The Exchanges

Q Right, so how do you think, you know, that, yeah, how should people think about, should I build a B to C thing or a B to B thing?

A Um, I'll tell you the way that I think about it, um, is, um, I think B to C is often like a bigger opportunity, can scale to be really massive, but go to market and finding product market fit, and then growing the thing is really, really hard, and you have to catch lightning in a bottle, and your odds of failure are much higher, the size of success could be larger, I mean, depending on what you choose to do, you know. Um, so there you're optimizing for, like, high dynamic range, and there are a lot of things that are, like, super difficult about it. I think B to B, you know, the dynamic range is not as large. Like, you can go to a person and say, hey, tell me about your problems. Okay, cool. I'm going to build this thing that looks like this. If I do, and like does this thing for you. And can you just tell me if, if I do that, like, will you pay me for it? How much? And then you can like go and talk to a hundred other people and get their views. Then you can like build the thing that they tell you to build as long as you're good at interviewing people and sell it and make money. And it's great. It's amazing. Um, and so I think that you can build like a, It's like a tighter range, if that makes sense. Like a tighter range of outcomes. You're less likely to fall, you know, face first and just like face plant into the dirt. Um, but you're also probably not going to create like a h…

AI assessment note: “I think B to C is often like a bigger opportunity, can scale to be really massive”

Q Uh, in AI startup land and also in financial services, it's a extremely competitive space, right? You're the smartest people in the world are going after this space. How much do you as a product person think about competition? And, and like when you're, when you're building out your product roadmap, are you like mapping it to how the competitors work? How should people think about it?

A Uh, not directly, but I do think that, like, there was a point in my career when I felt really bad about copying anything. I was like, oh, no, you have to come up with everything yourself. And, like, I no longer believe that at all. Like, I think there are just certain primitives or certain ideas that become commonplace and, like, should be replicated by everyone. And you have to be, like, you know, it's, like, impossible that you're gonna come up with all of those yourself. And so, It's like a big world, you know? Um, and so I think you have to watch what all your competitors are doing and other people in adjacent spaces. And when they do something really great or something really brilliant, be like, oh, okay. Like we've missed that. Or we should have prioritized that earlier. All right, well, let's do that. And I, and the way that I think about that now is like, these are just primitives. Like if somebody had said, well, I'm not willing to copy the wheel. You know, like that would be really stupid. Like if the only person who could ever produce a wheel is like the person who originally conceived of it, we would all just be in the dark ages, you know. Um, so I think that it, you have to kind of like look around and see what all the smart people are doing and have a deep respect for competition and the noise around you. Um, while, and this is so important, and I'm curious your …

AI assessment note: “Uh, not directly, but I do think that, like, there was a point”

Q or an AI startup, like, thinking about, okay, I need to give something to this group of micro-influencers that Is going to get them to create a 32nd or 15 second clip that is, you know, going to go viral. So what is that product? And then reverse engineering that, that's been really helpful in terms of, like, creating an AI startup that gets, like, millions of views and downloads.

A Yeah, I, I think that totally works. And, you know, it, it all depends on who your audience is, like, who you're trying to reach and what you're trying to do. A bunch of them are going to be on TikTok and, like, You know, that's a great way to get them into the funnel. And you can just like watch the top 10 clips at any given time and be like, okay, I'm going to like emulate this format. I know that the algorithm likes this right now. And I know a lot of people who are really successful at doing that right now. If you want to target like techie people, you probably want to do it on Twitter X. If you want to target Sales people, you probably want to do it on LinkedIn. You know, like, it just, you know, and there's a format, and there's a way of thinking, and a way of talking, and a way of interacting, which is, like, native to those platforms, and you have to behave that way to be successful on them.

AI assessment note: “Yeah, I, I think that totally works. And, you know, it, it all depends”

Q Okay, okay, I just have to ask, I know, I know, you know, before we leave, I know you spent some time with Mark Zuckerberg. What do you, what do you think Mark Zuckerberg is thinking about with AI and this whole AI world? Is he, you know, yeah, what's going through his mind?

A I don't know. I mean, I think that he's one of the smartest, most determined, most principled people I know, you know, um, and, and deeply thoughtful. And, um, he has a quality very few people have, which is, you know, he's also kind of fearless. You know, he like, once he makes a decision about a thing, you know, he, he kind of goes all in while still being open to input on the other side. You know, he's very, he's, I guess what I'm saying is very good at what he does. Um, I think that the strategy that they have right now of like open sourcing a bunch of models, you know, makes a ton of sense. It's like the right strategic move. If you're in their position, um, you're kind of like salting the earth in a, in a pretty meaningful way, you know, giving everybody access to this thing, um, instead of, so that it's no longer a war between like two or three or four Titans. It's just like, okay, we're all in this together, you know, um, this is very smart. Um, And we'll see. I think it's going to become, I think that one of the, you know, really significant advantages that Facebook has, and again, this is one of the reasons why, um, distributing the models in this way makes so much sense is, um, you know, Facebook has one of the largest permissioned training sets in the world. Um, it's not clear whether or not open AI really can train on all of the things that they train on. Um, there…

AI assessment note: “the strategy that they have right now of like open sourcing a bunch of models”

Q handle new, new situations. So, Then I wrote down, by the way, I wasn't planning on sharing this, so these are just some rough notes. Um, so I was like, okay, how do you build a smart agent that is going to be valuable? Start with one action that matters. Add LLM understanding. Build tight feedback loops, learn from every interaction, and expand slowly. What do you think about that?

A I think that's exactly right. And I, like, I think about, okay, so these are all things that I've had to do in the last week, right? I've also had to deal with scheduling, which is really annoying. And I was really struck, I had a really amazing, like, personalized AI, like LLM experience recently, where I used, I was trying to find this investor I was talking to, who's Japanese. I remember that when I spoke with him, he was in California. And he's on the board of some Japanese financial conglomerate, but I couldn't think of the name. Um, and so it's going to be a nightmare to find in, in superhuman, but I realized that there was ask AI and I just said all of that. And it came up with the wrong answer the first time. And, uh, I said, no, no, no. The guy I'm talking about has a Japanese name. And it came up with the right answer the second time. And it was the first time that I've had like a Really personalized, like, I'm just talking out loud, trying to, like, find recall moment where, like, an AI actually really solved that problem for me. And we use AI, I'm building this thing called Sling Money right now, and we use AI for a bunch of stuff for customer service, for, for financial crime, um, compliance reasons, you know, all sorts of things. It automates repetitive tasks, which are a real pain, like going and we use this thing called, um, Greenlight to do EDD, like enhanced d…

AI assessment note: “I think that's exactly right. And I, like, I think about, okay”