They would need to use escapes or something for actually like for any time that XML tags are actually enclosed.
Maybe we use an alternative sequence.
Instead of using escape tags for actual XML tags, we can use something else unlike them, but I mean Claude uses it anyway.
Here’s my point.
I want to make a representation of a looming session
that wasn’t too long, that wasn’t like 8,000 plus tokens,
that you could upload the entire tree, a high value one, very much pruned,
so that the entire tree, if given as context to even ChatGPT, would be very enlightening.
applications
And of course, to Claude, et cetera, you could actually generate trees that are quite complex,
If you can provide Claude with markdown looming, markdown based looming, like on daily notes,
that you have pruned well,
that has been proven to be very valuable,
and that you can contextualize as one made for another user,
like you can even do this for a fictional character, or for characters in existence, like people in existence that have accomplished things in the past, like Mark Zuckerberg would be an obvious one, right?
Like, you could loom on his notes in the past, but make them very realistic,
use Markdown and then it’s like a representation of a thought process
and you could then have Opus or Sonnet or Haiku probably even generate additional ones
like within the world sim or in some other similar kind of context
or just in any context, check out this stuff.
benefits
Like it’s just really interesting.
And I want you to, I want you to behold this information.
Like it’s, it’s something that you can not, like to a human, you couldn’t explain it without walking them through the whole session.
But to a language model, if you can get it dense, if you can get all of the things like prune it well, represents it well in a low, low token way, you can convey so much meaning, so much meaning.
That would be like tremendous.
Like you get, you get so much more out of a model.
In a way that we couldn’t even understand right now, but we definitely could.
I have no doubt whatsoever that we could do that.
If you pruned it well and if it was high value, it’s a more efficient way to prompt in a very unusual way.
It’s kind of like low-level programming versus Python.
It’s not about lines of code.
It’s about meaning you know it’s but it’s kind of the opposite i guess like like with a low level thing no it’s not it’s not the opposite actually because it’s going to take many lines but if you make those lines good you can do things that are way i don’t know it’s not a perfect analogy but you can do things that are much deeper than you can with a high level language of like pure english it’s still english but you’re just using structure to encode and there’s more behind it, even behind the actual words spoken and it’s about the connections between those words and the multiversal nature of what you’re looking at, like the way that it’s mind expansive.