This Week in Startups - AI Robots with Purpose with Jake Loosararian of Gecko Robotics | E1947

Episode Date: May 11, 2024

This Week in Startups is brought to you by… Vanta. Compliance and security shouldn't be a deal-breaker for startups to win new business. Vanta makes it easy for companies to get a SOC 2 report fast. TWiST listeners can get $1,000 off for a limited time at http://www.vanta.com/twist DevSquad. Most dev agencies only offer developers. Why? Because product management is hard. Get an entire product team for the cost of one US developer plus 10% off at http://devsquad.com/twist. Hubspot Podcast Network. Hubspot Podcast Network presents a new podcast called The Next Wave. The Next Wave is your personal Chief AI Officer, bringing fresh takes, industry insights and a trustworthy perspective on how to implement AI to grow your business. Just search for “The Next Wave” on YouTube or in your favorite podcast app. “The Large Language Model Race with Pete Huang, Founder of The Neuron” episode: https://www.youtube.com/watch?v=8elHTM9cOOA * Todays show: Jake Loosararian of Gecko Robotics joins Jason to discuss the purpose-designed robots being built at Gecko (3:01), the value of “bear hugging” your key customers (23:39), the “BOM” of current and future robots (1:02:14), and more! * Timestamps: (0:00) Jake Loosararian of Gecko Robotics joins Jason. (3:01) The purpose-designed robots being built at Gecko. (7:03) Origin story behind Jake’s startup. (9:45) Vanta - Get $1000 off your SOC 2 at http://www.vanta.com/twist (10:38) Show us the robots! (17:43) Details on the frequency of needed infrastructure inspections. (19:24) DevSquad - Get an entire product team for the cost of one US developer plus 10% off at http://devsquad.com/twist (23:39) The value of “bear hugging” your key customers. (27:42) Bridging physical and digital with Gecko’s Cantilever digital twins. (29:36) Hubspot for Podcast Networks - The Next Wave: https://www.youtube.com/@TheNextWavePod “The Large Language Model Race with Pete Huang, Founder of The Neuron” episode: https://www.youtube.com/watch?v=8elHTM9cOOA (33:50) Extending the value of inspections with fixed sensors. (46:47) Opening up the conversation to the AI data collected. (54:44) Why some robotics companies fail while others succeed. (1:02:14) The “BOM” of current and future robots. (1:06:29) Where humanoid robots may first be “employed”. * Subscribe to This Week in Startups on Apple: https://rb.gy/v19fcp * Check out Gecko Robotics: https://www.geckorobotics.com/ * Follow Jake: X: https://twitter.com/loosararian_j LinkedIn: https://www.linkedin.com/in/jake-loosararian-23981350/ * Follow Jason: X: https://twitter.com/Jason LinkedIn: https://www.linkedin.com/in/jasoncalacanis * Thank you to our partners: (9:45) Vanta - Get $1000 off your SOC 2 at http://www.vanta.com/twist (19:24) DevSquad - Get an entire product team for the cost of one US developer plus 10% off at http://devsquad.com/twist (29:36) Hubspot for Podcast Networks - The Next Wave: https://www.youtube.com/@TheNextWavePod “The Large Language Model Race with Pete Huang, Founder of The Neuron” episode: https://www.youtube.com/watch?v=8elHTM9cOOA * Great 2023 interviews: Steve Huffman, Brian Chesky, Aaron Levie, Sophia Amoruso, Reid Hoffman, Frank Slootman, Billy McFarland * Check out Jason’s suite of newsletters: https://substack.com/@calacanis * Follow TWiST: Substack: https://twistartups.substack.com Twitter: https://twitter.com/TWiStartups YouTube: https://www.youtube.com/thisweekin Instagram: https://www.instagram.com/thisweekinstartups TikTok: https://www.tiktok.com/@thisweekinstartups * Subscribe to the Founder University Podcast: https://www.founder.university/podcast

There's so much sex appeal to building new things. And in 10 years, we'll have this really cool new autonomous thing, drone, walking, you know, humanoid that's going to solve all of these problems. But the problem is, there's a lot of issues going on today. And so the approach to solving and using, you know, specific robots for specific jobs is actually just to earn the right to begin building really cool robots that are able to do, like, you know, more interesting things. But yeah, I get the business model right. And the business model has to incentivize and make a CEO or a CFO give a fuck about, you know, how useful, these industry 4.0 principles and tools are. Because right now that's not true.

