Original Research
Pushing the boundaries of machine intelligence through original research and open collaboration.
Focus Areas
Foundation Model Optimization
Techniques for training and serving foundation models more efficiently — quantization, distillation, custom kernels — with the goal of running production-grade intelligence on workstation-class and edge hardware.
Autonomous Swarm Engineering
Large-scale agent orchestration where autonomous agents are the primary contributors to engineering work, not just assistants. A two-tier model — orchestrator plus ephemeral worker swarm — built to enforce engineering standards at scale.
Cognitive Architecture
Memory architectures for long-running agents. We combine modern Hopfield networks, hierarchical temporal memory, and complementary learning systems theory to build agents whose effective context isn't bounded by their token window.