HEX: Humanoid-Aligned Experts for Cross-Embodiment Whole-Body Manipulation

HEX addresses a critical bottleneck in the humanoid robotics space: the lack of coordination in Vision-Language-Action (VLA) models which typically treat robotic limbs as independent actors. By introducing a 'humanoid-aligned universal state,' HEX attempts to create a standard interface for humanoid control that can scale across different hardware (cross-embodiment). The 17 forks against 0 stars in just 8 days suggest high internal/academic velocity and interest from the research community, likely associated with a significant paper release. However, its defensibility is low (4) because the core 'moat' is an algorithmic approach (MoE for body parts) that is easily replicated by well-funded frontier labs like NVIDIA (Project GR00T), Tesla (Optimus), and Google DeepMind. These entities possess the massive datasets and compute resources required to perfect this style of coordinated control. While HEX offers a sophisticated framework for cross-embodiment, its long-term survival depends on becoming an open-source standard for smaller humanoid startups before the major platforms consolidate the 'Humanoid Foundation Model' market.