Dream Machines: Physical AI, Robotics, Being Human
00:45:39 · Evan Helda, Charles Wong
[00:45:39] S1: Mhmm. [00:45:39] S2: And so that was how we sort of grew that that that share of the pie over time. You know, post, you know, physical AI cabin explosion, now you've got, you know, huge proliferation of, you know, humanoid robotics, you know, robotics and factories, manufacturing, logistics. Right? And now everybody's trying to ship product, you know. Now, know, early stage startups can add... Robotic startups can actually raise enough money where they can, you know, ostensibly build manufacturing capability and do it themselves. Right? And so now the incentives are so different. And so now, you know, it's a it's a good time, I'd say, for dev tools in a space like us. [00:46:18] S1: That's great.
00:46:18 · Evan Helda, Aravind Kandiah
[00:46:18] S1: That's great. [00:46:19] S0: The main thing to look at, like, is really what is the organization trying to optimize for? Right? It's like, are they trying to optimize for reliability and performance, or are they trying to optimize for, like, you know, hey. We wanna try a whole bunch of different architectures to figure out what is the most performant one and what's the right way. So, like, there's, the product companies and then there's, like, the research companies. Right? And we're seeing a lot more product companies come up now who are, like, chasing reliability with a sense of urgency. And they're like, okay. We want to win. Therefore, we need to move as fast as possible. Right? But research is more of that. Okay. There's a whole bunch of questions we need to answer. What's the different types of data types we can try? What's all these different things that we can do? Right? So there's really different... Within robotics, each customer and each type organization has, like, very different incentive structures. [00:47:11] S1: Yeah. It's a bit interesting exercise to go through each of them and map that out by by vertical stage of the business. That's a very useful thing for any salesperson, I think, to to do and think about as they as they go on an approach. So as you approach different companies who have different incentive structures, I'm I'm experiencing this in my day job. There's... There definitely is a not built here tendency. [00:47:35] S0: Yeah.
00:47:35 · Aravind Kandiah, Evan Helda, Charles Wong
[00:47:35] S0: Yeah. [00:47:37] S1: And a lot of them, they they they do hire great talent. They they... They're raising a ton of money. They can build it. A lot of them build it because things didn't exist. But now there's all these tools like you guys coming out to do. What... When you know you're competing against internal tooling that already exists, they've maybe already built, or you hear they've already built something, what are the differentiating moats from like a tooling perspective, infrastructure, UI, UX perspective, whatever it might be that you guys are touting to really help them get off their internal development effort. And and I guess this is another way to think about, like, I'm curious to get into the weeds a bit of the product and the tech and, like, how you guys are differentiating from a lot other companies out there doing this stuff. Like, I'm seeing pale domains doing it in AV. You have applied intuition, I think, does a lot of this stuff. Honestly, I'm seeing like world models, like the world labs are starting to come after this space. And so, yeah, tell me how you think about competing, guess, is another way to ask the question. [00:48:46] S2: Yeah. And Aravind's looking at... You know, just before Aravind gets into the weeds on the features and stuff, I think [00:48:51] S0: Yeah. [00:48:51] S2: We... Some of our biggest customers have internal simulation teams, like, that we routinely talk about and go up against in some in some fashion. Right? [00:49:03] S1: And
00:49:03 · Evan Helda, Charles Wong, Aravind Kandiah
[00:49:03] S1: And [00:49:05] S2: and really... And, you know, this is almost less of a robotic specific conversation or AI specific conversation, just like product specific... Product... More broadly product principles. Right? Mhmm. [00:49:16] S0: And I [00:49:16] S2: think one thing that we really try to hold ourselves to is how do we make their lives easier? Right? Not just the robotics developer, the user, but also, you know, the economic buyer, the procurement guy, the director of autonomous, all the way up the chain. Right? How do we make this a no brainer for them compared to whatever they're about to spend on or whatever whatever they're about to do? Yep. And so which is why, like, when we compare ourselves against some, you know, simulation tooling that's out there, we we... Like, one of our positions is that you don't need to hire, you know, 30 people in your sim team to run this and build this capability for you. What Bifrost, above all, purports is that you just need to provide one AI developer, one user, and you can start generating data within five minutes. Right? And just the ease of onboarding, the ease of use from a non three d, non SIM expert is usually more than enough to get our foot in the door. Right? Once your foot is in the door, then the game begins. Right? Because now you're being attacked from all sides. There's internal politics. There's, like, new new tools coming out, and that's where it gets kind of interesting, I guess. Sure. Aravind?
00:50:28 · Aravind Kandiah, Evan Helda
[00:50:28] S0: Yeah. So most of our customers have raised, like, 1 to $2,000,000,000 already. So, you know, they also have the money to go and build it, and most of also, like, try to build it themselves. The reality is, first of all... And maybe this is a thing that may most of you don't wanna hear, but you're just gonna ship faster and better. Right? That's first of all. And two, just make it super easy to use. But most important of all, I think you need to have a very opinionated take on how it needs to be solved. And our opinionated take is... We call it, like, the autonomous simulator. Right? What we saw previously at almost every simulation company is all the scenarios had to be hand built. Everything had to be done by hand. I have to go in. I have to define what three d objects have to be. I have to define how the scenario plays out. As a result, there's this category called simulation engineering, which is you hire a whole bunch of people to manually program the simulator to do a whole bunch of stuff. And as a result, you're you're hiring, like, 30 to 10 people just to serve, like, some internal teams. And most of the time, the internal team can barely even keep up and serve the entire organization. Mhmm. Especially with, like, these larger organizations who are, like, you know, scaling... Hyperscaling, there's, like, inside, there's, like, 15 different platform teams, and this one team cannot provide the data to help these, like, 15 teams. [00:52:00] S1: Mhmm.
00:52:00 · Evan Helda, Aravind Kandiah, Charles Wong
[00:52:00] S1: Mhmm. [00:52:00] S0: So we come in and say that, hey, guys. You shouldn't be building scenarios by hand, first of all. Your data, your real world data, should drive your simulator. So it's a data driven simulator. It should inform what scenarios you build. It should inform that rails to sim gap for the photorealism. Your simulator should be driven by real world data, and that's how you also close the SIM gap, but also how you do less work. Right? And you can produce higher quality data. So this concept of the autonomous simulator is really an opinion thing. It's like, how high conviction are you on this direction? And you're willing to commit all these engineering resources to build the thing. And most teams don't have that high conviction. They just do the simple thing which say, let's just build a simulator from scratch. There's procedural generation. Use this. Use that. And it's all the... On paper seems like the right thing to do, but they're not taking the high risk back, which is like, oh my god. Okay. How do we build real world data pipelines that can define... Auto define all these things that, like, generate these scenarios at scale? [00:53:05] S2: How does that [00:53:06] S1: how does that data flow work when you are trying to let real world data define the simulator and define the synthetic data, like, from, you know, if if you're talking to an engineer that's trying to build this, like, can can you break that down for us a little bit?