Skip to main content
Long-term directions include a portable dedicated-server move, optional external or local GPU acceleration, a local resident model, home embodiment, sensors/world modeling, and controlled self-improvement. Each is an adapter or capability gate, not a rewrite of canonical cognitive ownership. The permanent server is deferred and not selected. RunPod/external GPU is an optional accelerator candidate and must never hold irreplaceable cognitive data. Hardware or provider selection remains outside accepted V0.1 scope.