Tribeca Venture Partners

We're overfunding intelligence generation. We're underfunding economic reorganization. Here's the uncomfortable prediction: a meaningful percentage of today's AI unicorns will not be top-20 companies of the AI era. Not because AI isn't transformative. But because most capital is still concentrated in the layer that proves intelligence — not the layer that reorganizes the economy around it. The sharpest counterargument: "The models are becoming the coordination engines themselves." Partially right — some will make the leap. But Microsoft is the only infrastructure incumbent in modern tech history that has successfully jumped to the coordination layer multiple times. For every Microsoft, there are dozens of Ciscos, Nokias, and IBMs that dominated one layer and never captured the next. And I don't mean workflow orchestration — connecting apps, automating triggers, moving data between systems. That's middleware and it gets commoditized. I mean economic coordination: systems that restructure how industries allocate capital, deploy resources, and make decisions. The difference between a stage manager and a conductor. One moves things along. The other shapes the outcome. There's a deeper problem. We're treating today's token-based LLM paradigm as the settled architecture for intelligence itself. It probably isn't. We still haven't solved reasoning, planning, or world models — and when those breakthroughs come, they may arrive through fundamentally different approaches. Tokenless architectures. Continuous reasoning. Paradigms we're only beginning to explore. Today's model leaders face a two-front war: a coordination layer emerging above them and the architectural ground shifting beneath them. We've invested at this intersection as early investors in companies like AppNexus, AlphaSense, and ACV Auctions. The pattern is always the same. The companies that compound aren't the ones with the most impressive raw technology at launch. They're the ones that become structurally load-bearing inside an industry — and that position holds regardless of which model architecture wins underneath. The internet didn't reward whoever moved the most packets. It rewarded whoever controlled commerce on top of the packets. AI will not reward whoever generates the most tokens. It will reward whoever reorganizes the economy around intelligence. The biggest risk right now isn't missing AI. It's mistaking technological inevitability for architectural finality. They are not the same thing. If you're building tokenless AI architectures, post-LLM reasoning systems, or coordination infrastructure for the agentic economy — DM me. I read every one. AlphaSense Gather AI Radical AI ACV Auctions AppNexus