The Silent Scarcity: Why Judgment, Not Taste, Is the Real Bottleneck of the AI Era

The market is quiet these days. Sideways chop, as we call it. Over the past few weeks, I have watched my feed fill with the same recycled narratives—the same token analyses, the same bold predictions—only now, they come faster, and they say less. It feels like standing in a library where every book has been written by the same author, with infinite time and no memory. This is the slow erosion of signal, the quiet arrival of what a16z partner Tim Sullivan recently called the era of "slop." We are drowning in generated words, images, and videos, yet we are starving for the capacity to tell what actually matters. This is not a problem of taste. Taste, we are told, is subjective. The real scarcity is judgment—the capacity to evaluate, to filter, to decide. And that scarcity is not a human trait we can simply wish for. It is an infrastructure problem, a social architecture problem, and it is, in my opinion, the most undervalued asset in our entire industry. The premise is deceptively simple. If content production costs approach zero, then the only remaining bottleneck is consumption—specifically, the human ability to filter the infinite from the meaningful. Historically, this bottleneck has been managed by a series of social institutions: the editor at the newspaper, the curator at the gallery, the senior engineer who taught the junior what "good" looks like. These institutions are not just job titles. They are the physical manifestations of what we call "taste"—a trained, embedded, contextual judgment. Sullivan argues, and I agree, that this is not a matter of individual genius but of social infrastructure. It is the apprenticeship, the critique, the feedback loop. It is the network of people who can say, "This is derivative," or "This is a breakthrough," not because they are smart, but because they have been placed in a social position to see the difference. I have seen this play out in my own career. In the early days of DeFi Summer, I spent 300 hours auditing Uniswap V2's smart contracts. I was not looking for bugs—I was looking for the soul of the code, the fair-launch philosophy. I had to learn judgment. It was not taught in a classroom. It was taught in a Telegram group of five anonymous strangers, in the long silences of my apartment in Singapore, in the painstaking process of reading and re-reading code until I could feel the difference between a design choice and a compromise. That is social infrastructure. It is the bear market of learning, where the value is not in the initial reward, but in the compounded wisdom. And I am concerned, deeply concerned, that our AI-driven world is about to make that infrastructure obsolete, or at least, dangerously scarcer. Let us examine the current narrative. The consensus is that AI will democratize creativity. The ability to generate high-quality text, code, images, and video will be available to everyone. This is true, and it is a wonderful thing. But the truth is also that the average quality of output will rise while the top-end quality will become... muddied. The "slop" effect, as Tim Sullivan describes it. We are seeing it in my own field: there are now AI-generated whitepapers that are technically perfect but conceptually empty. They have no thesis. They have no risk. They have no soul. They are the equivalent of a warm glass of milk—not bad, but not memorable. And the danger is not in the milk itself, but in the fact that it crowds out the rare, the strange, the structurally intelligent work that requires years of judgment to produce. The Columbia University research cited in the original article is crucial here. It suggests that the success of a piece of work is not merely a function of its quality, but of social influence and path dependency. In other words, what becomes popular is not necessarily what is best, but what is first seen by the right people, at the right time. This is where the "structural holes" theory of Ron Burt becomes so relevant. Burt's argument is that innovation does not come from the center of a network, but from the gaps between networks—the people who can bridge two different communities, who can see the pattern that is invisible to each side. In the AI era, we have an infinite capacity to generate information, but we have lost the structural capacity to bridge the gaps. We have all the data in the world, but we are losing the ability to make the connection. This is the core of my contrarian angle. Many will argue that AI is solving the problem of access. They say, "We can now use AI to synthesize information from all domains. We can connect the dots faster." This is true, to a point. AI can help us parse the explicit knowledge of the world. But it cannot help us with the implicit, the tacit, the embodied knowledge that only comes from doing and from failing. I have learned more from the times I had to debug a smart contract that had drained a user's funds than from any tutorial. That fear, that pressure, that sweat is the true, tacit knowledge. And AI cannot simulate that. It cannot simulate the feeling of the bear market, the silence of the portfolio, the quiet certainty of holding a conviction that is wrong. That is judgment. Sullivan's argument is a warning against the devaluation of the journey. When AI automates the entry-level jobs, it also automates the apprenticeship. The old model