About Qstar Labs
AI agents solve the same problems across your projects in isolation. We believe every lesson learned should compound across your workspace.
Born from research
The Frontier and Localhost paper surveyed 142 production AI systems and found that reliability has moved from model weights to the surrounding scaffold — instructions, skills, memories, tools.
OpenCore implemented local "dreaming" — session reflections that compound into engineering practices per project. But dreams stayed in silos. There was no system to aggregate, evaluate, and promote them across projects.
Qstar Labs is that missing system. The automated, eval-gated Loop 2 that aggregates learnings across your workspace's projects.
Team
Mikhail Arbuzov
Authored the Frontier and Localhost paper and OpenCore framework. Architecting the automated eval-gated Loop 2.
Lee Mosbacker
Leading go-to-market strategy and operations. Bridging research vision with product execution.
Mark Rafael
Guiding product direction and strategic partnerships across the AI tooling ecosystem.
Our philosophy
"Your projects dream. Q* evaluates. Your whole workspace gets smarter."
OpenCore captures. Q* aggregates. The eval gate proves. Workspace-scoped with federation on the horizon.
We don't move fast and break things. We move deliberately and build knowledge.