Chris Dixon on How to Build Networks, Movements, and AI-Native Products

The Three Exponential Forces Driving Tech

The technology industry is fundamentally different from others because it’s driven by powerful, exponential forces that can overwhelm any tactical product decision. Founders and investors must first understand this landscape to position themselves on the right side of these curves. There are three primary exponential forces:

1. Moore’s Law: The classic example of compounding improvement, where semiconductor performance doubles roughly every 18 months. This principle extends broadly to other computing resources like storage and networking. Apple’s genius with the iPhone wasn’t just the initial product, but seeing and riding this exponential curve of hardware improvement.
2. Composability: This is the engine behind the rise of open-source software. By making software open, it becomes like “Lego bricks” that anyone can reuse and build upon. This harnesses the collective intelligence of the global community, allowing a project like Linux to grow from a hobby into the world’s dominant operating system.
3. Network Effects: A service becomes more valuable as more people use it. This force powered the growth of foundational internet services from email to Facebook. A network often starts small and limited—like Facebook being exclusive to Harvard—before expanding and achieving global scale.

The recent disruption caused by technologies like ChatGPT is a modern example of these forces at play. Neural networks were considered “toys” a decade ago, but their exponential improvement caught incumbents like Google, with their established business models, in an awkward position. The key is to bet on these curves of improvement.

Strategy: “Come for the Tool, Stay for the Network”

One of the biggest challenges in building a network is the “cold start problem”—no one wants to join an empty network. A powerful tactic to overcome this is to first build a valuable single-player tool that attracts users on its own merit.

• Instagram’s Origin: People forget that Instagram’s network was not the initial draw. It first attracted users by offering free, high-quality photo filters (the “tool”), which were paid features in other apps. It then piggybacked on existing networks like Twitter for distribution. Only after reaching a critical mass did its own network become the core, defensible asset.
• Modern Examples: This pattern is visible in modern productivity tools like Figma and Notion. They are highly useful for an individual, but their value and stickiness increase exponentially through collaborative, multi-player features.

For AI founders, this is a crucial concept. The space is currently filled with innovative tools, but many lack defensibility. Layering in a network is often the answer to moving a product beyond a fad and creating long-term engagement, but it must be a useful addition, not just a feature tacked on for its own sake.

Defensibility in the Age of AI

While network effects are the gold standard, defensibility in the modern tech landscape is multi-faceted.

• Brand and Inertia: The power of brand is often underestimated. ChatGPT became a household name almost overnight, creating a powerful moat through brand recognition and consumer inertia, even without traditional in-product network effects.
• Externalized Networks: The internet itself is now a mature, built-out platform. As a result, network effects can be “externalized.” When a tool like Midjourney becomes popular, an entire ecosystem of YouTube tutorials, influencers, and community guides springs up around it. This external network creates a powerful defensibility that lives outside the core product.
• Capital as a Moat: In the capital-intensive AI space, performance begets funding. After proving traction, raising a billion-dollar round can itself become a moat, providing the resources to stay on the cutting edge of a rapidly advancing field.

Platform Shifts and Navigating the “Idea Maze”

We are in the midst of a major platform shift driven by AI. Unlike previous shifts like iOS where the platform’s rules were explicitly defined, the properties of the AI platform are emergent and constantly being discovered.

This relates to the concept of the “idea maze,” which posits that a startup’s success depends on both the initial idea (choosing the right maze to enter) and the execution (navigating that maze’s twists and turns).

• The Netflix Example: Netflix entered the right maze with the core hypothesis that the internet would lead to subscription movies. However, they had to be incredibly agile in their execution, pivoting from mailing DVDs, to digital streaming, and finally to producing original content as the landscape shifted.

For AI startups, the maze is incredibly dynamic. The “meta-process” of AI improvement is scaling exponentially, creating immense opportunity. The challenge is that incumbent “god models” could subsume specific use cases. The primary strategy for startups is to go deep into a specific domain, building an edge through specialized knowledge, a strong brand, or a loyal user base that a general-purpose model cannot replicate.

From Skeuomorphic to Native AI

New technology platforms typically evolve through two phases:

1. Skeuomorphic Phase: The new medium imitates the old one. Early films were shot like stage plays; the early internet consisted of digitized brochures. Much of today’s AI feels skeuomorphic—image generation automates the work of a human illustrator.
2. Native Phase: Creators invent a new grammar and new forms of media that could only exist on the new platform. This led to the unique language of cinema and user-generated content on YouTube.

We are likely still in the skeuomorphic phase of AI. The truly “native” phase is yet to come and will likely be surprising. It might involve entirely new art forms we can’t yet imagine. Even the current “prompt-to-media” interaction feels like a temporary step. Articulating complex aesthetic preferences with words is inefficient. A more native interface might leverage a user’s data directly—like using a Spotify library to generate new music—bypassing the need for clumsy linguistic prompts.

The Critical Importance of Open-Source AI

Open-source software has been a profoundly democratizing force, enabling startups to build on a free, powerful software stack and making technology accessible globally. Ensuring a healthy open-source ecosystem for AI is critical to prevent the technology from being consolidated in the hands of a few large companies.

While AI’s capital requirements present a challenge, the outlook for open-source AI is cautiously optimistic. A likely and positive outcome is an equilibrium where open-source models are consistently a generation or two behind the state-of-the-art. These models will still be powerful enough for the vast majority of startups and consumer applications, ensuring that innovation isn’t gatekept and that new companies can continue to be built.

• Tags:
• Venture Capital
• AI Strategy
• Open-Source AI
• Platform Shifts
• Network Effects
• Exponential Growth
• Consumer Tech