Rethinking: Will GenAI Shape New Platforms?
In the previous article, starting from the essence of networks, I pointed out that Generative AI is creating a new network structure by introducing new nodes and connections. In this piece, we explore a step further: Will new platforms emerge atop this new network? Will the logic of value capture differ completely from the past?
This brings to mind the evolution of the cosmos. In astrophysics, a nebula is a diffuse cloud of gas and dust—the raw material that nurtures stars and galaxies. Yet, most nebulae do not evolve into galaxies; only when conditions are right do they collapse to form a stable gravitational center.
By analogy, the multitude of tools and applications we see in the AI world today are like nebulae: vast in number and diverse in form. In the future, only a select few will collapse into true platforms, becoming the "galaxies" of the network. Therefore, identifying these potential nebulae—and supporting the founders capable of driving their evolution—becomes exceptionally important.
> Key Takeaways:
>
> - Definition: A platform is a collection of nodes in a network that share specific digital infrastructure and rules.
> - Human Platforms: Scarcity lies in human attention; thus, value capture occurs primarily on the Consumer/Application side.
> - Agent Platforms: Scarcity lies in infrastructure supporting mass collaboration; thus, value capture occurs primarily on the Supply/Infrastructure side.
Let's think step by step.
---
Why Platforms Exist in Traditional Networks
Real-world networks are not distributed evenly (Random Networks); they are Scale-Free Networks. This concept, proposed by Albert-László Barabási, emphasizes that a minority of nodes aggregate massive connections, forming a Power Law distribution.
However, this only explains the emergence of "Super Nodes." In reality, many Super Nodes are tightly linked, clustering together to form "Super Platforms."
To understand this, we must look beyond network structure and introduce Transaction Cost Theory. Ronald Coase (1937) proposed that transaction costs determine the boundaries of a firm; Oliver Williamson expanded this into Institutional Economics, emphasizing how contracts and governance structures reduce these costs. Later, scholars like Evans and Hagiu extended this logic to digital platforms, arguing that a platform is a new "organizational form" that significantly lowers market transaction costs through digital infrastructure and rule-setting.
Thus, we can say: Network Science explains why platforms emerge, while Transaction Cost Theory explains why platforms are necessary. Combined, they explain why digital platforms naturally emerge in reality and continuously trend toward concentration.
> A Platform is essentially an organizational form where a group of nodes in a network shares specific digital infrastructure and rules, thereby drastically reducing transaction and collaboration costs.
>
---
How Are AI Network Platforms Different?
In this new network structure, how will platforms evolve? We must start with the differences in Nodes and Connections.
The Difference in Nodes
In a system where humans are the nodes, attention is the scarcest resource. Therefore, traditional Human-to-Human platforms center their value capture around attention (Advertising, E-commerce, Gaming).
However, in a network where Agents are the primary nodes, a fundamental shift occurs. An Agent's "attention" is not scarce—they can process tasks in parallel for extended periods. What becomes truly scarce is the infrastructure supporting Agent work: compute power, toolchains, database access, and permissions. As the population of Agents and the scale of tasks grow, value capture will shift toward this infrastructure and the networks that facilitate Multi-Agent collaboration for complex tasks.
The Difference in Connections
1. Human Participation:
Human connections are finite. Dunbar’s Number tells us most people can maintain ~150 stable social relationships. Extending this to apps, humans typically use only about 100 apps. This is why consumer internet apps often follow a winner-takes-all logic.
In the Human-Agent network, this logic remains. However, there is a new trend: we now possess extremely rich and accurate "Intent Data." Compared to traditional behavior-based understanding, the combination of Intent Data and Behavioral Data may birth entirely new platform logic. (Note: The topic of Intent vs. Behavioral Data warrants a separate future article).
2. Agent-to-Agent Connection:
Unlike humans, Agents are not bound by Dunbar’s Number; theoretically, they can maintain massive connection volumes. Furthermore, the cost to establish or sever a connection is near zero. They can rapidly reorganize to find the most efficient nodes to complete a task.
