Byung-Chul Han tends his garden in Berlin. He is a world-renowned thinker. He has the luxury of refusal, the luxury of not owning a smartphone, of listening to analog music, of choosing contemplation over optimization. His analysis is brilliant. His life is admirable. But he is analyzing the world from a position that allows him to say no, to disconnect and not be wholly on an island.
There is a developer in Lagos who does not have a garden.
What Claude Code makes possible for her is important. Before AI coding assistants, building a software product required either a team or years of training in multiple programming languages, frameworks, and deployment systems. The developer in Lagos had the ideas. She had the intelligence. She had the ambition. What she did not have was the infrastructure: the team, the capital, the institutional support, the network of mentors and investors that turns a talented individual into a shipped product.
Claude Code changed the equation. Not completely. Inequalities of access, connectivity, and capital remain real. But the floor rose.
The argument for democratization cannot be made from a remote office. That’s why I felt the sprint to CES was so necessary, and why we took Station on the road across Europe, and why I flew to Trivandrum in February, and why I encourage all of my employees to not just engage with AI, but try building with it, whether they have coding experience or not. And it’s why I need to clarify once again that this is not about replacement, because you can’t replace the questions and ingenuity that lead to the remarkable things a person with new resources can create.
Now, I want to be concrete about what we did in February, because stating “twenty-fold productivity” as fact is a bold claim but one I stand by, and I can articulate the reality underneath it. On Monday, a team of three began building a feature for Napster Station’s multi-modal speaker detection system that had been on the backlog for four months. The estimate, under normal conditions, was six weeks of development time. By Wednesday afternoon, they had a working version – a working, tested, deployable version. It wasn't just about accelerating their existing output; it was allowing each of them to unleash different disciplines and achieve things they could have never dreamed of being able to do on their own. The twenty-fold multiplier is a bit misleading. It’s not just an increase of existing output by 20x it is a widening of the output people can create across a much broader problem space with multiple disciplines they might not be proficient in before.
The senior engineer from the Trivandrum training, the one I described in Chapter 1 who spent his first two days oscillating between excitement and terror, became the test case for what democratization means. His expertise did not become irrelevant. It became the judgment layer that directed the tool. Years of deep knowledge about systems architecture, about what works and what breaks, about the thousand decisions that separate a prototype from a product – all of that mattered more, not less.
The tool did not replace the engineer. It made him exponentially more potent. And the capability that mattered most was the layer that had been masked by implementation labor his entire career. It was obvious to me that the more capable the person was, the more robust the output they got out of Claude. While entry level peoples output looked very similar, the more advanced developers created more intricate and differentiated solutions. They brought more of themselves to the partnership. The juniors signal was less visible in the output as the AI did most of the work of directing the process.
Alex Finn, whose year of solo building I described in Chapter 2, is the test case. Han reads auto-exploitation. I read something more complicated: A person who could not have built this product at all five years ago. A person whose ideas had no path from imagination to reality. A person for whom the imagination-to-artifact ratio dropped from infinity to a conversation.
Is the pace sustainable? Almost certainly not: 2,639 hours, zero days off. The cultural dams need building. But the capacity itself, the capacity of a single individual to build something that serves real users and generates real revenue without institutional backing or a technical co-founder or a year of runway, that capacity is new. And its implications extend far beyond American tech culture.
The developer population worldwide has crossed forty-seven million, and the geography of that population is shifting faster than any previous decade. The fastest growth is in Africa, South Asia, and Latin America, the places where the gap between imagination and artifact has historically been widest, where brilliant ideas have routinely died for lack of the institutional infrastructure to realize them. There is a very clear dampening of this growth as people question if they should even enter the profession that is about to be dominated by thinking machines (more on this later). But this massive cohort has a decision to make. Run for the woods (flight) or pivot to leverage this new found superpower to do more (fight).
Most ideas fail simply because their drivers give up before they get there. But what if getting there was that much faster? How many ideas would survive the journey?
