Excelent! That sounds like a very usefull personal benchmark then. You could test llama v2 by copying in different lengths of snippets from that conversation and checking how usefull you find its outputs.