Working on laconic, a "Ralph loop for Research"
API costs for LLM are getting ridiculous. By managing a 4K context window, I can get tiny local models to do advanced research tasks that require dozens of searches and building a knowledge graph for $0.