Everyone says the code is final. They are wrong.
Here's the thing about AI content: the marginal cost of production has hit zero. A model can generate a thousand blog posts, a hundred videos, or an entire social media feed for the price of a coffee. And yet, we're drowning in something that isn't content—it's slop.
Last month, a16z partner Tim Sullivan published a piece that cut through the noise. His thesis: the true scarcity in the AI era is not 'taste'—it's the social infrastructure for developing judgment. This isn't a feel-good essay about human creativity. It's a structural analysis of what breaks when production costs collapse. And for anyone trading on information asymmetry, it's a signal worth dissecting.
Greeks don't measure this kind of risk. But the market will price it eventually.
The Context: When Production Hits Zero, Distribution Becomes Everything
Sullivan's argument rests on a historical pattern that repeats with mechanical precision. Grub Street in the 18th century. Cheap penny presses in the 19th. Television in the 20th. Blogs and social media in the early 21st. Every time content production costs drop, the same debate erupts: is this destroying quality?
But there's a difference this time. AI's marginal cost is lower than anything that came before. Not incrementally lower—orders of magnitude lower. The barrier to entry for producing content that looks credible is now effectively zero.
This creates a specific market structure problem. When supply becomes infinite, attention becomes the scarce resource. And when attention becomes scarce, the mechanism for filtering quality becomes the battleground. Sullivan's point is that we're not prepared for this shift.
Columbia University research cited in the piece shows that social influence and path dependency—not intrinsic quality—often determine what becomes a hit. In an AI-saturated environment, this dynamic intensifies. The algorithms that distribute content are not designed to find truth. They're designed to maximize engagement. And engagement metrics reward the mediocre, the sensational, and the emotionally manipulative.
This isn't a technical failure. It's a structural one.
The Core: Judgment Is Not Taste—And That Changes Everything
The piece makes a critical distinction that most commentators miss: taste and judgment are not the same thing. Taste is the ability to recognize quality. Judgment is the ability to make decisions under uncertainty with incomplete information. Taste is passive. Judgment is active.
Sullivan argues that we've been fixated on taste as the differentiator in the AI era. But taste is relatively easy to cultivate—expose yourself to enough good work and you develop a sense of what works. Judgment, on the other hand, requires something far more complex: a social infrastructure that includes mentors, feedback loops, and the kind of apprenticeship models that have been eroded over the past decade.
This is where the analysis gets sharp. Ron Burt's structural holes theory, cited in the piece, argues that innovation comes from bridging gaps between different communities. The people who can move between worlds—who can see what the tech world knows and what the finance world knows and synthesize them—are the ones who create disproportionate value.
AI can help you access information across domains faster than ever. But it cannot teach you how to evaluate that information. It cannot teach you which signals matter and which are noise. That knowledge is tacit. It lives in the heads of experienced practitioners, and it is transmitted through apprenticeship, not through data sets.
Here's the uncomfortable implication: the institutions that built this infrastructure are dismantling it. Companies are replacing entry-level jobs with AI. Those entry-level jobs were the training ground for judgment. They were where young professionals learned to distinguish the signal from the noise, where they made mistakes in low-stakes environments, where they developed the pattern recognition that becomes judgment.
Take that away, and you create a judgment vacuum. In ten years, we won't have a shortage of people who can produce content. We'll have a shortage of people who can tell good content from bad.
The Contrarian Angle: The Scarcity Narrative Has a Self-Interested Source
Now let me push back on the frame itself.
Sullivan is a partner at a16z. The firm has invested billions in AI companies. When a venture capitalist tells you that 'judgment infrastructure' is the next scarce resource, you should ask: who benefits from that narrative?
The answer is: anyone who can build and monetize that infrastructure. If judgment becomes the bottleneck, then the companies that can provide judgment-as-a-service—content verification tools, expert networks, quality assessment platforms, training programs—become the next investment cycle's winners.
Code is law, but bugs are justice. The 'bug' here is that the judgment scarcity narrative conveniently positions a16z's portfolio companies as the solution to a problem they helped create.
The more interesting question, which Sullivan's piece doesn't fully address, is whether AI itself can augment judgment rather than just replace entry-level work. The models I've audited over the past year—the ones with real technical depth, not just marketing gloss—are getting better at reasoning. Not just generating text, but evaluating it. If that trajectory continues, the judgment infrastructure might be partially automated before the apprenticeship models can be rebuilt.
That would make the current scarcity temporary, not structural. And it would change the investment thesis significantly.
There's also a darker angle that deserves more attention than it gets. If judgment becomes concentrated in a small number of institutions—the ones that can afford to build the social infrastructure—we're creating a new form of gatekeeping. The NFT floor is a feeling, not a number, and the same applies to judgment. It's subjective. It's culturally embedded. It's not something you can standardize and sell without losing what makes it valuable.
The Takeaway: Position for the Judgment Gap, Not the Content Wave
For traders and builders, the actionable insight from Sullivan's analysis is straightforward: the content production arbitrage is closing. The people who made money by using AI to produce more content faster are going to find that the market is flooded. The next edge will come from those who can filter, validate, and judge.
The structural signal to watch is the entry-level job market. If companies continue to replace junior roles with AI, we're building a ten-year time bomb. The senior people who should be mentoring the next generation will retire. The infrastructure for developing judgment will be gone. And when that happens, the value of anyone who actually has judgment—who can look at an AI-generated analysis and spot the flaw, who can evaluate a protocol's code and identify the vulnerability—will be enormous.
That's the trade. Not in AI tokens. Not in content platforms. In human capital that has been systematically undervalued.
Will the market price this correctly? Markets are efficient in the long run. But the long run has a habit of taking longer than you expect. The question isn't whether judgment becomes valuable. It's whether you'll have it when everyone else realizes they don't.
The signal to watch: which institutions are rebuilding their apprenticeship models, and which are cutting them for short-term earnings? The answer tells you who's positioned for the next decade. And who's about to be left behind.
