Author: Naman Bhansali

Translator: Shen潮 TechFlow

Shen潮 Introduction: In the early stages of new technology dissemination, people always generate a kind of illusion of “technological egalitarianism”: When photography, music creation, or software development become easy, will competitive advantages disappear? Naman Bhansali, founder of Warp, combines his personal experience of crossing from a small town in India to MIT with his entrepreneurial practice in the AI-driven payroll sector, revealing a counterintuitive truth: The more technology lowers the threshold (Floor), the higher the industry's ceiling (Ceiling) rises.

In this era where execution has become cheap and can even be “vibecoded” by AI, the author believes that the true moat is no longer simply traffic distribution, but rather the hard-to-counterfeit “taste”, deep insights into the underlying logic of complex systems, and the patience to compound returns over a ten-year scale. This article is not only a cold reflection on AI entrepreneurship but also a powerful argument for the power law that “common technology leads to aristocratic results.”

The full text is as follows:

Every time a new technology lowers the entry threshold, the same predictions inevitably follow: Since everyone can do it now, no one has an advantage anymore. Camera phones have made everyone a photographer; Spotify has made everyone a musician; AI has made everyone a software developer.

This kind of prediction is always half right: the bottom line (The floor) has indeed risen. More people are creating, more people are releasing products, more people are joining the competition. But this prediction always overlooks the ceiling (The ceiling). The ceiling is rising faster. And the gap between the bottom line and the ceiling — that is, the difference between median levels and top-level levels — has not narrowed but is instead widening.

This is the characteristic of power laws: it does not care about your intentions. The technology of egalitarianism always produces aristocratic results. It is always so.

AI will not be an exception but will even behave more extremely.

The Evolution of the Market Structure

When Spotify launched, it did something truly radical: it made it possible for any musician on Earth to access distribution channels that were previously only accessible through record companies, marketing budgets, and great luck. The result was an explosion in the music industry — millions of new artists emerged, billions of new songs were released. The bottom line has indeed risen as promised.

But what happened afterward was that the top 1% of artists now capture a larger share of plays than during the CD era. It did not shrink; it grew. More music, more competition, and more ways to seek quality content led audience members, no longer constrained by geography or shelf space, to gravitate towards the best works. Spotify did not achieve music egalitarianism; it merely intensified this tournament.

The same story has played out in writing, photography, and software. The internet has birthed the largest number of authors in history but has also produced a more brutal attention economy. More participants, higher stakes at the top, the same basic pattern: a tiny minority captures the vast majority of value.

We are surprised by this because we are used to linear thinking — we expect productivity improvements to be evenly distributed, like pouring water into a flat container. But most complex systems do not operate this way; they never have. Power law distributions are not quirks of the market or technical default; they are the default settings of nature. Technology did not create it; it simply revealed it.

Think of Kleiber's Law. Among all living beings on Earth — from bacteria to blue whales, spanning 27 orders of magnitude in weight scale — metabolic rate is proportional to the 0.75 power of body weight. The whale's metabolism is not proportionate to whale size. This relationship is a power law and holds with remarkable precision across nearly all life forms. No one designed this distribution; it merely represents the form that energy takes when following its inherent logic in complex systems.

The market is a complex system, and attention is a resource. When friction disappears — when geography, shelf space, and distribution costs no longer buffer — the market converges to its natural form. This form is not a normal distribution bell curve but a power law. The story of egalitarianism coexists with aristocratic outcomes; this is why every new technology catches us off guard. We see the bottom line rising, and we assume the ceiling follows at the same pace. The truth is not so; the ceiling is accelerating away.

AI's push in this process will be faster and more ruthless than any previous technology. The bottom line is rising in real time — anyone can release products, design interfaces, and write production environment code. But the ceiling is also rising, and faster. The question worth asking is: what truly determines your ultimate position?

When Execution Becomes Cheap, Aesthetic Becomes Signal

In 1981, Steve Jobs insisted that the internal circuit board of the first Macintosh had to be beautiful. Not the appearance, but the inside — that part the customer would never see. His engineers thought he was crazy. But he wasn’t crazy. He understood something that could easily be dismissed as perfectionism but is actually closer to some sort of proof: the way you do anything is the way you do everything. Someone who can make the hidden parts beautiful is not performing quality; he simply cannot tolerate releasing any inferior product.

