This is mostly about AI, but since all my samples are in Blazor I felt it was appropriate to post it here.

I work on mostly business applications and use a lot of retrieval augmented generation however I recently discovered the power of knowledge graphs and recently wrote a blog post about it and would love to hear feedback.

More Powerful than AI RAG: Building Lightweight Knowledge Graphs

Basically if you have any data schema that has related entities you can easily build a knowledge graph saved as simple in memory json, and expose that to your AI through function tool calling.

Then allow your end users to ask questions and query that data in ways unachievable without a knowledge graph. This can also be used to update and make changes to your source data.

Quick background, I've gone to college to learn to write and I have written a lot of fiction and nonfiction over decades.

I've also created AI writing programs. My main concentration now is on quality. I believe the reason why most AI writing sucks is it was trained on 'everything' and in general 'everything' is actually is pedestrian.

So, AI is more like a enhanced grammar checker it is incapable of producing what I call 'Quality' and by quality I mean writing that really grips you and moves you.

I am focused right now on the essence of quality so that I can feed these guidelines to the AI. I know there are prompts based on things like 'save the cat', but following those rules just produces 'decent' not gripping and emotionally moving on their own.

I am interested in thoughts on this.