[jfreds]

Having gone though exactly this exercise recently comparing a homegrown vector db against Qdrant, I’m wholeheartedly in agreement that getting a working solution FAST, and then spending a decent amount of time interrogating it (with help of LLM), is my favorite learning pattern