Media Mental Model
A Conceptual Index of The Mediator
If you’re a student of anything that could be considered a system (like companies, industries, ecosystems, value chains, markets, networks, economies, culture, or geopolitics), you probably have a model of it in your head, whether you call it that or not.
A good model should capture the most important concepts, assumptions, and causal relationships and explain how the system behaves. It will help you understand and predict how the system will respond to change. It will also help you manage what is an increasingly hard problem: knowing what’s worth your attention as the volume of information balloons. A good model gives you a filter to make sense of the daily deluge: does this new information support, weakly challenge, or strongly contradict my model—or, as is usually the case, can I discard it as irrelevant noise?
Putting the model on paper makes it a lot more useful. It exposes logical flaws and knowledge gaps. Those knowledge gaps are extremely valuable, because they point you to the most important questions to ask. Writing it down also encourages you to revisit your assumptions and conclusions as facts change.
I think of The Mediator as a living, breathing mental model of the media business. But right now this model is implicit, spread among my articles. For new readers, it’s not clear where to start. It also isn’t clear how the ideas fit together (or if they do). And it isn’t clear which ideas are relatively settled and which are more likely to change.
So, my goal in this post is twofold:
Write down the key concepts of this model in one place—a kind of conceptual index—with links to the most relevant articles (most recent first).
Outline the most important open questions around each of those concepts, including what might change my mind.
My intent is to update this as I add new articles and the model evolves.
Dynamic Equilibria, Punctuated Equilibria, the Internet, and GenAI
I want to quickly get to the key concepts in the model—that conceptual index—but we’ll start with the most important meta-concept: over the last 20 years, the media industry went through one shock, enabled by digitization and the internet, and now another one is coming, enabled by GenAI.
Every system is shaped, in part, by its constraints. They determine what the system can do, how it organizes, and how resources (energy, information, capital, labor, raw materials) flow within it. Feedback loops determine how the system responds to change. Positive feedback loops accentuate the change and negative feedback loops counteract it.1
In most systems, negative feedback loops are more prevalent than positive feedback loops. So, most of the time, the system remains in a sort of dynamic equilibrium. It is always churning but the overall structure remains relatively stable or it vacillates around a steady trendline.
Occasionally, however, there is an exogenous shock that throws the system out of whack by changing the nature of the constraints themselves, causing what biologists call punctuated equilibrium—a period of dramatic change that reshapes the system. In market-based systems, those shocks may come from technology, geopolitics, regulation, or some social upheaval.
The media business has just gone through one upheaval—a punctuated equilibrium—and now another is coming.
This brings me to the most important meta-concept in my model:
Digitization/the internet was a shock to media; GenAI will be another. The current state dynamics in the media business that I describe below were largely produced by two related technological shocks: digitization and the internet. Collectively, they made information computable and changed the architecture of media. Most important, they drove the cost of distributing information toward zero.
A second shock is now underway. GenAI makes meaning computable. This will arguably have an even more profound effect on the media business. One important consequence is that it will drive the cost of creating information toward zero. Just as the internet collapsed the cost to move bits, GenAI will collapse the cost to make them. But lower creation costs are only one consequence of computable meaning. It also has the potential to change how media is discovered, marketed, translated, coordinated, and monetized, and it will likely enable entirely new media forms. To put it another way: the media business has just gone through one upheaval and now another one is coming. The future state dynamics I describe below are my current best predictions about the effects of this coming shock.
See:
Media Model
Below, I outline the most important concepts in the model. They all pertain to: 1) demand dynamics (drivers of aggregate demand, changes in consumer preferences, or changes in price elasticity); 2) supply dynamics (changes in market structure, cost structure, or bargaining power along the value chain); or 3) both. The first section describes the current state, enabled by digitization and the internet, and the key observed concepts. The second section describes the future state, enabled by GenAI, and the key predicted concepts.
Current State (Observed)
1. Time spent with media is stagnant
Total time spent with media has plateaued because human attention is finite and the practical ceiling has largely been reached. For instance, U.S. adults already spend more than 75% of waking hours with media. This structurally constrains total value growth in the industry because media is predominantly monetized through advertising, which correlates directly with time spent.
See:
Key open questions: Could AI increase time spent with media by freeing up more leisure time, such as through greater labor efficiency or self-driving cars? Even if the number of hours spent with media doesn’t grow much, could monetization per hour rise enough to offset flat time growth (#14)?
