The Innovation Civilization Podcast
00:26:32 · George Sivulka, Waheed Nabeel
[00:26:32] S1: Yeah. I think that there's definitely gonna be a moat on taste. There's gonna be a moat on good ideas. I don't think technology was ever a moat. Technology is not the hard part of building a business. So a lot of, like, business founders that go out and try to find their tech cofounders, oh, tech is such a hard part. So no. That is not the case. Humans are the hard part, and I think that will continue to be the case. [00:26:51] S0: Yeah. That's a good point, actually. That's a great point. So one of the questions actually, one of our listeners, Josh Godwin asked. So Josh is asking basically that you've got some goaded investors on your cap table. And there's Eric Schmidt. There's Peter Field. There's Jerry Yang. So it's fascinating to see how these tech titans are shaping the landscape. Given their diverse backgrounds, I'd be curious to know how their unique perspectives influenced the AI company's strategic direction. So what sort of impact do they have on your product road map, on your technical road map, on your vision?
00:26:51 · Waheed Nabeel, George Sivulka
[00:26:51] S0: Yeah. That's a good point, actually. That's a great point. So one of the questions actually, one of our listeners, Josh Godwin asked. So Josh is asking basically that you've got some goaded investors on your cap table. And there's Eric Schmidt. There's Peter Field. There's Jerry Yang. So it's fascinating to see how these tech titans are shaping the landscape. Given their diverse backgrounds, I'd be curious to know how their unique perspectives influenced the AI company's strategic direction. So what sort of impact do they have on your product road map, on your technical road map, on your vision? [00:27:29] S1: I think that at the end of the day, investors, especially at the stage of career as some of the folks that you've mentioned, it's less in the weeds and far more about what they represent and how they push you and how there's a dialectical tension between what you think and what they think and getting to the truth. And so there's a lot of wisdom, but I think that wisdom and maybe the push to be more contrarian, or the push to be different is actually what helps make me a better entrepreneur. But it's never a black or white, or hey, you should do this, or you shouldn't do this. It's almost always a dialogue and engaging with some of the best minds to ever do it. [00:28:05] S0: Do you have any examples that come to your mind of interactions which kinda shifted your perspective like that?
00:28:05 · Waheed Nabeel, George Sivulka
[00:28:05] S0: Do you have any examples that come to your mind of interactions which kinda shifted your perspective like that? [00:28:13] S1: Yeah. One of them was very early on, I had this idea that I would create one piece of software and then another piece of software to capture a market. And so I had to effectively do two miracles to come up with with an answer. Create two companies in order to have the market that I wanted to. And one of the folks that you mentioned sat me down and said, hey. Never build a company that requires two miracles to work. One is enough. And revise your go to market plan and your product plan so that you're not trying to do too many things or orchestrating a bunch of dominoes to fall. Keep your business as simple as possible. And the simpler that it is, the faster you'll be able to grow. Was a hard lesson. Yeah. It was good. [00:28:53] S0: So you mentioned about you're fusing LLMs and you're also using... You're basically building that multi agent kind of platform. So what sort of differentiates you from some of your competitors and especially in a world where kind of agents are getting better and better every day at just doing normal stuff, the kind of stuff that Hebiah might be doing. So generally, level of the agents when they could trip over with basic Excel formulas, now it can just do much better, basically. So what's the differentiator for Hebia here, basically, in that world?
00:28:53 · Waheed Nabeel, George Sivulka
[00:28:53] S0: So you mentioned about you're fusing LLMs and you're also using... You're basically building that multi agent kind of platform. So what sort of differentiates you from some of your competitors and especially in a world where kind of agents are getting better and better every day at just doing normal stuff, the kind of stuff that Hebiah might be doing. So generally, level of the agents when they could trip over with basic Excel formulas, now it can just do much better, basically. So what's the differentiator for Hebia here, basically, in that world? [00:29:34] S1: Yeah. I think that's a great question. And, yeah, I'd say that the the big differentiator for Hebia is the fact that we're building a very different paradigm to the operator agents or to the... What the foundation models are working on in terms of capabilities research. We're basically saying that the limitation to agentic applications is not just the ability for an AI to move a mouse around the screen or for the AI to call a tool, but really the ability for AI to, for example, retrieve information very accurately over massive amounts of documents, over unstructured and private data. And we're more of an applied AI reasoning lab towards information retrieval than just capabilities research. And as capabilities get better, actually, Hebbia improves at the same time, but we've built this fundamental substrate and orchestration and agentic layer that makes us so much better actually at at at doing tasks as the models improve.
00:29:34 · George Sivulka, Waheed Nabeel
[00:29:34] S1: Yeah. I think that's a great question. And, yeah, I'd say that the the big differentiator for Hebia is the fact that we're building a very different paradigm to the operator agents or to the... What the foundation models are working on in terms of capabilities research. We're basically saying that the limitation to agentic applications is not just the ability for an AI to move a mouse around the screen or for the AI to call a tool, but really the ability for AI to, for example, retrieve information very accurately over massive amounts of documents, over unstructured and private data. And we're more of an applied AI reasoning lab towards information retrieval than just capabilities research. And as capabilities get better, actually, Hebbia improves at the same time, but we've built this fundamental substrate and orchestration and agentic layer that makes us so much better actually at at at doing tasks as the models improve. [00:30:31] S0: Okay. That's interesting, basically. There's one interesting question, basically. What trend of development when it comes to AI do you think people are not paying enough attention to, but you think is important?
00:30:31 · Waheed Nabeel, George Sivulka
[00:30:31] S0: Okay. That's interesting, basically. There's one interesting question, basically. What trend of development when it comes to AI do you think people are not paying enough attention to, but you think is important? [00:30:44] S1: Oh, okay. That's a good one. I would say one thing would be the geography of data centers. Okay. Where data centers are is very important. I think that the Trump Sam Altman $500,000,000,000 investment project Stargate is... Was really just a bit of a nothing burger where, yeah, obviously, everyone, they're probably gonna make that investment anyways, etcetera, etcetera. But as data centers and access to power, I e, like electric power to power those data centers becomes more and more scarce, the location of data centers and the geopolitical implications of the location of those data centers will become increasingly important in terms of who controls compute and who controls effectively AGI. So that'll be interesting. And then another thing that I think is a really interesting trend is training scaling laws. So the ability to have more data and more parameters and to have a more accurate model come out are very interesting. But inference scaling laws, which is the ability to run more models and have more accuracy, is a new type of scaling law that has been powering a lot of paradigms like test time compute that was pioneered at Hebia with our matrix product. And I think that will be a massive unlock for more accurate and more able and capable AI systems.