llama2.c running on original iPhone (240k)07-10 status: draft240k LLM runs on the very first iPhone.← How I made PrimeIntellect’s toploc 70x faster WIP A mildly cursed 3.5× triton tl.reduce optimization →