Capital for what is already underway.
Manifesto
AI-native, not AI-layered. Currents writes first or co-first checks into companies that could not exist without current-generation models in the loop. Four vectors of conviction. A single point of view, held long enough to be tested, written down often enough to be accountable.
Thesis
Four vectors. One frame. Currents writes selectively into AI applications and AI-services vertical plays where the company is AI-native rather than AI-layered. The vectors below are where conviction concentrates.
Agent infrastructure
The substrate beneath agent applications: orchestration, memory, tool-use grammars, evaluation harnesses, the runtime plumbing that determines whether agents are toys or operators. Most agent applications today are wrappers waiting to be commoditized. The infrastructure layer is where margin can durably accrue. Currents writes early into the primitives that outlast the current model generation.
Dual-use defense
Software and autonomy with both a defense buyer and a commercial buyer. Procurement reform is real, budget authority is in place, and talent migration into the category is now structural rather than novel. Currents backs founders building for both edges. Defense without commercial discipline drifts. Commercial without defense underprices the largest single buyer in the world.
Physical AI middleware
The translation layer between models and matter. Robotics, manufacturing, energy, logistics. The bottleneck is rarely the model and rarely the hardware. It is the middleware that lets a perception stack drive a real actuator inside a real facility under real liability. Currents backs the connective tissue, not the robot and not the foundation model.
Data center intelligence
Compute is the binding constraint on the AI economy, not a line item. The data center stack, power, cooling, scheduling, observability, capacity markets, is being rebuilt at speed. Currents backs the software intelligence threading through that buildout, sitting between hyperscaler capex and the operators keeping the floor running.
GP
Currents is run by Clayton Bryan. Years investing institutionally across hundreds of early-stage companies, with a working network of operators, allocators, and founders built across the prior cycle. The fund reflects a single point of view, held long enough to be tested, written down often enough to be accountable.
Write us
lp@currents.vc for allocators.
founders@currents.vc for companies.