Brian Maucere

with the larger codecs, that there are many, many things in coding.

And this would be for somebody who's more of a computer engineer who need those light flash type fixes that don't need the full power.

And this allows them to move at, well, a thousand tokens per second at a much faster rate

to me, this sounds very much like when I would use haiku versus when I would use sonnet or opus, you know, like you don't need all of them for all things.

In fact, I'm, I'm running into this immediately with, um, my claw that AI, which I've been sort of just playing with, right.

Like because, and, uh, one of the things I saw doing was chewing through dollars.

Like, so I started at like 20 bucks and it was like pretty quickly down to 1750.

And I was like,

In the world of tokens, $3.50 is a lot, right?

$3.50 is a lot.

And so I was like, what is happening?

And my class, I came back and said, oh, well, listen, I inject the entire context.

And it had broke it down for me to the point where it was bringing in 300,000 words every time I talked to my class.

And I was like, well, we, this is not sustainable.

This is, we cannot continue to do this.

And my clock kind of came back and said, noted, heard, you know, and this is how I think we can fix this.

So I'm just playing.

I don't mind the three dollars.

It's worth the play.

Right.