... and the language that we really should be intentional with using when we're helping students tackle those larger words.

Yes.

So this is another thing that may be a contribution or it might be something that people just hate.

But I think I, I think I've actually helped shift the conversation about this a little bit.

So syllable division for the less initiated, most, most educators know at this point, but what it's about is when you have particularly a word with, you know, two or more syllables, how do you divide it so that you can predict the pronunciation of the vowels in it, right? So there are two really common patterns.

There's the VCCV, where you have a vowel, two consonants, other vowel.

So like rabbit is a good example of that.

... and where the basic premise for that came from.

And maybe talk about what you think is going to have to happen in the next six to 12 to 18 months, whatever it might be, for agents to actually start to become kind of more of a household thing, right? Like every business that you walk in, it's almost like, oh, yeah, they've got agents, right? Which I'm not sure is the case today, but I'd love to hear your opinion

on that.

Yeah, I actually, so I guess the first thing to do is like define an agent, right? Because I think what a lot of people do is conflate automation that has like an LLM sprinkled in, in one step of the automation with like agents, right? So the way that I think about agents today is that they are the orchestrator of the workflow.

And so if they have a task list, they have a certain success criteria, they have a set of tools that they have available to them to go ahead and execute those tasks and meet the success criteria.

And it's the expectation that it's going to use its judgment and reasoning and intelligence to orchestrate the workflow and get it done.

Now, what's ironically happened is in fall 24, early 25, everyone was talking about agents and back when like the models weren't even that great right like the junk from three point uh gpt 3.5 to four and four oh was massive and then everyone's thinking okay like if we just like give the agent an instruction set uh give it the set of tools like it can go and do the things uh fantastic The problem with that is you go into production and all of a sudden it's extremely unreliable and you're just not going to use it, right? Because it's just not getting the job done every single time the way you expect.

And so what you start to do is make it more deterministic.

And you mentioned a couple of technologies that you're playing around with, you know, like Hermes and in some other ones, but like, where do you see like the blockers?

So I'm wondering if you can share about what your research found about like the consistency or inconsistency of the syllable division rules.

Yeah, definitely.

So, so the study that I did was actually an analysis of words, not of kids.

And so what I did was to look at like over 10,000 English words that were common in grades one through five, and I found words that had either the VCCV pattern or the VCV pattern.

And for those who don't know what I'm talking about, I mean, so, um, you know, everyone here is a literacy person, but I'll just say it for people who are less literacy, you know, as familiar, maybe don't know about these.

So anyway, so the VCCV means vowel, consonant, consonant, vowel, a word like rabbit, A-B-B-I.

And a vowel, consonant, vowel word is a word like tiger, T-I-G-E.

You're, you're in the, the poly/multisyllabic part here, so [laughs]