slm

slm: cagataycali's research direction on self-learning/online-learning models — "a model whose weights change while it runs," i.e. continual/in-context weight updates rather than frozen inference. Conceptually the deepest node in cagatay's self-modification family tree: G0 agi.diy (runtime tool self-extension, code not weights) → G1 slm/harness-optimizer/strands-coder → G2 tiny.technology.