I noticed an interesting paradox that is currently puzzling everyone: when a new technology makes something accessible to everyone, why does inequality only grow?

Spotify allowed any musician to distribute their tracks. The result? The top 1% of artists captured an even larger share of streams than during the CD era. The internet created more creators in human history, but the attention economy has become harsher. Photography, programming, now AI — each time, the same story.

We are used to thinking linearly, expecting growth to distribute evenly. But complex systems don’t work that way. It’s not a bug of technology; it’s nature. Look at Kleiber’s law — the metabolism of all living organisms from bacteria to whales follows a power law. No one designed it; it’s simply how energy organizes itself in complex systems.

Markets are also complex systems. When friction disappears — geography, logistics, distribution costs — the market converges toward its natural form. And this form is not a Gaussian curve but a power law. That’s exactly why each new wave of technology catches us off guard.

But here’s what’s interesting: when execution becomes cheap, aesthetics become a signal. Remember how Jobs insisted on beautiful printed circuit boards inside the first Macintosh? Parts no one sees. His engineers thought he was crazy. But he understood something important: the way you make hidden parts — is the way you make everything.

The last ten years in SaaS were different. Execution standardized so much that the winner was whoever distributed and sold best. The product hardly mattered. Go-to-market strategy trumped even mediocre solutions. The signal of aesthetics was drowned out by the noise of growth metrics.

AI changed everything. Now anyone can create a functional product, a beautiful interface, working code in an hour. The question is no longer whether it’s convenient. The question: is it truly outstanding? Does the creator know the difference between good and great? Is it important enough for them to see it through, even if no one demands it?

This is especially critical for systems handling payroll, tax reporting, employee data. These are not apps you test and abandon. Switching costs are real, and the consequences of errors are serious. Companies conduct all possible trust checks. And a beautiful product is one of the loudest signals: the people who built it put effort not only into the visible but also into the invisible.

During the SaaS boom, the dominant force was the rational optimizer, well-versed in metrics. Founders came from sales, consulting, finance. They lived in spreadsheets, knew about NDR, ACV, magic number. And they were absolutely right for that time.

But that was an era of constraints. AI removes those constraints and creates new ones. Now, the rare resource is not distribution but the ability to see opportunity — and realize it with standards of aesthetics and conviction.

That’s why technical founders now have an advantage. Not just engineering skill, but insight. They see different entry points. They look at a system everyone considers “perpetually complex” and ask: what’s needed for true automation? And most importantly — they can build it themselves.

I remember, when I was in my twenties, I looked at the startup scene and saw that deep insights seemed irrelevant. The market rewarded go-to-market, not product. Creating technologically perfect things seemed naive. Then, at the end of 2022, everything changed. ChatGPT showed what years of research couldn’t express: the curve bent. A new S-curve began.

Phase transitions do not reward those who adapted best to the previous phase. They reward those who saw the endless possibilities of the new phase before others understood its value.

That’s when I founded Warp. The specific task: in the US, there are over 800 tax authorities, each with its own requirements. No API, no programmatic access. For decades, every provider handled this by hiring people. Thousands of experts manually managed systems not designed for scale. Traditional giants like Paychex embedded complexity into their model instead of eliminating it.

In 2022, I saw that AI agents were fragile, but I also saw the improvement curve. Someone deeply immersed in large-scale systems and observing model evolution can make a confident bet: the technologies that are fragile now will be incredibly powerful in a few years.

We built an AI-native platform from scratch, starting with the most complex workflow — one that traditional giants could never automate due to architectural limitations. Today, that bet is paying off.

But the key here is pattern recognition. Technical founders in the AI era see different entry points, make different bets. They look at a system everyone considers “perpetually complex” and ask: what’s needed for real automation? And then — crucially — they can build it themselves.

But there’s one more factor that decides everything. And most AI founders make catastrophic mistakes here.

In entrepreneurial circles, a popular meme: you have two years to get out of the bottom layer. Launch fast, attract funding quickly — or fail. I understand where it comes from. The speed of AI creates a sense of existential threat. The window seems incredibly narrow.

But that’s a mistake. Speed of execution is critical — it’s even reflected in my company’s name. But speed of execution does not equal narrow vision. The founders who will create the most valuable companies in the AI era are not those running for two years, but those running for ten and enjoying compound interest.

Because the most valuable elements of software — private data, deep customer relationships, real switching barriers, regulatory expertise — take years to build. You can’t copy them quickly, regardless of capital or a competitor’s AI capabilities.

