Y Combinator Startup Podcast
00:23:50 · Jared Friedman, Garry Tan
[00:23:50] S2: Yeah. It's like pretty bananas. I... Do you wanna talk about anything, Garry? [00:23:53] S0: I mean, the big ones, I mean, I think AfterQuery and DataCurve, both really really great. I mean, there are probably too many to name that are honestly, like, maybe don't even wanna be mentioned because Yeah. What, you know, once you have something that's working, you almost don't want people to know. I think that it's kind of natural to understand this, though. I mean, data is one of the legs of the scaling law, and, you know, much has been made of compute. But without the data, how are you gonna make these models that much better? The RL environment thing is interesting. I mean, there's a lot there. I mean, there's a lot of, like, pure customization that's happening for specific use cases. Like, you'll you'll have, like, RRL environments for finance, for instance, and someone can go very, very infinitely deep with that. And it's like a little bit of expertise, it's a bunch of computer science, it's some systems work. But RL seems to be, I mean, one of the big engines for how... I mean, people are maybe bench... Benchmark maxing a little bit more than they should, but it costs money to do it, and it's seemingly here to stay in terms of how big model companies are gonna approach it.
00:24:58 · Diana, Jared Friedman
[00:24:58] S1: Reportedly, the big labs are spending about $1,000,000,000 on this. It's not a very known fact, but there's there's actually a real business to be built around this. Our real environment is the current flavor of it, and there's things with long term horizon tasks that are getting built up. And I think that is starting to also emerge in robotics. The labs also wanna solve the problem of getting AI to work on the physical world, so they need a lot of the environments in the real world. So things with ecocentric data, tele op, robotic tasks, starting to merge as, like, big data category where labs are spending $8.09 figure deals with these companies. We had a number of companies in the batch that work on that and been able to close close revenues in that in that space. Companies like... In the current batch of summer twenty six, there's Praxis Robotics. There's one that I'm working with that has, like, a network of places across the world where industrial manufacturing gets done. They collect data from that. There's this other company that Brad worked with called Deepreach that also has data that local entrepreneurs in... Across the world do. [00:26:11] S2: And Human Archive in Winter twenty six. Yeah. There have been a 100 of these companies recently. [00:26:16] S1: Right.
00:26:16 · Diana, Garry Tan
[00:26:16] S1: Right. [00:26:17] S0: I think, like, if I were gonna prognosticate, like, one of the things going back to the, you know, all systems of record need to be AI harnesses, they might also need to start training their own models. And that's where things like River AI or Tinker start becoming really interesting. Out of the box, like, you can sit there in Claude Code or even open... I use OpenClaw to train my own models, is very fun. It'll do its own data cleaning and everything. But to date, like, that hasn't been a huge factor, but I can see that becoming a much much bigger factor. I mean, when you have proprietary data and you can train... I mean, the opens... Open weight models are really nearly frontier. If you can, like, sort of special purpose train these things to do even better than what the frontier can do. Like, that... That's gonna be really, really powerful.
00:27:06 · Diana
[00:27:06] S1: I think this is actually gonna be even bigger in robotics. I mean, this is a hypothesis not proven yet. But robotic foundation models and robotics, I think, have a very different characteristics versus LLMs. LLM is... The whole thing is you model reality as language. And for robotics, you model reality in the physical three d space, which has way more degrees of freedom. And perhaps in order to get robots to work in a specific vertical, like, let's say, robots that do operations in data centers. I have this company called Boost Robotic that build robots for data centers, like, doing the cabling. It is possible for these robots to work. It is... It's better to get a model that's fine tuned and trained on custom data that just works in that environment. Because the the thing that's also challenging for robotics, they need to be in real time and respond very quickly to to the stimuli and have an action plan, which is different than LMs. LMs, you can have this this feature where you can just let it go and come back. But for robotics, you can't be... Because if if... I don't know. Let's say you connect that cable to to data center, and then someone comes in and, like, knocks the robot out, and things could get connected to the wrong plug, let's say.
00:28:17 · Jared Friedman, Diana
[00:28:17] S2: Yeah. My understanding is that all the YC companies that are using physical intelligence as models to deploy robotics, they're all fine tuning the PIE models. I don't think any of them are able to use the PIE models out of the box. Even though it's a great, like, starting point, you have to actually fine tune it for, like, your specific case, like, data center cables in order for it to work. [00:28:34] S1: Mhmm. You work with this company, Ultra. Right? [00:28:36] S2: Yeah. They start with the Pie model, but then they have, like, thousands of hours of footage of, like, putting things in boxes that makes it really good at putting things in boxes.
00:28:36 · Jared Friedman, Garry Tan, Diana
[00:28:36] S2: Yeah. They start with the Pie model, but then they have, like, thousands of hours of footage of, like, putting things in boxes that makes it really good at putting things in boxes. [00:28:44] S0: I've heard the argument basically that, you know, you could look at ClaudeCode, like ClaudeCode can use its its code transcripts to figure out who the top coders are, and you can take that and turn it around and, you know, basically train the next coding model to be even better. If you happen to own TikTok, you happen to have all of the data on what people watch and click on and what's compelling, and you can use that to make much more compelling videos in Seedance. So, you know, that's already been happening. I just, you know, I think that that that trend is gonna continue in a fairly spectacular way from here. So one of the things that we've been noticing, I think all of us have, is that frankly, some of the most powerful and badass founders that we've been seeing lately, they're... Might be in their late thirties, forties, even fifties. I mean, there's a sort of resurgence of the experienced founder. A lot of people seem to say that they wanna be YC for solo founders, but it turns out YC is the YC for solo founders. Diana, you have a few stats that you found surprising. [00:29:49] S1: One of the shocking stats from analyzing the septic companies from a year ago, we used to only have about 5% of the companies accepted be solid founders, and now we're over 18, 19%, which is a huge... This is the highest spike that we've seen.