BUILDING FOR AN AI-NATIVE✦ ECONOMY.
We are entering the first decade in history where software does not just respond it acts. Our AI thesis is built around the founders rewriting how work, capital, and intelligence move through the world.
For decades, software has been a tool something humans pick up, point at a problem, and put down. That definition is ending. The next decade of software will be agentic: systems that observe, decide, and act with increasing autonomy across complex domains.
We believe this shift will produce the largest reorganization of economic activity since the cloud. New companies will not just be built faster they will be built fundamentally differently, with AI agents as employees, customers, and counterparties.
Where we invest
Our AI thesis spans four areas of conviction. We back founders building at the layer where intelligence becomes infrastructure durable, defensible, and economically aligned with the operators who depend on it.
- Agentic infrastructure The runtime, identity, payment, and coordination primitives that let autonomous agents act in the real world without breaking it.
- Vertical AI applications Domain-specific systems built by founders who understand both the AI stack and the unglamorous workflows it will replace.
- Developer tooling Evals, observability, fine-tuning, and inference infrastructure for the teams shipping production AI systems.
- AI-native economies New markets, marketplaces, and financial systems built around agents as first-class economic participants.
Software stops being a tool. It becomes a counterparty.
What we look for
The next generation of AI companies will not be defined by model access. The frontier labs will compete on capability; the durable companies will compete on distribution, data, and execution. We back founders who understand that AI is the means, not the moat.
We look for teams with deep domain insight or strong systems instincts, founders who have built before, and a clear answer to why this company can only be built now. We are happy to write the first check, and happier still to be useful long after.
What we don't believe
We are skeptical of thin wrappers around foundation models, of "AI for X" without a real customer wedge, and of the assumption that bigger models alone will solve harder problems. The interesting work is happening at the seams where AI meets messy human systems and has to actually deliver.
BUILDING IN AI?✦
If you're building applied AI infrastructure, agents, or vertical AI systems we'd like to hear from you.
PITCH YOUR STARTUP