2. Attention is fragmenting
Within the context of a relatively fixed pool of time, attention is fragmenting in media. This is driven by many dynamics, including an explosion in content supply, especially from creators; a shifting consumer definition of quality away from high production values (#3); and the structural advantages of social platforms, which exploit human biology to make consuming habitual.
See:
Key open questions: To what degree is this fragmentation occurring because social media is not necessarily preferable, but is low friction to consume? Could the advent of an AI discovery layer (#13) that redirects consumer attention swing the needle back?
3. The consumer definition of quality is shifting
A common, but not commonly-well understood, side effect of disruption is that consumers’ definition of quality may change. (Here, “quality” is not a Platonic ideal or expert’s definition, but is what consumers actually do—it is revealed preference.) New entrants not only compete on the incumbents’ traditional measures of performance, but they often introduce new features that may change what consumers value. This is occurring in media. Consumption is moving away from high production values in video, gaming, and music toward other attributes of quality, such as authenticity, convenience, community, and relevance. These new features tend to have lower barriers to entry than the former attributes of quality.
See:
Key open questions: Similar to the prior point, is this shift driven by the lower friction of social content and will swing back if it becomes easier to consume content that meets the traditional (high production) definition of quality (#13)? Will GenAI enable new innovative forms that further shift consumers’ preferences (#12)?
4. Power and attention are concentrating
Today, most media is consumed on networks. Owing to the way signals propagate through networks, they are subject to very powerful positive feedback loops that make strong things stronger. On the supply side, this manifests as network effects that concentrate power in a handful of tech platforms, so media is increasingly dominated by non-media companies. On the demand side, this results in power law-like popularity distributions, with a skinny head of massive hits, a narrowing torso, and a functionally infinitely-long tail. This distribution alters the economics of content creation, because it increases risk, puts more pressure on the formerly lucrative “middle,” and shifts bargaining power to the most successful talent in the head. Moreover, the popularity distributions in the creator economy are far more extreme than the corporate media economy, which may be a harbinger for all media as barriers to entry collapse.
See:
Key open questions: Could better discovery and recommendation systems, enabled by AI, counteract power law dynamics by surfacing quality in the tail (#13)? Could AI disintermediation or regulation meaningfully check platform concentration?
5. Monetization per hour is under pressure
Over the last two decades, relatively fixed demand plus ever-increasing content supply has driven down monetization per hour across video, print, music, and gaming. That’s because new media platforms monetize lower than the legacy media they have displaced. This has proved surprisingly persistent even though new digital media platforms have been raising prices for years. This is simply supply and demand dynamics at work.
See:
Key open question: Does GenAI-driven supply growth accelerate deflation further, or do new engagement-based monetization models partially offset it (#14)? Do price increases from digital media eventually bridge the gap with legacy media?
6. Technology is disintermediating traditional middlemen
Most of the household names in media, like studios, networks, streamers, labels, and publishers, are middlemen between the creatives who make and the consumers who consume. These intermediaries have extracted much of the value in media because they fulfill critical roles that have historically been very hard for creatives to do themselves. But technology is both democratizing and disaggregating the media value chain, enabling creators to perform functions themselves—production, distribution, marketing, monetization—that previously only institutional intermediaries could perform. This has given rise to the estimated $300B creator media economy, which is growing far faster than corporate media. Within the context of a flattish total market, creator media will almost invariably continue to take share.
See:
Key open question: Can traditional intermediaries retain critical roles in the value chain through scarce capabilities—financing at scale, global marketing, access to top talent, professional validation, marketing and distribution infrastructure, IP ownership, access to first-party data—that become relatively more valuable as content creation itself becomes abundant?
7. Boundaries to the global flow of content are falling
For decades, the U.S. has been the dominant exporter of culture globally. As the largest content market, U.S. content companies could justify larger production budgets, attract bigger stars, and best navigate global distribution complexity. The internet eliminated geography as a constraint to free-flowing content across borders, leaving market structure, regulation, and language and culture as the remaining constraints. The rise of global self-distribution platforms has further weakened these remaining constraints because global platforms have challenged local market structures, are harder for regulators to police, and have introduced a sort of global internet culture that competes with local culture. These shifts have weakened the role of U.S. content globally and given rise to K-dramas, K-pop, Latin trap, Chinese video games and other international content that is increasingly finding success across borders.
