https://github.com/EDrTech/Working-memory-depth-recurrence https://gitlab.com/erikrudec-group/Working-memory-depth-recurrence https://codeberg.org/erikrudec/Working-memory-depth-recurrence/ This is a demonstration, in pure python, of a different way of making, well, AI. No backprop, no gradients, no weight transport, only local rules. Everything learns on one graph, and you can run all of it on almost anything. Have you ever seen an LLM solve the S4 or S5 card shuffle problem? I hav...