This matters because trust is hard to build but easy to fake in a short time. We constantly run heuristic judgments, trying to discern who is genuinely excellent and who is merely performing excellence. Credentials help but can be manipulated; pedigree helps but can be inherited. What is truly hard to counterfeit is taste — that is, a persistent, observable, heightened adherence to some unasked standard. Jobs did not have to make the circuit board beautiful. He did it, and that act itself tells you how he would behave in the unseen places.

For most of the past decade, this signal was somewhat obscured. In the golden age of SaaS (roughly 2012 to 2022), execution became so standardized that distribution became a truly scarce resource. If you could efficiently acquire customers, build a sales machine, and achieve the “Rule of 40” — the product itself was nearly irrelevant. As long as your go-to-market strategy was strong enough, you could win with a mediocre product. The signals from aesthetics were drowned in the noise of growth metrics.

AI has fundamentally changed the signal-to-noise ratio. When anyone can generate a functional product, a stunning interface, and a runnable codebase in an afternoon, whether something “works well” is no longer a differentiator. The question becomes: is this thing truly outstanding? Does this person understand the difference between “good” and “insanely great”? Even if no one is forcing them, do they care enough to bridge that last little gap?

This is particularly true for business-critical software — those systems handling payroll, compliance, and employee data. These are not products you can casually trial and abandon by the next quarter. Switching costs are real, failure modes are serious, and those deploying the systems are responsible for the consequences. This means that before signing contracts, they will run all trust heuristic judgments. A beautiful product is one of the loudest signals it can send. It says: the people who built it cared. They care about the parts you can see, which means they’re likely to care about the parts you can’t see.

In a world where execution is cheap, aesthetics are proof of work.

What the New Phase Rewards

This logic has held true, but the market environment of the past decade has rendered it almost invisible. Once upon a time, the most important skill in the software industry was even unrelated to the software itself.

From 2012 to 2022, the core architecture of SaaS became established. Cloud infrastructure was cheap and standardized, and development tools matured. Building a functional product, while difficult, was a kind of “solved difficulty” — you could manage it through hiring, follow established patterns, and achieve the passing grade as long as resources were sufficient. What was truly scarce and could distinguish winners from mediocre players was distribution capability. Can you efficiently acquire customers? Can you establish repeatable sales actions? Do you know enough about unit economics to stoke the flames of growth with salary increases at the right time?

The founders who thrived in that environment mostly came from sales, consulting, or finance backgrounds. They were well-versed in metrics that sounded like a foreign language ten years ago: net dollar retention rate (NDR), average contract value (ACV), magic number, Rule of 40. They lived in spreadsheets and sales pipeline reviews, and in that context, they were correct. The peak SaaS period birthed peak SaaS founders. It was a rational evolutionary adaptation.

But I felt stifled.

I grew up in a small town in a state in India with a population of 250 million. Only about three students from all of India get into MIT annually. Without exception, they all come from expensive prep schools in Delhi, Mumbai, or Bangalore — institutions specifically established for that purpose. I am the first person in my state's history to enter MIT. I mention this not to boast, but because it is a microcosm of my argument: When entry barriers are limited, pedigree predicts outcomes; when entry barriers are open, deep people always prevail. In a room full of those from notable pedigrees, I was a chip that won through depth. That is the only betting method I knew.

I studied physics, mathematics, and computer science, fields where the deepest insights come not from process optimization but from recognizing truths that others miss. My master’s thesis was about straggler mitigation in distributed machine learning training: how to optimize this constraint without compromising the overall integrity when running large-scale systems.

When I looked at the entrepreneurial world in my early twenties, what I saw was a picture where these deep insights seemed irrelevant. Market premiums favored go-to-market rather than the product itself. Building something technically excellent seemed a bit naive — it was seen as interference with the “real game” (that is, customer acquisition, retention, and sales velocity).

Then, by the end of 2022, the environment changed.

What ChatGPT demonstrated — in a more intuitive and striking way than years of research papers — is that the curve has bent. A new S-curve has begun. Phase transitions do not reward those who adapt best to the previous phase; they reward those who discern the limitless possibilities of the new phase before others can price them.

