Peter Deng: belief

Transcript context

…Well, let's actually spend more time there because a lot of people are thinking about this. It feels like everybody's trying to start a company these days with AI enabling so much more. And so I think a lot of people are just curious, where should they spend time? And so I think this is actually really interesting. So, what I'm hearing here is two things to think about to create any kind of moat, defensibility against, say, foundational models coming to your lunch or in other companies. What sort of data can you acquire that is proprietary and create a flywheel to generate more of that data? And then the other piece is how do you fit into a very specific, basically, vertical that you understand really well that fits into their existing workflow? Is that...? I'm probably right. Yeah. Well, it's, again, this is something we can unpack for a long time. Because with any product that you want to build, there's going to be incumbents that have distribution advantages. But I do have this thesis that there are certain- ... have distribution advantages, but I do have this thesis that there are certain products that will be able to break through those advantages of the distribution of the other companies, but you have to overcome a pretty high bar of your product has to be so much better. I think that's one thing. But yeah, I think that data flywheel thing is really interesting because the models will get really good at whatever data you show it, and that's one of the things that people just think that AI is such a magic wand. But no, it's like if it's been trained on the right data, it's going to do the thing that it's been trained on. It's very malleable. Being very mindful of the data that you have access to to start your flywheel going and what you can do to keep on going with that flywheel is going to be a critical thing for anyone who's starting a company today. Let's make that even more specific. When you talk about this, I think about... The CEO of Windsurf was on the podcast and we talked a lot about how they've all this really unique data about which recommendations of code snippets people accept and reject and they actually launched their model I think based on that. Is that example? Any other examples to make this real?…

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