a16z
December 18, 2025

Two Futures | Runtime 2025

This a16z manifesto cuts through the noise, declaring the current moment a defining period for intelligence itself. It presents a clear choice: build the foundational systems for AI, or risk obsolescence. The message is a direct call to action for those ready to shape the next era of technology.

Unprecedented Scale, Unmatched Ambition

  • "There's never been a technical project of this complexity and this scale ever. It has all led to this."
  • Historical Convergence: All prior technical advancements, from early computing to the internet, have culminated in this moment, enabling projects of previously unimaginable scope.
  • Resource Intensity: Building these systems demands immense "fuel"—silicon, advanced models, and intellectual capital. Consider building a global brain: it requires more than just code; it needs specialized hardware, novel architectures, and a coordinated effort across disciplines.
  • Complexity as a Feature: The intricate nature of current AI and foundational tech is not a bug; it's a characteristic of building intelligence from first principles.

Building the Invisible Layers of Thought

  • "The invisible layers, the systems that quietly interconnect other systems, and the tools upon which all other tools are built. AI was born here. It grew up here. It will lead here."
  • Foundational Infrastructure: The most impactful work occurs in the underlying systems—the protocols, frameworks, and core models that enable all higher-level applications. Think of the internet's TCP/IP; most users never see it, but it underpins every digital interaction.
  • AI as a Core Utility: AI is not merely an application; it is becoming a fundamental building block, a "tool upon which all other tools are built," much like electricity or the operating system.
  • Domain-Specific Intelligence: The future requires specialized AI models "fashioned for every domain," moving beyond general-purpose systems to highly optimized, contextual intelligence.

Two Futures: Build or Be Left Behind

  • "Two paths, two futures, teaching sand to think or being left in the dust. You have the opportunity to build the foundation for intelligence itself. It is time to build."
  • Binary Choice: The current era offers a clear choice: actively participate in "teaching sand to think"—creating intelligence from raw computational power—or face irrelevance.
  • Defining Opportunity: This is a rare chance to contribute to the "foundation for intelligence itself," akin to the early days of personal computing or the internet.
  • The Builder's Moment: The message is an urgent imperative for builders to step up. This is a "moonshot" moment, a grand challenge with profound implications for the future of human and machine cognition.

Key Takeaways:

  • Strategic Implication: The next decade's value will accrue to those building foundational AI infrastructure and the "invisible layers" that connect intelligent systems.
  • Builder/Investor Note: Focus capital and talent on core AI models, specialized domain intelligence, and the underlying computational fabric. Superficial applications risk rapid commoditization.
  • The So What?: This is the defining period for the architecture of global intelligence. Participation now determines future influence and relevance.

Podcast Link: https://www.youtube.com/watch?v=4bJOe2tuLb8

This episode confronts the monumental scale of AI's ambition, presenting a stark choice between actively shaping intelligence and succumbing to obsolescence.

The Unprecedented Scale of AI Ambition

  • AI represents a technical project unmatched in complexity and scale throughout history.
  • The pursuit aims to generate boundless digital realities, signifying a new frontier for creation.
  • Past technological advancements culminate in this singular moment, setting the stage for future intelligence.
  • “There's never been a technical project of this complexity and this scale ever. It has all led to this.”

The Bifurcation of Futures

  • Humanity faces a binary choice: contribute to the development of artificial intelligence or risk irrelevance.
  • "Teaching sand to think" metaphorically describes the process of imbuing inert matter (silicon) with intelligence.
  • The alternative implies a failure to adapt, resulting in technological and societal stagnation.
  • “Two paths, two futures, teaching sand to think or being left in the dust.”

The Foundational Pillars of Intelligence

  • Core components include "silicon and light models," representing hardware and optical computing advancements.
  • These models must be "fashioned for every domain," indicating the need for specialized AI architectures.
  • "Invisible layers" and "systems that quietly interconnect" refer to the middleware, APIs, and protocols forming the AI stack.
  • The development of "tools upon which all other tools are built" highlights the importance of foundational AI frameworks and platforms.
  • “It takes silicon and light models fashioned for every domain. The invisible layers, the systems that quietly interconnect other systems, and the tools upon which all other tools are built.”

The Imperative to Build

  • The present offers an unparalleled chance to construct the fundamental infrastructure for artificial intelligence.
  • This endeavor represents a "moonshot," demanding significant innovation and investment.
  • Individual contributions now will leave a lasting "footprint in dust," shaping the future of intelligence.
  • “You have the opportunity to build the foundation for intelligence itself. This will be your moment, your moonshot, your footprint in dust. It is time to build.”

Investor & Researcher Alpha

  • Capital Reallocation: The emphasis on "silicon and light models" signals a continued, intensified capital flow into advanced semiconductor manufacturing, photonics, and novel computing architectures beyond traditional GPUs. Investors should scrutinize early-stage ventures in these foundational hardware layers.
  • Research Bottleneck Shift: The focus on "invisible layers" and "systems that quietly interconnect" points to a critical research gap in interoperable AI middleware, secure multi-modal data pipelines, and robust AI operating systems. Research into these systemic integration challenges will yield significant returns.
  • Obsolete Paradigms: A purely application-layer focus without engagement in foundational infrastructure or novel compute paradigms risks obsolescence. Researchers and investors solely pursuing incremental improvements on existing LLM architectures, without considering the underlying "sand to think" challenge, will find their efforts marginalized.

Strategic Conclusion

This episode asserts that the future hinges on building the foundational infrastructure for intelligence itself. The industry must now prioritize the creation of robust silicon, light-based computing, and interconnected systems to avoid technological stagnation.

Others You May Like