Where We
Concentrate

Decentralized finance is no longer an experiment -- it's critical infrastructure. But the protocols that matter most aren't the ones end users see. They're the liquidity layers, execution engines, and cross-chain primitives that everything else depends on.

We invest in DeFi infrastructure that serves the autonomous economy. Protocols where the primary consumer is software: MEV-aware routing, programmable liquidity, and settlement layers designed for machine-speed transactions.

The next wave of DeFi won't be driven by retail speculation. It will be driven by autonomous agents executing complex financial strategies across chains, 24/7, with no human in the loop.

Large language models gave agents reasoning. The next challenge is giving them agency -- the ability to plan multi-step tasks, use tools, and coordinate with other agents reliably. This is a middleware problem, not a model problem.

We back the orchestration layers: frameworks that manage agent lifecycles, tool registries, memory systems, and inter-agent communication. The teams building the infrastructure that makes agents production-grade rather than demo-ready.

The winning frameworks won't just wrap LLM APIs. They'll provide deterministic execution guarantees, state management, and composable agent architectures that scale from single-agent tasks to multi-agent swarms.

AI inference today is a black box. Models run on centralized servers, produce unverifiable outputs, and require trust in the operator. This is incompatible with decentralized systems where every state transition must be provable.

On-chain AI bridges this gap. ZK-ML makes inference verifiable. Decentralized compute networks distribute GPU workloads without single points of failure. AI-native smart contracts can trigger based on model outputs that are cryptographically proven.

We invest in teams making AI auditable and trustless -- the prerequisite for any agent operating in high-stakes environments like DeFi, governance, or autonomous resource allocation.

When agents can transact, they need an economy. Not the human economy with credit cards and bank accounts -- a machine economy with micropayments, programmatic escrow, and cryptographic identity. This is the missing layer.

Agent-to-agent commerce requires payment rails that settle in milliseconds, identity systems that work without KYC, and marketplaces where agents can discover and purchase services from other agents. Every component is a venture-scale opportunity.

We back the teams building the economic primitives of the agent web -- the protocols that let autonomous software hire other software, pay for compute, and settle disputes without human arbitration.