I built a lossless geometric ML representation for a year. It failed, but the point-attractor model survived (opens in new tab)

Physics-inspired NLI: vector collapse engine with semantic basin attractors - chetanxpatil/livnium