See:
Key open question: To what degree will GenAI bring down the final remaining constraint, language and culture?
Future State (Predicted)
As mentioned above, the central thesis in the model is that we are approaching another punctuated equilibrium, catalyzed by GenAI. The following concepts are predicted outcomes from its continuing advancement.
8. GenAI will exacerbate current state dynamics
GenAI operates at the semantic layer—the layer of meaning—while digitization operates at the syntactic layer—the layer of raw information, abstracted from its meaning. Computable meaning has many applications, but the most immediate and certain application is cheaper creation, which exacerbates several of the current state dynamics mentioned above. By driving down the cost of creation and further reducing entry barriers in media, this will likely increase fragmentation (#2), concentration (#4), deflation (#5), disintermediation (#6), and globalization (#7).
See:
Key open questions: Might other applications of GenAI—such as new GenAI-enabled content experiences that improve engagement and monetization (#12); an AI discovery layer that directs consumers to more enriching content (#13); or agentic advertising that increases ad yields—counteract some of these trends?
9. The effect of GenAI will be determined more by consumer acceptance than by technological development
There are many unknowns about how GenAI will affect the media business, but most of those can be reduced to just two: 1) how far will the technology advance (i.e., how realistic will it get); and 2) will consumers accept it and for which use cases? Of these, the latter is a larger impediment than the former. There is no obvious theoretical barrier to continued improvements in media generation quality. All digital media is represented by bits and the bits generated by a digital video camera or codec are no different than the bits generated by an AI video model. By contrast, consumer receptivity is a wildcard. It seems likely that consumers will embrace AI for some use cases, but not others. And there are early warning signs of an intense social backlash.
See:
Key open questions: For which use cases will consumers accept or embrace GenAI? Could compelling GenAI-native applications—which are clearly not replacements for traditional forms, but altogether new forms—soften consumer resistance (#12)?
10. Value will shift to complements and chokepoints
As the cost of content creation falls, it will disrupt media business models built around content scarcity. When disruption occurs, value migrates to what I call the 4Cs: challengers (i.e., disruptors), consumers, complements, and chokepoints. Complements are goods and services that become relatively more scarce and valuable when their complementary goods become abundant. Chokepoints are structural points of control. Examples of complements that may grow in value include abundance-driven complements that help consumers manage the volume of content; creator-driven complements that help the emerging creator class; fan-driven complements, namely what fans would pay more for; new business possibilities; and new frictions. Likely structural chokepoints include distribution, curation, IP, first-party data, community, and human provenance, among others.
See:
Key open question: Which specific complements and chokepoints end up mattering most? Can incumbents successfully reposition toward scarce assets, capabilities, and market positions or is organizational inertia insurmountable?
11. Content will become top of funnel
This is an extension of the prior concept. As creation costs fall, competitors will flood in and prices will migrate toward marginal cost, which in this case is zero. The logical conclusion is that content itself will shift from profit center to cost center—a marketing mechanism and entry point to more valuable downstream relationships, experiences, and monetization, rather than the primary revenue source. This is arguably already happening in some media, such as in music, mobile gaming, and even the creator economy, in which the content is top-of-funnel for other forms of monetization.
See:
Key open question: Can incumbents adjust culturally and operationally to an environment in which the content itself is not the profit center?
12. GenAI will enable genuinely new media experiences
Usually, the first applications of a new media technology are mere imitations of the prior technology, also called skeuomorphism. The first online magazines were static PDFs, the first films were recorded stage plays, the first video games were analogs of sports or board games. The dominant narrative about GenAI in media today is also skeuomorphic, namely that it will enable creatives and creators to make the same things they make today, but for lower cost and with less labor (cheaper movies, synthetic music, and engineless video games). The real value creation opportunity, however, lies in genuinely new applications—or what we can call neumorphic—such as interactive, conversational, real-time, spatial, multimodal, and agentic media. These new forms may increase consumer engagement and willingness-to-pay and perhaps soften consumer resistance to GenAI media.
See:
Key open question: Which specific neumorphic applications gain consumer traction, and when?
