, and it's part of how Stripe out-executed better-funded competitors.
The pattern is the same every time. The compounding phase you're aiming for is powerful but automatic. The threshold that unlocks it is small, manual, and only crossable by hand. We break the mechanics of this down further in Do Things That Don't Scale, Graham's companion essay.
Learning: The Superlinear Return Anyone Can Claim
Startups and market caps make for dramatic examples, but they're not the version of superlinear returns most people can act on tomorrow. The one that is, and the one Graham keeps returning to, is learning.
Learning compounds for a plain reason. "When you first start learning something, you feel lost," Graham writes. "But it's worth making the initial effort to get a toehold, because the more you learn, the easier it will get." Every concept you understand becomes a hook that new concepts can attach to. A reader with ten years of accumulated context can absorb a dense paper in an afternoon that would take a beginner a month. The knowledge you have makes the knowledge you're acquiring cheaper. That's the definition of a superlinear return, and unlike fame or market timing, nobody has to grant you access to it.
The catch is the same as every compounding curve: the early phase feels flat, and most of what you read leaks away. Highlighting a sentence and never seeing it again does almost nothing. The compounding only starts when your reading builds a durable, connected base you actually return to. That's the entire premise behind Glasp's web highlighter. When you highlight across articles, PDFs, and books, the highlights become a searchable personal library instead of ink you forget. You can layer Glasp's AI chat on top to ask questions across everything you've saved, which is compounding made literal: your past reading answers your present questions.
Video is the same story. A one-hour talk holds three usable ideas that vanish unless you capture them. YouTube Summary pulls the transcript, timestamps, and key points so an hour of watching turns into notes you keep. And because Glasp highlights are public by default, you also tap the community of other readers, which is Graham's exponential audience engine pointed at learning instead of fame. For readers who want the science of why capture beats passive consumption, spaced repetition for readers covers the retention side in depth.
A Playbook for Positioning Yourself
Understanding superlinear returns is worthless if you don't change what you do. Graham's advice is scattered through the essay, but it collapses into a few concrete moves.
Pick a field that has superlinear returns at all. Some work is inherently linear, and no amount of brilliance changes that. Look for fields where a few big winners visibly outperform everyone else. That gap is the signature of a superlinear curve.
Follow curiosity to choose within it. Graham is blunt about the compass: "When in doubt, follow your curiosity. It never lies, and it knows more than you do about what's worth paying attention to." Curiosity isn't a soft nicety. On a superlinear curve you have to work at the far end to see real returns, and you only reach the far end on things you'd obsess over anyway. This is the same engine behind How to Do Great Work.
Take multiple shots, and start early. Because thresholds are unforgiving, any single attempt has a real chance of missing. "The solution is to take multiple shots," Graham writes. "Which is another reason to start taking risks early." Youth is an asset here not because young people are smarter but because they can afford more attempts before the stakes rise.
Always be learning. His flat rule: "Always be learning. If you're not learning, you're probably not on a path that leads to superlinear returns." A role that stops teaching you has quietly become linear, no matter what it pays.
Here's the contrast in one view.
| Situation | Return shape | The right move |
|---|---|---|
| Hourly or quota work | Linear | Trade it for something that teaches or compounds |
| Building an audience or asset | Exponential | Stay in the game through the flat early phase |
| Winner-take-all competition | Threshold | Aim to clear the line, not to be generically good |
| Deep expertise over years | Both combined | Concentrate, follow curiosity, take repeated shots |
For founders specifically, the same logic shapes what to build and how to find it, which is the subject of How to Get Startup Ideas.
The Uncomfortable Truth: Superlinear Means Unequal
The essay doesn't pretend this is all upside. Superlinear returns have a shadow, and Graham names it plainly: the steeper the return curve, the greater the variation in outcomes. A world that pays superlinearly is a world with wider gaps between winners and everyone else. The same math that makes compounding beautiful for the winner makes it brutal for the person one threshold short.
That's why he's careful about who should lean into this. The strategy suits two groups: people who are genuinely good enough that they'll come out ahead in a higher-variance world, and young people who can afford to take risks before they have much to lose. For everyone else, he notes, being part of a pool can be the safer and wiser choice. This isn't a "hustle harder" essay. It's a map of where the steep curves are, plus an honest warning that steep curves cut both ways.
There's a second caveat to add on your own behalf. Superlinear returns reward the specific over the generic, so the advice is useless without a real commitment to one thing. Spreading yourself thinly across ten fields keeps you on the flat part of ten curves at once. The people who benefit are the ones willing to look overcommitted and underdiversified for years. That's uncomfortable, and it's supposed to be. If it were comfortable, the far end of the curve wouldn't be so empty.
Frequently Asked Questions
What are superlinear returns in simple terms?
Superlinear returns describe any situation where output grows faster than the effort you put in, so doing twice the work can pay off ten times as much or more. Paul Graham argues this is the normal shape of the most important outcomes in life, even though we're usually taught that effort and reward move in a straight line.
What are the two causes of superlinear returns?
Exponential growth and thresholds. Exponential growth is compounding, where success in one round makes the next round easier, like an audience that recruits more audience. Thresholds are step functions, where crossing a performance line unlocks a large jump and falling short pays almost nothing, like winning a race or a customer by a hair.
How can a normal person, not a startup founder, benefit from this?
Learning is the most accessible superlinear return, because knowledge compounds and nobody has to grant you access to it. Concentrate on a field you're genuinely curious about, build a durable base of notes and highlights instead of forgetting what you read, and stay in it long enough to compound. Tools like Glasp's web highlighter and YouTube Summary exist to keep that base from leaking away.
Why does Paul Graham say to "do things that don't scale"?
Because the compounding phase of a superlinear curve is powerful but the first threshold has to be crossed by hand. Airbnb photographed listings door to door and Stripe installed its code on customers' laptops in person. Those manual, unscalable efforts got each company across the line where growth could start compounding on its own.
Is chasing superlinear returns risky?
Yes, and Graham says so directly. Steeper return curves produce more unequal outcomes, so the same math that rewards winners punishes those who fall one threshold short. He suggests the strategy fits people who can afford to bet on themselves, especially the young, while noting that being part of a pool is the safer choice for many.
Conclusion: Pick Something That Compounds
The reason "Superlinear Returns" lands isn't that it promises outsized rewards. It's that it explains why the world already works the way it does, and gently accuses most conventional advice of misreading the curve. Effort and reward aren't a straight line. They bend, driven by compounding and by thresholds, and the biggest outcomes cluster at the far, lonely end where few people bother to compete.
You don't need to found the next Airbnb to use this. You need to find one thing worth compounding, cross its first threshold by hand, and then refuse to quit during the flat early stretch where the returns haven't shown up yet. For most people, the surest version of that bet is learning, because knowledge compounds and no one can take the curve away from you.
If you want to start compounding today, make your reading durable. Capture what you highlight with Glasp's web highlighter, turn hours of video into notes you keep with YouTube Summary, and question everything you've saved with Glasp's AI chat. Small, consistent inputs, on a curve that bends in your favor. That's the whole game.