So I quit my job and founded Warp.

This bet was very specific. There are over 800 tax authorities in the United States — federal, state, local — each with its own filing requirements, deadlines, and compliance logic. There are no APIs, no programmatic access interfaces. For decades, every payroll provider has handled this issue the same way: by piling on people. Thousands of compliance experts manually navigate these systems that were never designed for scalable operation. The traditional giants — ADP, Paylocity, Paychex — have built entire business models around this complexity; they do not solve the complexity but absorb it into their employee numbers and pass costs onto customers.

In 2022, I could see AI agents were still fragile. But I could also see the improvement curve. Someone who is deeply immersed in large-scale distributed systems and closely observes model evolution can make a precise bet: that at the time, fragile technology would become immensely powerful in a few years. So we bet: to build an AI-native platform from first principles, starting with the most challenging workflows in that category — workflows that traditional giants could never automate due to architectural limits.

Now, this bet is paying off. But a broader point is pattern recognition. Technical founders in the AI era not only have engineering advantages but also have insight advantages. They can see different entry points and place different bets. They can examine a system that everyone assumes is “permanently complex” and ask: what is required for true automation? Then, crucially, they can build the answer themselves.

The sovereigns of the peak SaaS era were rational optimizers under constraints. AI is removing these constraints and imposing new ones. In this new environment, the scarce resource is no longer distribution, but the ability to insight possibilities — and to build them to proper standards with aesthetics and conviction. But there is also a third variable that determines everything, and that is where most AI era founders are making catastrophic mistakes.

A Long Game at High Velocity

There is a meme circulating in the entrepreneurial circle today: you have two years to escape the permanent bottom. Build fast, raise fast, either exit or die.

I understand where this mentality comes from. The speed of AI evolution creates a sense of existential crisis, and seizing the window of opportunity feels extremely narrow. Young people casually seeing overnight success stories on Twitter assume that the essence of the game lies in speed — the winners are those who can run the fastest in the shortest time.

This is correct on a completely erroneous dimension.

Execution speed is indeed crucial. I believe this wholeheartedly — it is even etched in the name of my company (Warp). But the speed of execution is not synonymous with shallow foresight. Founders who can build the most valuable companies in the AI era are not those who cash out after sprinting for two years. They are those who sprint for ten years and enjoy the compounding.

Short-sightedness is flawed in that the most valuable things in software — proprietary data, deep customer relationships, genuine switching costs, regulatory expertise — require years to build up and cannot be quickly replicated, regardless of how much capital or AI capability competitors bring to bear. When Warp handles payroll for multistate companies, we are accumulating compliance data across thousands of jurisdictions. Every tax notice resolved, every edge case handled, every state registration completed is training a system that becomes increasingly difficult to replicate over time. This is not a feature; it’s a moat that exists because we have deeply dug for long enough at an extremely high quality to generate quality density.

This compounding is invisible in the first year. It becomes faintly visible in the second. By the fifth year, it is everything in the game.

Frank Slootman, the former CEO of Snowflake, who has established and scaled more software companies than anyone currently does, succinctly articulated this: get used to being “uncomfortable.” Not for a short race, but as a permanent state. The “fog of war” in the early stages of a startup—that sense of directionlessness, incomplete information, and the necessity of making action decisions — does not dissipate after two years. It merely evolves, and new uncertainties replace the old ones. Enduring founders are not those who find certainty but those who learn to move clearly within the fog.

Building a company is incredibly brutal, a brutality that is hard to convey to those who have not experienced it. You live in a constant state of slight fear, punctuated from time to time by higher levels of terror. You make thousands of decisions in the face of incomplete information, acutely aware that a series of wrong decisions can lead to termination. The “overnight successes” you see on Twitter are not only outliers in a power law distribution but extremes among outliers. Optimizing your strategy based on these cases is like training for a marathon by studying the results of those who ran five kilometers the wrong way, stumbling upon the finish line.

So why do it? Not because it is comfortable, not because the odds are good. But because for certain people, not doing so feels like not truly living. Because the only thing worse than the fear of “building something out of nothing” is the silent suffocation that comes from “not even trying.”

