In 2024, I had an idea for an AI company called Fumble AI.
The idea was to build specialized models and AI systems for organizations. Rather than give every company the same general-purpose intelligence, I wanted to build systems around what each organization actually needed.
I never launched Fumble AI. But the idea stayed with me, and the technical questions behind it kept getting deeper.
By 2026, during my graduate work in artificial intelligence at the University at Buffalo, I was exploring model architectures, inference systems, memory, and orchestration. That work raised a more fundamental question: why should the architecture of intelligence be treated as something fixed, when it is one of the most important things we can design? And if we can design that structure, how do we know what it actually does inside?
An arch does not stand because one stone is extraordinary. It stands because of the structure that connects its parts. The arrangement changes what the individual pieces can do.
Arcence is building Research OS to specify, test, and understand model architectures and mechanisms, Foundry OS to help organizations turn requirements into a controlled model-engineering and deployment workflow, and the Assurance Plane to keep evidence, provenance, and governance attached at every step.