Daice Labs

Brookline, Massachusetts, USA · 11-50 employees · daicelabs.com

The Problem Current AI systems are brittle in production: they forget context, are costly to retrain, hard to audit, expensive to adapt, and difficult to align with the people creating value. Our Approach Founded by MIT CSAIL scientists, we build hybrid AI infrastructure for long-horizon work: composite systems that pair frontier models with neurosymbolic methods and bio-inspired system design. Our systems are built to improve, generalize, and adapt, supporting verification and attribution. Our Product Daice Labs is the continuous-learning AI infrastructure for governed work. Human teams supervise, co-build, and co-own outcomes—products, discoveries, and innovation. It is the integrated system for continuous collaboration: persistent context, domain-specific environments, sandboxed execution, and a co-ownership layer for governance and attribution. For selected projects, we provide compute and support, while builders keep the majority of commercial upside. At the core is an adaptive, verified compositional substrate for continuous learning. Verified results become shared primitives—reusable building blocks for new compositions. Compositions become parts of larger compositions, and patterns discovered in one domain transfer into others. This is the foundation for adaptive discovery: capabilities that compound across problems, domains, and model generations. Our Name Pronounced “dice”—exploration & programmable adaptation. The question is no longer how to build. It's what to bu