The data shows a paradox: AI content production costs are approaching zero, yet the value of what gets produced is collapsing. Every day, millions of AI-generated articles, images, and videos flood the digital ledger. The marginal cost of creation has dropped to near-nothing. And yet, the market is not rewarding producers. It's punishing them.
Consider the ledger of the last two years. The number of AI-generated 'slop' pieces — low-quality, low-effort content designed to game algorithms — has exploded. YouTube, TikTok, and even LinkedIn are drowning in it. The signal-to-noise ratio is deteriorating. The data shows that content platforms are facing a crisis of trust: users cannot distinguish between a human-written analysis and a machine-generated summary. The result? A liquidity crisis of attention. When confidence in content quality breaks, engagement dries up.
This is the backdrop for a recent essay by a16z partner Tim Sullivan, who argues that the true scarcity in the AI era is not 'taste' — the aesthetic sense that many assumed would be the last human frontier — but the social infrastructure required to develop judgment. Judgment, in this context, is the ability to verify, contextualize, and decide what is worth reading, watching, or believing. It is not a static skill; it is a dynamic capability that requires training, feedback loops, and exposure to diverse perspectives.
As an options strategist, I see this as a classic market structure problem. The AI content generation market has become a commodity market. The barrier to entry for producing content is now effectively zero. But the barrier to entry for producing good content — content that passes the 'judgment filter' — remains high. This is a structural imbalance. The market is flooded with supply, but the demand side — the human capacity to filter, verify, and judge — is not scaling at the same rate.
Sullivan's argument is grounded in historical precedent. He draws a line from Grub Street to the penny press to the rise of television, blogs, and social media. In each era, a drop in content production costs led to a surge in low-quality output, followed by a societal adjustment. The adjustment never came from technology alone. It came from the development of new social norms, new institutions, and new training systems that helped people navigate the increased volume. The same is happening now, but with a critical difference: AI is not just increasing volume; it is also replacing the entry-level jobs that historically served as the training ground for judgment.
This is where the market analysis gets interesting. In the traditional financial world, junior analysts spend years learning how to read balance sheets, identify anomalies, and develop a 'feel' for the market. They are trained through apprenticeship: senior mentors review their work, correct their assumptions, and gradually delegate more responsibility. The same applies in journalism, law, medicine, and finance. Judgment is not innate. It is built through a structured process of exposure, feedback, and repetition.
But here is the problem. Companies are now using AI to replace those entry-level roles. Why hire a junior analyst to summarize earnings calls when an LLM can do it in seconds? Why hire a junior reporter to write a first draft when a generative model can produce a passable piece? The cost savings are immediate and measurable. But the long-term liability is hidden: the pipeline for developing senior judgment is being severed.
This is the 'judgment gap' — a structural risk that the market has not yet priced in. Let me be clear: this is not a theoretical concern. In my own work, I have seen the impact of this gap. In 2022, during the Terra Luna collapse, my team survived because we had a circuit breaker in place — a pre-coded rule that halted trading 30 seconds before the crash. That rule was not created by an AI. It was created by a human who had spent years studying market microstructure and risk management. That human had developed judgment through years of exposure to volatile markets, through mentorship, and through the painful process of making mistakes and correcting them.
Now, imagine a future where that human is not trained. Where the junior roles that would have provided the training are automated away. The market would be left with a generation of traders who understand the mechanics of options pricing but lack the judgment to know when to override the model. This is not a hypothetical. It is the inevitable outcome of the current trajectory.
Sullivan's point is that we need to invest in the social infrastructure that supports judgment development. This includes mentorship programs, apprenticeship models, and — critically — the creation of institutional structures that reward verification and critical thinking over raw production. He cites Columbia University research showing that social influence and path dependency play a significant role in what becomes popular. This is not just about algorithms. It is about human networks. Ron Burt's 'structural holes' theory is relevant here: innovation and insight often come from bridging different communities. AI can help us access information across domains, but it cannot replace the human ability to synthesize that information into a coherent worldview.
The contrarian angle is this: everyone is talking about 'taste' as the new currency. The narrative is that AI can generate anything, so the only remaining differentiator is having good taste — a sense of what is beautiful, interesting, or valuable. This is a seductive idea, but it is wrong. Taste is a component of judgment, but it is not the same thing. Taste is subjective; judgment is objective. Taste is about preference; judgment is about assessment. You can have great taste and still make terrible decisions. You can appreciate a well-constructed argument and still fail to recognize a flawed premise. Judgment requires a framework for evaluating evidence, weighing risks, and making decisions under uncertainty. This is not something that can be developed by consuming more content. It requires active engagement, feedback, and the kind of adversarial testing that only comes from being in a community of practitioners.
The second contrarian point is that the 'judgment infrastructure' is not just about education. It is also about tools. We need to build verification systems, quality filters, and audit trails for AI-generated content. This is not a niche concern. It is the foundation for any functioning market. In the financial world, we have auditors, rating agencies, and regulators. In the content world, we have editors, fact-checkers, and peer review. These institutions are not perfect, but they provide a baseline of trust. Without them, markets fail. The same will happen in the AI content economy if we do not build equivalent structures.
So, what is the takeaway? The market is mispricing the value of human judgment. As AI content production costs continue to fall, the demand for verifiable, high-quality content will rise. The bottleneck will not be production; it will be curation. The winners will be those who can build — or invest in — the infrastructure that supports judgment. This includes AI content verification tools, expert networks, and training programs that prioritize critical thinking over rote memorization.
In the short term, I would watch for signals: Are major AI companies launching 'content verification' features? Are platforms introducing quality scores? Are enterprises redesigning their training pipelines to focus on judgment development? These are the leading indicators of a structural shift.
The final point is a question: In a world where content is free and judgment is scarce, who will own the infrastructure for judgment? The answer to that question will determine the next decade of the digital economy. Ledger books, not feelings, settle the debt. Audit the code, then audit the intent. Liquidity dries up when confidence breaks. Build the infrastructure, or be left with the slop.
