AI is not just LLMs, frontier models, RAG, MCP.
AI.AI (rolls right of the tongue) is an exploration at the intersection of JEPA, latent and energy based models, and user-freedom.
The Future of Compute is Geometric
The Deterministic Engine: Traditional AI agents operate in unpredictable, bloated environments. We replace this with a highly optimized, deterministic game loop fed by a strict to-do pipeline. Every action is mapped to a mathematically verifiable "legal cause-effect graph," making system behavior 100% auditable and predictable.
The Hardware-Vision Bridge: We continuously capture the graphical desktop and generate a lightweight vector model of the visuals. Crucially, we map this directly to local system energy use and hardware "ramp-up" profiles, creating a perfect correlation between UI states and electrical energy.
Code is Just a UI: Take Python for example. We don't run Python. Instead, we execute purely mathematical, geometric functions directly against the vector model of the machine's state. Python is no longer an execution engine; it is simply edited and projected as a separate state representation. Python becomes a human-readable skin draped over geometry. Code is just UI.
Cutting-edge research in JEPA and energy-based models is applied to the small scale. We have questions to answer, like what happens when everyone can model then edit their existing system?
Check out these resources to learn more about what motivates our efforts:Electromagnetic geometry
Recursive, non-fractal, idempotent system states.