The first generation of AI applications looked like chat. Every model call happened inside a conversation with a person waiting on the other end. Speed became the dominant objective, and infrastructure evolved around it.
Agents change the optimization problem. Agents plan, reason, call tools, retry, and execute long-running tasks without waiting on human interaction. A single task may generate hundreds or thousands of model calls. As agents become more capable, inference volume grows. Infrastructure needs to enable agents to do more without penalizing the way they operate.
Distributed computing isn't new. Researchers and companies have explored ways to aggregate idle computers into useful distributed systems for decades. Those efforts offered valuable lessons about coordination, incentives, and reliability. One thing was consistently missing: a killer application that naturally benefited from a global network of idle compute. AI changes that.
Agent workloads create sustained demand for inference at a scale previous applications never did. At the same time, millions of capable laptops and PCs spend most of their lives underutilized. If those machines can participate in inference reliably, they become a new supply layer for AI.
To solve that problem, we built ARC (Activations, Routing, and Control), the distributed inference network that powers Aquaduck. ARC exists to make ordinary laptops and PCs practical for AI inference. It coordinates independently operated machines into a single inference network, handling routing, scheduling, and orchestration so builders can access distributed intelligence through a single OpenAI-compatible API.
Join us
We're AI and platform nerds from Together AI, Apple, Block, and Royal Bank of Canada. We've spent our careers building AI tools, financial systems, developer platforms, and consumer software. Aquaduck brings those experiences together around a single problem: building infrastructure for where AI is going.
We're still at day zero. If you're interested in distributed systems, AI infrastructure, intelligent routing, or building something that doesn't have an existing playbook, join us: careers@aquaduck.ai.