When Warp processes payroll for companies across multiple states, we accumulate compliance data across thousands of jurisdictions. Every resolved notification, every edge case, every registration with government agencies — all of this trains the system, which over time becomes harder to copy. It’s not just a feature. It’s a protective barrier, existing because we’ve worked long enough with extremely high standards.

This compound interest is invisible in the first year. In the second year, it only hints at itself. By year five, it becomes the core of the game.

Frank Slootman, former CEO of Snowflake, put it this way: you must get used to constant discomfort. It’s not a sprint; it’s a state. The fog of war in early stages — the feeling of losing direction, incomplete information — will not disappear in two years. It simply evolves. New uncertainties replace old ones.

Successful founders are not those who found confidence, but those who learned to move clearly in the fog.

Building a company is a brutal process. You live in constant slight fear, sometimes interrupted by stronger terrors. You make thousands of decisions with incomplete information, knowing that a series of mistakes will lead to collapse. Those “overnight successes” on Twitter are not just outliers in a power law; they are extreme outliers among outliers. Optimizing strategy around them is like training for a marathon by analyzing people who got lost and accidentally ran five kilometers.

Why do it? Not because it’s convenient. Not because the odds are high. But because for some, not doing it is not truly living. Because the only thing worse than the fear of creating something from nothing is the quiet suffocation of not even trying.

And if you guessed right, if you saw a truth others have not yet recognized, if you acted with aesthetics and conviction over a long enough horizon — the result will be not only financial. You create something that truly changes how people work. You create a product people love to use. You hire people who thrive and reveal their full potential.

This is a ten-year project. AI will not change that.

So, what will the software architecture of the future look like? Optimists say AI creates abundance — more products, more value. They’re right. Pessimists say AI has killed competitive advantage — everything can be copied in an hour. They’re also partly right.

But both sides are looking at the bottom. No one is looking at the ceiling.

In the future, there will be thousands of one-off solutions — small, functional, AI-generated tools. Many won’t even be companies, just internal projects. For low-entry barrier software categories, the market will become truly democratic. Competition will be fierce, profits thin.

But for mission-critical software — systems handling cash flows, tax compliance, employee data, legal risks — the situation is entirely different. These are workflows with extremely low error tolerance. When payroll fails, taxes are audited, insurance is interrupted — real consequences follow.

For these workflows, companies will continue to trust providers. And the “winner takes all” dynamic will be more extreme than ever. Not only because of network effects, but because an AI-native platform that scales and accumulates private data from millions of transactions and thousands of compliance scenarios has a compounding advantage, making it nearly impossible for followers to start from scratch.

Entry barriers are not just features. They are quality, built over time through high standards in areas where mistakes are costly.

This means consolidation in the software market, surpassing the SaaS era. I expect that in ten years, the HR and payroll market will not have twenty companies with a few percent each. It will be two or three platforms with the overwhelming share of value, and a long tail of one-off solutions that will hardly amount to anything.

The same model will appear everywhere where compliance complexity, data accumulation, and high switching costs are at play.

Top companies look similar: founded by technical experts with a sense of product; built on AI-native architecture from day one; operating in markets where current giants cannot provide a structural answer without destroying their existing business.

They made a unique early prediction, saw the truth created by AI that others hadn’t yet appreciated, and held on long enough for the power of compound interest to become evident.

I founded Warp in 2022 because I believed: the entire stack related to personnel management — payroll, tax compliance, benefits, onboarding, equipment management — is built on manual labor and outdated architectures that AI can fully replace. Not just improve. Replace.

Big players built billion-dollar businesses by absorbing the complexity of personnel management. We are building a business by eliminating complexity at its root.

Three years confirmed this bet. We processed over 500 million in transactions, are growing actively, serving companies creating critical technologies. Every month, accumulated compliance data, edge cases handled, integrations built make the platform harder to copy and more valuable to clients.

Our competitive advantage is still early-stage, but it has formed and is accelerating.

I share this not because success was predetermined — in a world of power laws, nothing is — but because the logic that brought us here is what I described: see the truth, dig deeper than others, build standards you hold without external pressure, and stay long enough to see if you were right.

Companies that stand out in the AI era will be those who understood: access has never been a scarce resource, insight yes; execution has never been protection, taste yes; speed has never been an advantage, depth yes.

Power laws don’t care about your intentions. But they reward the right ones.