Buy or wait on local LLM hardware during the memory crunch · HackerLangs