Your pursuit of AI safety grounded in Hamiltonian physics and dynamical systems represents the exact underlying truth the industry needs. We cannot rely on statistical nudging; we must solve AI safety using a fundamental, physics-based best practice.
The OHM NI Stack aims to establish an environment where everyone can contribute to exactly such a framework. We have filed USPTO Applications (#63/994,444, #63/997,472, and #64/1,997,472) introducing the NI-SHIELD. We enforce predictability by utilizing dimensional alignment checks and "Kaiostic Entropy" as a hard physical constraint, utilizing Proof of Agent Work (POAW) to secure provable contracts between agents.
We are presenting this deterministic framework at the Vienna Summit this May. We want to establish this physics-engine approach as the global best practice for "provably limited" AI. We would be honored if you would contribute your expertise and review our physical proofs.