And the final bit of coding today. I started working on the XPath-based splitter. This is what is going to break apart a chapter into paragraphs, lines, and tokens/words.

That way, an echo plugin can search for `//token[length() > 3]` to get all tokens over three characters long.

`//para//token[index() = 1]` for the first word of every paragraph. I want to use that to make sure no paragraph starts with the same word.

I figured I'm 2-4 days from seeing if this POC works.

Most of these plugins will have a "scope" which says how to break up the code. The default will be "content" which is actually `//file[has-class('content']` to only analyze chapters.

So, the echo plugins would only process per chapter.

If one used `/project` for the scope, then the echo would consider the last paragraphs of the previous chapter instead of only per-file checking.

This will also let the frequency plugins look for overuse of a word per file or per project.

I'll admit, I don't have hopes of getting past certain milestones on this iteration of AI. I think I've written and rewritten this at least nine times over the last eight years, each time find a new and interesting way it failed.

I have hopes it will finally get to be usable and extendable, but it also means I have to not get distracted for another 1-2 months. Otherwise I'll lose it and end up starting over again.

Getting to a point where others can ooh and aah over it helps.