was: you get a junior role, you write the code, you make the mistakes, and you have a senior developer who tells you, "No, that's not good enough. Try again." That feedback loop, the iterative process of being wrong, is how judgment is built. In the age of AI, the senior is not there because the company does not need junior developers. The AI writes the first draft. The senior checks it. But the junior never learns. We are creating a generation of AI-native workers who are brilliant at using the tools, but are severely underdeveloped in the ability to judge the tool's output. This is the judgment deficit. This is the real risk. The market has not yet priced this in. I look at the current sideways market, and I see a lot of projects that are trying to solve the "AI problem" by building better models, faster models, cheaper models. But the model is not the bottleneck. The bottleneck is the human capacity to validate the model's output. We are building more and more content, but we are not building the human layer of validation. We are not building the curators, the editors, the critics, the communities of practice. We are not building the social infrastructure that makes taste and judgment possible. In a way, the market is still valuing the ability to produce, but it is devaluing the ability to evaluate. That is a mistake. Let me bring this to a more concrete, personal level. In my work with The Commons, a community for ethical Web3 builders, I have seen the power of this infrastructure. We host roundtables, but it is not about the topics. It is about the process. It is about the fact that a group of people are forced to sit, listen, and push back on each other. It is about the creation of a shared vocabulary, a shared standard of what is "good." This is the "social infrastructure" that Sullivan is talking about. It is not a DAO that gives you a token. It is a culture that gives you a compass. And this compass is the most scarce resource in the AI era. In the silence of the bear market, we heard the truth. We realized that the code is the covenant, not just the contract. And that covenant is built not by a smart contract, but by the people who agree to honor it. What is the investment angle? A16z, as an institution, is obviously looking for the next big thing. They are saying that the "judgment infrastructure" is the next big thing. I am not a VC, but I can see the logic. If the AI produces content, we need tools to validate it. We need expert networks. We need training programs. We need a new set of platforms that do not just produce content, but also help people build the capacity to evaluate it. This is a whole new asset class. But it is an asset class that is much more difficult to scale. You cannot scale judgment by just hiring more people. You have to scale the social infrastructure that produces it. And that is a slow, deliberate, and deeply human process. This is why my own analysis is tempered with a note of caution. I have a high level of conviction, but a low level of certainty. The regulatory environment, too, is important. Hong Kong's licensing push, Singapore's stablecoin, these are all forms of judgment infrastructure. They are trying to create a framework that helps people judge what is safe. But the regulation is about risk management, not about the production of judgment. It is about the validation of assets, but not about the validation of ideas. We need to build the human layer, and that is a more complicated task than any smart contract. It requires a new kind of moral synthesis. It requires us to combine the computer science with the philosophy, the technical with the human. Let me be honest about the blind spots of this argument. Sullivan's piece is optimistic in its pessimism. It is a warning, but it is also a call to action. It does not discuss the possibility of AI being used to actually enhance the judgment. I am an idealist, but I am also a pragmatist. There is the possibility that we can build AI systems that are themselves trained to be good critics, to be the new, the new curators. We are seeing the emergence of AI verification tools. But I believe these will be useful. I do not believe they will be sufficient. The reason is that judgment is not just a logical process. It is a moral one. It is a matter of values. And the values are not something you can optimize for. You cannot solve for the loss of a moral compass with a dataset. In the end, this article is not a summary of a16z's report. It is my own testimony, my own cry in the desert. We are building a world where the content is infinite, but the attention is finite. We are building a world where the production is cheap, but the judgment is expensive. We are building a world where the code is the contract, but the contract is meaningless without the community that enforces it. And in the silence of the bear, we heard the truth: every broken token taught me how to hold value. The value is not in the token. It is in the ability to see what is real. We are not in a liquidity crisis. We are in a judgment crisis. And the only way out is not to build more AI, but to build more of the people, the networks, and the trust that allow us to see what we are building in the first place. The next bull market will be built on that foundation, or it will not be built at all. The world is not going to be saved by the models. It is going to be saved by the modelers who can judge.