Recently, I’ve been pondering an interesting paradox: whenever new technology lowers barriers, people always say, "Now everyone can do it, so the advantage disappears." Cameras on smartphones make everyone a photographer, Spotify makes everyone a musician, AI makes everyone a software developer. It sounds reasonable, but the reality is quite the opposite.

The baseline indeed rises—more people participate in creation, more products are released. But the ceiling rises even faster. What’s the result? The gap between median and top-tier levels actually widens. This is the strange law of power-law distribution: technologies that promote equality tend to produce aristocratic outcomes.

Spotify is the best example. It broke the distribution monopoly of record labels, allowing any musician on Earth to reach a global audience. And the result? The top 1% of artists now capture a larger share of plays than during the CD era. Not smaller, but larger. More music, more competition, yet listeners gravitate toward the very best works. Spotify didn’t democratize music; it just intensified this race.

Writing, photography, software—all tell the same story. The internet has spawned the most authors in history, but also created a harsher attention economy. We’re surprised because we tend to think linearly—that productivity gains distribute evenly like water pouring out. But complex systems don’t work that way. Power-law distributions aren’t market quirks—they are the default setting of nature.

The question has shifted. When execution becomes cheap—anyone can generate functional products, beautiful interfaces, and runnable code in the afternoon—what truly distinguishes you?

The answer is aesthetics.

Steve Jobs insisted that the internal circuit boards of the first Macintosh had to be beautiful, even though customers would never see them. His engineers thought he was crazy. But he wasn’t. He understood: the way you do anything is the way you do everything. Someone who makes even hidden parts beautiful isn’t just performing quality—they have a character that can’t tolerate releasing subpar work.

Trust is hard to build but easy to fake. We constantly run heuristic judgments to figure out who is truly excellent and who is just performing excellence. Credentials can be manipulated, backgrounds inherited, but genuine aesthetic—an enduring, observable commitment to a high standard for some unspoken criterion—is the hardest to counterfeit.

In the SaaS era of the past decade, this signal was obscured. Execution standards became standardized, distribution became the real scarce resource. As long as your go-to-market strategy was strong enough, mediocre products could win. The signals of aesthetic quality were drowned out by the noise of growth metrics.

AI changes the signal-to-noise ratio. Now, “whether it’s usable” is no longer a differentiator; the question is: is it truly excellent? In a world where execution is cheap, aesthetics become the proof of effort.

My own experience confirms this. I grew up in a small town in India, the first in my state’s history to get into MIT. In a room full of alumni from prestigious backgrounds, I relied on depth. I studied physics, math, computer science—gaining insights from truths others overlooked, not from process optimization.

When I saw ChatGPT at the end of 2022, I realized the curve had bent. A new S-curve had begun. The phase transition no longer rewards those best adapted to the previous stage, but those who can see the infinite possibilities of the new stage before others do.

That’s why I founded Warp. I saw that in the U.S., there are over 800 tax agencies, each with its own filing requirements. For decades, every payroll provider handled this the same way: by staffing up. Traditional giants built business models around complexity—they didn’t solve it, they just turned it into more employees.

But I could see the improvement curve of AI agents. Someone deeply experienced in large-scale distributed systems could make a precise bet: that the fragile technology of today would become immensely powerful in just a few years. So we started from first principles, building an AI-native platform, targeting the most difficult workflows.

That bet is paying off. But more broadly, it’s about pattern recognition. The founders of AI-era technology not only have engineering advantages but also insight advantages. They see different entry points, make different bets. They can scrutinize systems deemed “permanently complex” by everyone else, ask: what’s needed for true automation? And then build the answer themselves.

But here’s a key variable: most AI-era founders are making a disastrous mistake.

A popular meme in the startup world is: “Escape the permanent bottom layer in two years.” Build fast, raise fast, or exit or perish. I understand where this mindset comes from. The rapid pace of AI evolution creates a survival crisis; seizing the window feels razor-thin. Young entrepreneurs see overnight success stories on Twitter and naturally think the game is all about speed.

But they’re right—on the wrong dimension.

Speed of execution is indeed critical—it’s even reflected in my company’s name. But the founders who will build the most valuable companies in the AI era aren’t those who cash out after two years, but those who sprint for ten and enjoy compound growth.

