Show HN: Make a free 3.8B model as reliable as one 7× bigger at parsing datapypi.org4 pointspcoz22 days agohttps://github.com/pcoz/llm-feedback-control