LegendreGPT: Compressing a transformer into 15.7 MB with Legendre polynomials (github.com)

so we're coming closer to my dream of `npm install llm` and have some small agentic flows unlocked on modest VPS hardware soon? one can hope :-)