Andy Halliday

So if you're designing the front end of an application and the other one, you know, has to do with React, right?

I made the language that is used in application development.

Well, those have thousands and thousands of downloads already.

I think the top one was 25,000, but it quickly drops off so that even on the first page, the number of skills downloaded by skill drops off to the hundreds and then very quickly down to the dozens.

And so if you're downloading one of these very popular ones, you can...

I think, be pretty confident that it's tested.

It's been used by a number of people and there's no blowback on it that went back to Vercel saying, hey, this was junk and should be at the top of the charts there.

Yeah.

So I'll paint a distinction that sits in my mind, Greg, which is like in a skill you can have as part of the skill that one of the first things it does is it goes to a repo and checks for the latest information that's available in that repo.

that's not something that a custom GPT could do, I don't think.

A custom GPT isn't quite that agentic.

You have to populate the custom GPT with a bunch of knowledge documents and then the instruction set.

And those are the two main components of that.

And it creates a reusable application that can do what it's focused on

repeatedly for a user.

A skill is a little more atomic in my mind, which is there's going to be a whole bunch of different skills that you might assemble over the course of time.

And the cloud version to stay within their ecosystem, the cloud code or cloud code work version can decide on its own

To look through the skills library and see if there's a skill or get the skill that's requisite to the task that it's going to have.

And there isn't, I don't think in the chat GPT world, there is a way to have the agent, like the chat GPT agent mode,

automatically review all of the custom GPTs that you have in your library and say, which one of these might be helpful here?