A student in Dhaka can now access the same coding leverage as an engineer at Google. Not the same salary. Not the same network. Not the same institutional support. Not the same safety net if the project fails. But the similar leverage, the same capacity to turn an idea into a working thing through conversation with a machine that does not care where you went to school or who your parents know or which accent you speak English with.
I am not claiming AI eliminates inequality. It does not. But it threatens a class of privilege more than the disenfranchised for sure. Sit with that; it’s not something we are used to seeing.
Access requires connectivity, and connectivity requires infrastructure that billions of people do not have. It requires hardware that costs more relative to local wages in Lagos than in San Francisco. It requires English-language fluency, because the tools are built by American companies, trained on predominantly English data, and optimized for the workflows of Western knowledge workers. The cost of inference of these frontier models is very high. The tokens rendered that constitute the thoughts output could be cost inhibitive even to the affluent developer in San Francisco. But these barriers will fall fast as models reach a certain rubicon that is already better than advanced humans and then get optimized to reduce them. This process has already started and soon this level of more than human capability will be dirt cheap. The friction will ascend to a certain level past that level it will be rare air that only few humans can occupy then none. It is not the roof of this tower. We will make it to the more than human level on our journey together this time.
The democratization is real but partial, and the partiality should not be hidden behind the grandeur of the claim. What I am claiming is more modest and more defensible: AI tools lower the floor of who gets to build.
They make it possible for people who were previously excluded from the building process by lack of skills, capital, or lack of institutional access, or lack of years of specialized training, to participate. The expansion of who gets to build is the most morally significant feature of this technological moment.
Han gardens in Berlin and describes the degradation that smoothness causes. But for the engineers in that room in Trivandrum, smoothness is not the enemy. The barriers between their creativity and its expression? Years of friction that had nothing to do with productive struggle and everything to do with access to resources? Those are the enemy.
A philosophy of friction that cannot account for the rising floor has told only half the truth. The privileged half.
The developer in Lagos does not need more friction. She has plenty. Unreliable power grids. Limited bandwidth. Economic precarity. Distance from the centers of capital and institutional support. What she needs is the smoothness that AI provides: the removal of barriers between her intelligence and its expression.
But democratization has a companion argument, and it is an economic one. When the cost of production approaches zero, what happens to quality?
This is not a new question. It has been asked at every technological transition that reduced the cost of making things. When Gutenberg's press made books cheap, the scholars worried that the flood of written material would drown out the most important works. When the internet made publishing free, the editors worried that the deluge of content would water down the type of writing that made a difference. When streaming made music distribution essentially costless, the musicians worried that the ocean of available sound would make it impossible for quality to surface.
In every case, the concern was legitimate. In every case, the flood came, and the noise increased. The incumbents fought with all their resources to stop the river. And in every case, the resolution was not less abundance but the need for better human judgment, curation, criticism, taste.
All of these are mechanisms for applying judgment to abundance. They are dams in the river, redirecting the flow toward quality. Far from being a solved problem, yet understanding the core ingredient of how to amplify human agency persists.
The age of AI is no different. When the cost of execution approaches zero, when anyone can produce anything that can be described, the premium shifts from the capacity to build to the capacity to decide what deserves to be built. The executor was the scarce resource in the old economy. The creative director and judgment of what to build is the scarce resource in the new one.
Judgment is the capacity to evaluate, to discern, to choose wisely among possibilities. It is taste applied to decisions. It is the ability to look at ten possible products and know which one deserves to exist, not because you can measure its market size but because you understand, in some deep and partially inarticulate way, what people need and what would serve them well.
The implications are immediate.
For organizations: The most valuable people will not be the most technically skilled. They will be the people with the ability to be the orchestrators, the creative directors, the multi-disciplinary thinkers.
For education: The emphasis must shift from teaching students to produce toward teaching them to see outside the fishbowl and think widely.
For individuals: The career question is no longer, "What can you do?" but, "What can’t you do?" and most importantly “What is worth doing!”
AI does not change what judgment requires.
It changes what judgment is worth.