A from-scratch, world-model-first, recursively self-improving learner that makes unknowns known, ever more cheaply. No pretrained weights. Entity 'Nous', open-ended growth, learned L1/L2 representations, and a physics-of-knowing thread.
Runs in your browser, served from our own domain. No account, no install. If it misbehaves in this frame, open it on its own.