Key Takeaways

  • Union Square Ventures' (USV) "Rebel Alliance" thesis argues that the AI opportunity is massive and expansive, contradicting the "Death Star" meme of value consolidating into a few large labs.
  • Generational companies will emerge “up and down the stack,” focusing on capital-light application-layer innovations and specialized user experiences, like Sunno.
  • A significant shift sees agents moving from 'personified people you hire to cloud infrastructure,' demanding new, sophisticated agentic systems.
  • Founders should target crucial infrastructure for this agentic future: model routing, orchestration, memory, and durable networks acting as APIs.
  • The call is to avoid hyper capital-intensive AI ventures and instead pursue more capital-light business models that leverage existing foundational models.

The AI "Rebel Alliance": Beyond the Death Star Labs

Nick Gman of Union Square Ventures (USV) introduces a direct counterpoint to what he calls the AI "Death Star" meme—the idea that a few giant labs will eventually consolidate all value in artificial intelligence. Gman champions the "Rebel Alliance" thesis, asserting the opportunity is "so big and so massive and so expansive and there are so many forces pushing for this expansive motion that there are going to be companies, you know, in in all these components of the stack." This perspective shifts focus from building core models to recognizing the vast ecosystem emerging around them. USV believes true generational companies will be built "up and down the stack," particularly through capital-light application-layer innovations and specialized user experiences. Think products like Sunno, which offer distinct value by building on top of existing AI capabilities rather than competing with them head-on.

Agents Shift from People to Cloud Infrastructure

Gman points to a subtle but profound shift already underway: agents are moving “from sort of people you like personified people you hire to cloud infrastructure.” This isn't just a conceptual leap; it suggests “the whole economy but maybe the whole internet getting rewritten on the agentic stack and with an agentic approach.” As agents transition from human-like roles to automated cloud components, a new set of infrastructure needs emerges. Gman states, “every company, every enterprise who's building in an agentic style is going to need model routing, model choice, orchestration, ... memory, ... some sort of harness.” These are the foundational components for managing, directing, and scaling the autonomous systems that will underpin much of the future internet. This creates a fresh greenfield for builders.

Capital-Light Bets in the Agentic Future

For founders looking to build in this space, Gman offers clear guidance on capital deployment. He advises against competing in the realm of “hyper capital intensive AI companies which is the big labs.” Instead, founders should seek endeavors that are “generally speaking, you're looking for things that are more capital light.” This means focusing on the infrastructure and application layers surrounding core models, rather than the models themselves. The real opportunity lies in developing sophisticated orchestration tools for agents, building specialized APIs that tap into emerging agentic networks, or crafting highly specific user experiences that solve acute problems for a defined audience. The key is to find leverage points within the AI stack that demand less upfront investment but still address critical, unmet needs within the expansive ecosystem.

What to Do With This

This week, pinpoint a specific process in your business or industry that an AI agent could significantly improve. Instead of trying to train a new AI model, outline the "harness" this agent would require: how it would route queries to the right model, orchestrate tasks, and manage its memory. Focus your next build on creating that capital-light orchestration layer or a specialized user experience that wraps around existing models, rather than investing heavily in foundational AI research.