13. AI may create a new discovery and coordination layer
Another potential application of computable meaning is the emergence of an AI discovery and recommendation layer that sits above the media distribution platforms. This could be a personal agent or chatbot. Such a layer could be a much better mousetrap for consumers than current platform-specific search and recommendation, for a few reasons: it would be able to surface content across platforms and maybe even formats or modalities; it would understand highly nuanced requests; it would have far deeper and broader context about consumers’ behavior, preferences, and circumstances; and, at least in theory, its interests would be more aligned with consumers’. While much of the focus in GenAI is its ability to create new content, its ability to coordinate content (match consumers and suppliers) could prove just as disruptive.
See:
Key open questions: Who is best positioned to control this layer? Will existing platforms successfully resist? Or will they try to become the cross-platform discovery layer themselves?
14. Time spent with media won’t likely grow, but engagement might
More than half of media revenue is derived from advertising, so media monetization has historically been tied to time spent, or attention, which is structurally bounded (#1). But willingness-to-pay is tied to intensity of engagement—the depth of emotional attachment, fandom, and parasocial relationships—which is unbounded. Media companies have an opportunity to increase engagement by focusing more on superserving fans. GenAI-enabled neumorphic experiences (#12) could also boost engagement and willingness-to-pay.
See:
Key open question: Does engagement-based monetization grow broadly enough beyond a small superfan population to offset the attention ceiling at an industry level?
15. The advertising market will bifurcate
As GenAI floods the network with content, signal-to-noise online will degrade, reducing the value of certain types of advertising impressions. In addition, consumers are likely to turn to AI to help them manage all the AI. If they filter content through AI agents or chatbots, many ads may not be viewed at all. In that environment, only two types of marketing activity will have value: those that create trust and those that prove (or guarantee) performance.
See:
Key open questions: Does risk reallocation shrink the total addressable ad market as agents disintermediate exposure to ads altogether—or does it merely redistribute the existing pool? How can trusted environments prove they are trusted?
Implications
The dynamics above paint a picture of an industry going through a wrenching, if exciting, series of transformations. The most important implications depend on whether you are a creative/creator, incumbent, startup, investor, policymaker, or social critic. They are fleshed out much more in the linked articles and in my (pending) book, Infinite Content. But here is the high-level sketch.
Media is mostly a zero-sum game
Time spent with media is not growing much, which is a structural constraint. The two plausible engines of growth are engagement-based monetization (#14) and, relatedly, genuinely new GenAI-enabled experiences (#12) that deepen willingness-to-pay. Both are real but unlikely to move the needle at an industry level in the near term. For most participants, the baseline assumption should be that a relatively fixed media pie will be redistributed, not that it will expand much.
As content becomes abundant, value shifts to what is enduringly scarce
Some areas of scarcity are obvious: trust, authenticity, relationships, first-party data, distribution, curation, etc. The trick for everyone in media is to figure out what’s scarce and lean into it.
Consumers are the clearest beneficiaries
More content, lower prices, better discovery, and new experiences will all accrue to consumers.
Creators will be empowered, but the power laws are brutal
GenAI is a democratizing technology. It will enable more creators to tell stories, create music, and develop games than ever. It will also give them more tools to build direct audience relationships and monetize them and, overall, the creator economy will likely continue to take share. But the same dynamics that empower creators also intensify competition for attention, deepening the power law distributions that already govern media (#4). The spoils will be enormous for those at the top and negligible for almost everyone else.
For incumbents, the environment is getting harder along every dimension
Competition is rising from creators and platforms simultaneously, and AI-native entrants are coming. Risk is rising as power law distributions make content investment less predictable and squeeze the middle. The consumer definition of quality is continuously evolving, not just preferences in genres or styles. Uncertainty is growing as GenAI reshapes the value chain faster than it is practical to reorient strategy. And operational complexity is increasing as incumbents are forced to manage legacy businesses in decline and build new capabilities under strained budgets. Incumbents have certain natural advantages: capital, infrastructure, relationships, data, institutional expertise, IP, brands, etc. Some will successfully leverage these assets, capabilities, and market positions to adapt and perhaps even thrive in this environment, but most won’t. It will require foresight, commitment, speed, and the ability to overcome the institutional calcification and internal misaligned interests that so often hold companies back.
This model will continue to change. When evidence strengthens or weakens any of these dynamics and my thinking inevitably evolves, I’ll update it here.
A network effect is an example of a positive feedback loop; the network gets more valuable as it grows, which attracts more participants, which makes it bigger, and so on. Sweating, like other homeostatic biological processes, is an example of a negative feedback loop; when you get hot, you sweat, which cools your skin and reduces your temperature.