In 2035 every process with have a 0.1B LLM running at 60x human capacity, with half the overhead and twice the work! ;)