I’ve been cobbling together my own personal agent for research and recall using data APIs and LLMs. I’d love to see if Nash could replace this long term.