They're not really of any educational value themselves because the model is so small and it's not rooted in any kind of factual basis - you'll probably be able to spot obvious errors in even very widely-covered events - but I'm working on a blog post about non-chat interfaces for LLMs and this is part of it.