The most valuable things in software—private data, deep customer relationships, real switching costs, regulatory expertise—take years to accumulate. No matter how much capital or AI capability competitors bring, they can’t quickly copy this. When we handle payroll for multi-state companies, we’re accumulating compliance data across thousands of jurisdictions. Every resolved tax notice, every edge case handled, every state registration completed trains a system that becomes harder and harder to replicate.

This isn’t a feature point; it’s a moat. It exists because we’ve cultivated quality at a very high level for long enough to generate density.

This compounding isn’t visible in the first year; it’s faintly perceptible in the second, and by the fifth year, it’s the entire game. Snowflake’s former CEO Frank Slootman put it well: get comfortable with “uncomfortable”—not for a sprint, but as a permanent state. The early “fog of war” in startups—directional uncertainty, incomplete information, decisions that must be made—won’t disappear after two years; it just evolves. The enduring founders aren’t those who find certainty, but those who learn to move clearly through the fog.

Building a company is brutally hard. You live in a constant state of slight fear, punctuated by higher-level terrors. You make thousands of decisions with incomplete information, knowing that a string of mistakes could end everything. The “overnight success” stories on Twitter are not only outliers in a power-law distribution—they are extreme outliers. Optimizing strategies based on these cases is like studying the results of people who ran the wrong way and accidentally finished a 500M, to train for a marathon.

Why do it? Not because it’s comfortable, nor because the odds are in your favor, but because for some, not doing so feels like not truly alive. The only thing worse than the fear of building something from nothing is the silent suffocation of never trying.

And—if you bet right, if you see truths others haven’t priced yet, if you execute with aesthetic and conviction over a long enough cycle—the results will be more than financial. You will build something that truly changes how people work. You will create products people love to use. You will hire and empower those who perform at their best here.

This is a ten-year project. AI will never change that—never has. What AI changes is the ceiling that those founders can reach in those ten years.

So what will software look like?

Optimists say AI creates abundance—more products, more builders, more value distributed. They’re right. Pessimists say AI destroys moats—anything can be copied in the afternoon. They’re partly right. But both focus on the bottom line, ignoring the ceiling.

The future will see thousands of point solutions—tiny, functional, AI-generated tools that solve narrow problems. For low-barrier, easily replaceable software categories, the market will democratize truly. The bottom line is high, competition fierce, profits razor-thin.

But for mission-critical software—systems handling money flows, compliance, employee data, legal risks—the situation is entirely different. These are workflows with extremely low fault tolerance. Payroll failures mean employees don’t get paid; tax errors mean the tax authorities come knocking; benefits disruptions mean real people lose security. Those choosing software must bear the consequences. This responsibility can’t be outsourced to an afternoon’s patchwork AI.

For these workflows, companies will continue to trust vendors. The “winner takes all” dynamic will be even more extreme than in previous generations of software. Not only because network effects are stronger, but because an AI-native platform that has accumulated proprietary data across millions of transactions and thousands of edge cases will have a compound advantage that makes “leapfrogging” almost impossible. The moat isn’t just features; it’s the quality built over long periods of high-standard operation in error-prone domains.

This means the software market’s consolidation will surpass SaaS. I expect in ten years, HR and payroll won’t have 20 companies with single-digit market shares. Instead, two or three platforms will dominate most of the value, with a long tail of point solutions barely making a dent. The same pattern will play out across every category with complex compliance, data accumulation, and switching costs.

Companies at the top of these distributions will look very similar: founded by technically talented people with genuine aesthetic sensibility; built from day one on an AI-native architecture; operating in markets where incumbents can’t restructure without dismantling their existing business. They will have made a unique insight bet early—seeing some unpriced truth created by AI—and stuck with it long enough for compound growth to become clear.

Warp’s three-year track record proves this bet. Since launch, it has processed over $500 million in transactions, growing rapidly, serving the most important tech companies in the world. Every month, the accumulated compliance data, edge cases handled, and integrations built make the platform more difficult to copy and more valuable to customers. The moat is still early-stage but already sizable and accelerating.

I tell you this not because Warp’s success is guaranteed—nothing in a power-law world is—but because the logic that brought us here is exactly the logic I’ve described throughout: seeing the truth more deeply than anyone else, building standards that can be maintained without external pressure, and sticking long enough to see if the compound effect is real.

In the AI era, the most excellent companies will be built by those who understand: entry has never been scarce; insight is. Execution has never been a moat; aesthetics are. Speed has never been an advantage; depth is.

The power-law law doesn’t care about your intentions, but it rewards correct ones.