Hi HN, we are building an open-source alternative to the "scale is all you need" paradigm. We believe the quadratic cost of Transformers ( O ( N 2 ) O(N 2 ) ) and their lack of physical grounding make them unsuitable for continuous, long-term autonomous agents. Genesis is an attempt to build Embodied AGI using: Architecture: Pure-SSM (State Space Models) using Mamba-3 for linear scaling O ( N ) O(N) . Inference Engine: Written 100% in Rust (using Oxidizr) for low-latency (<16ms) control loops. Learning: Homeostatic Active Inference (minimizing free energy/surprise) instead of just next-token prediction. We are simulating agents in voxel environments where they must survive, plan, and evolve social structures (validated against Turchin's Cliodynamics). The project is early stage (Phase 1: Survival). We are looking for Rust engineers and ML researchers interested in SSMs and ALife. White Paper: discord Feedback on our "3-Lane Temporal Architecture" is very welcome.