And — if you bet right, if you see truths that others have yet to price, if you execute with aesthetics and conviction over a long enough period — the results will be more than financial. You will build something that truly changes how people work. You will create a product that people love to use. You will hire and enable those who excel here in the business you’ve built with your own hands.

This is a ten-year project. AI cannot change this fact; it has never changed.

What AI changes is the ceiling (Ceiling) that the founders who can endure to see it through will achieve over these ten years.

The Ceiling That No One Is Watching

So, what will software look like on the other side of all this?

Optimists say that AI creates abundance — more products, more builders, more value distributed to more people. They are right. Pessimists say AI has destroyed the moats of software — anything can be replicated in an afternoon; defensibility is dead. They are also partially correct. But both sides are focused on the bottom line (The floor), while no one is paying attention to the ceiling (The ceiling).

Thousands of point solutions will emerge in the future — small, functional, AI-generated tools sufficient to solve certain narrow problems. Many of these are even not built by companies but by individuals or internal teams developing solutions for their own pain points. For certain low-threshold, easily replaceable software categories, the market will achieve true democratization. The bottom line is high, competition is fierce, and profit margins are thin as a gossamer.

But for business-critical software — those systems dealing with cash flow, compliance, employee data, and legal risks — the situation is entirely different. These are workflows with extremely low tolerance for errors. When payroll systems fail, employees do not get paid; when tax filings go wrong, the IRS shows up; when benefit contributions lapse during open enrollment, real people lose coverage. Those choosing the software must take responsibility for the consequences. This sense of responsibility cannot be outsourced to an AI cobbled together in an afternoon with “vibecoding.”

For these workflows, businesses will continue to trust vendors. Among these vendors, the “winner-takes-all” dynamic will be even more extreme than in previous generations of software. This is not only because the network effects are stronger (though they are indeed), but also because an AI-native platform that accumulates proprietary data through large-scale operations, millions of transactions, and thousands of compliance edge cases will have a compounding advantage that makes it nearly impossible for newcomers to achieve a “jumpstart” competition. The moat is no longer a set of functionalities but the quality accumulated over a long time running at high standards in a domain that punishes errors.

This means that the degree of consolidation in the software market will exceed that of the SaaS era. I predict that ten years from now, in the HR and payroll space, there won’t be 20 companies each capturing single-digit market shares. I expect two or three platforms to seize the vast majority of value, while a long list of point solutions will barely get a slice. The same pattern will occur in every software category where compliance complexity, data accumulation, and switching costs interact.

The companies at the top of this distribution will look very similar: founded by tech talents with genuine product aesthetics; built on AI-native architecture from day one; operating in markets where current giants cannot make structural responses without dismantling existing businesses. They placed a unique insight bet early on — seeing some truth that AI creates that has not yet been priced — and then stuck around long enough for compounding to become clear.

I have been abstractly describing such founders. But I know very well who he is because I am striving to become him.

I founded Warp in 2022 because I believe that the entire stack of employee operations — payroll, tax compliance, benefits, onboarding, equipment management, HR processes — is built on manual labor and old infrastructure, and that AI can completely replace them. Not improve, but replace. Traditional giants built billion-dollar businesses by absorbing complexity into their employee numbers; we will build our business by eliminating complexity from the source.

Three years have proven this bet. Since its launch, we have processed over $500 million in transactions, are growing rapidly, and are serving companies building the world’s most important technologies. Every month, the compliance data we accumulate, the edge cases we resolve, and the integrations we build make the platform harder to replicate and more valuable to customers. The moat is still in its early stages, but it is starting to take shape and is accelerating.

I tell you this not because Warp's success is fated — in a power law distribution world, nothing is destined — but because the logic that has guided us here is precisely the logic I have described throughout this text: See the truth. Dig deeper than anyone. Establish a high standard that can be maintained without external pressure. Stick around long enough to see if you're right.

Exceptional companies in the AI era will be built by those who understand these truths: Entry has never been a scarce resource; insight is; execution has never been a moat; taste is; speed has never been an advantage; depth is.

Power laws do not care about your intentions. But they reward the right intentions.