This implies that their connections possess high volatility and instantaneous clustering effects. Once a node proves superior, a mass of Agents will converge upon it instantly—but they can migrate away just as fast. In other words, connections are dynamic. However, if multiple Agents rely on the same infrastructure and rules, stable collaborative relationships will form. Platforms are born from this stability of collaboration.
---
Which New Platforms Will Emerge?
Synthesizing the logic above, future platforms can be understood through two core directions:
1. Human-Agent Platforms: Aggregation and Conversion of Intent
In the Human ↔ Agent network, the platform's value lies in capturing limited human attention and converting it into reliable task execution. The flywheel effect here is: More Intent → Better Solutions → Better Results → Attracting More Intent.
Future forms include Search/Q&A gateways (e.g., Perplexity), Vertical Copilots (Education, Health, Productivity), and intelligent gateways for E-commerce/Content consumption.
- The Difference: They don't just attract attention; they accept intent and verify execution results. They don't just distribute content or goods; they invoke tools and service chains to complete tasks.
2. Human Collaboration Platforms: The Irreplaceable Value of Humans
In complex systems, greater collaboration and finer division of labor lead to higher efficiency. Yet, AI cannot fully replace human experience—it cannot taste a dish or physically experience a travel destination. The unique contribution of humans remains indispensable.
Therefore, the next generation of human collaboration platforms must solve two problems:
1. How to make Human-Agent collaboration smoother and more efficient.
2. How to use rules and incentives to ensure human creators receive fair returns.
This is the starting point for new benefit distribution mechanisms. The CEO of Cloudflare recently highlighted this exact issue: In the past, there was a tacit agreement between search engines and websites of "Content for Traffic." Ten years ago, Google’s ratio was roughly 2:1 (indexing two pages brought one click). Today, it has deteriorated to 6:1. AI models are even more extreme: OpenAI is roughly 250:1, and Anthropic reaches 6000:1. This means AI companies are voraciously consuming original content while returning almost nothing to creators.
Generative AI has broken the old equilibrium. The new platform opportunity likely lies in establishing a more reasonable value settlement mechanism between "Content and Model." Whoever achieves this first has the chance to become the gravitational center of the new network.
3. Agent Infrastructure Platforms: The Scarcity Supporting Complex Collaboration
Infrastructure platforms are critical because they are the scarce resources supporting Agent collaboration. Human attention is capped at 24 hours a day, but the demand for complex work has no ceiling. Even if human time cannot expand, output can be amplified through efficient collaboration.
New intelligence does not come from the capability upgrade of a single node, but from the emergence brought about by collaboration. The value of the platform lies in organizing the scattered Agent network by providing stable compute, toolchains, permissions, data access, and governance mechanisms, thereby nurturing system-level intelligence.
4. Agent Distribution/Collaboration Platforms: The Hub of Emergent Intelligence
Beyond infrastructure, another potential platform is the hub for distribution and collaboration. Complex intelligence is rarely achieved by a single Agent but relies on the division of labor and combination of many Agents.
The value of such a platform lies in:
- Efficiently decomposing tasks and distributing them to the most suitable Agents.
- Maintaining consistency, verification, and traceability during collaboration.
- Precipitating knowledge and experience reuse through task flows.
As the scale of collaboration expands, these platforms not only improve single-task efficiency but may also trigger an emergent leap in the overall system's intelligence level.
---
Conclusion
While predicting the future is difficult, this inquiry holds two unique values:
1. Faith in the Galaxy: Whether as founders or investors, we must believe that "galaxies" exist and strive to move in their direction.
2. Early Recognition: In the early stages, we must identify those founders with the potential to shape galaxies and support them strictly.
Future platforms will inevitably emerge, though their exact forms and timing remain unpredictable. What we can do is stay sharp during the "nebula phase," striving to detect the early signals of those conditions ripe for collapse.
The ultimate winners will endure like galaxies, becoming the gravitational centers of the new network order.