Youssef Rizk: evaluation

Transcript context

…Interesting. I'm gonna move up. The question that we said was gonna be the meatiest question of this. So you chose this out of your list of ideas. And this is one of the things that a lot of AI founders are worried about, right? So the framing of this is, are you worried that you're a thin wrapper around 11 labs? What is your moat? That's a seminal question. I think, frankly, everyone in an ASR should convince themselves of this. Don't listen to me, and like, just make sure from first principles that you can derive this. But, I guess I would start by first saying, what is a mode? What is defensible? In theory, if we're just taking it, and this is trivially true, but the fact that someone built it means someone else can build it. Right, so modes tend to just be built around, like, you have a lot of network effects, or you have A really good product for this use case or, you know, something like that. And I think typically when people ask this question in the AI context, They're thinking of like, okay, you're a thin wrapper, you're an application layer thing. As opposed to you're one of the like underlying technologies or APIs that people use. Cool, I think that's fair. But I think the reality is that like, yeah, these APIs exist and they probably do serve a million different use cases. But they're not built to serve these million different use cases. So whenever you ask the question of modes, it always has to be with the perspective of who is the user I'm building this for. Right? I can use chat gpt to do half of my writing. But you know, but I don't know. Jasper claims that they do this much better for marketing. So it's tailored. I actually, you know, don't quote me on how well they're doing after chat gpt came out because they were really big before. Yeah, there's some negative data points, but…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.