I see this time and time again where I won't name like the AI companies, but these AI companies come in and say, well, like, completely turn on your head the way you're operating, you know, your entire business. And they'll come in for some contract that ends up expiring because it just did not produce. And that's the problem. You think you have all the information and data, but you're building your AI and your solutions off of ground truth.

It's actually not ground truth. This week in startups is brought to you by Vanta. Compliance and security shouldn't be a deal breaker for startups to win new business. Vanta makes it easy for companies to get a SOC2 report fast. Twist listeners can get $1,000 off for a limited time at vanta.com slash twist. Dev Squad. Most dev agencies only offer developers. Why?

Because product management is hard. Get an entire product team for the cost of one U.S. developer plus 10% off at devsquod.com slash twist. And HubSpot Podcast Network. HubSpot Podcast Network presents a new podcast called The Next Wave. The Next Wave is your personal chief AI officer, bringing fresh takes, industry insights, and a trustworthy perspective on how to implement AI to grow your business. Just search for The Next Wave on YouTube or in your favorite podcast app.

That's The Next Wave. All right. Welcome back to another episode of. this week in startups. We'd like to talk about innovation here, and AI's been on everybody's minds for the last two years. You know, it's just simply brilliant what AI can do, and we see it improving every week. Of course, there has been this dread of, oh, my God, what if AI plus robotics gets put

together and we have the Terminator films? The truth is, autonomous robots are coming, and they will have AI built in. You've probably seen figure or what Elon's working on with Optimus over a Tesla. This is going to change the world in my belief. And today we have a company that's been working on it for a little while. It's called Gecko Robotics. We have the CEO here. His name is Jay Lucerarian.

So tell me about the robots you're building. And for those of you not watching this week in startups on YouTube or the video version on Spotify, you can go over to YouTube, just type in this week and startups, hit the subscribe button. You'll find this video there under the videos tab. And you can actually see what we're doing. talking about here. Yeah, tell me what you're building with Gecko. Thanks having me on. I'm really excited to dive in on more my favorite topics with robotics and artificial intelligence and how it impacts the world. But started in college. I started a robotics company at a college

when I saw firsthand actually the state and how the physical world that we rely on every single day collapses and isn't always there for you. And this happened at a power plant where I got to see firsthand where power plant was having these massive shutdowns and the best way to stop it was sending a human into an endangered environment and trying to predict when these built structures, in particular this boiler, was going to fail. And the best way to do that was sending a human into a dangerous environment. The same year, I'd gone there. Someone had fallen and died doing this job, sturdy, dangerous, and typically talked about. And so I built a wall climbing robot in college to solve for the decay of critical infrastructure that we care, that we rely on so deeply

to live our lives every single day. So 11 years after that, here I am, still working on the same critical mission of protecting and helping to build new infrastructure, but more intelligently. So these are purpose-designed robots to do very specific tasks. You're not taking the approach

that Elon's taking a Tesla or the figure robot is taking of specifically a humanoid robot. These are robots that are designed for a specific function, like climbing up and inspecting a building, correct? The whole premise was it seems like we don't care that deeply or at least know that much about the built world that we rely on.

And that was the thought, you know, when I was in college, hey, we go over a bridge every day, hey, we rely on power plants, we rely on manufacturing facilities, ships to carry supplies all around the world. How do we know if those things are going to be around or there for us? Is it the right assumption to believe that the bridge I'm crossing is, you know, it's going to be structurally sound and not going to collapse? So back where you started the journey, you said, hey, infrastructure is the ideal customer profile for your startup and for this product robotics.

Your customer is essentially infrastructure and specifically infrastructure in the United States, which, for whatever reason, we seem to have not allocated enough resources towards. Yeah, it's, you know, in 2013 when I was in college designing the first robot, you know, I've read this report. It's a 3.34% of GDP around the world was spent on fighting corrosion. I was like, wow, that's a crazy $3.5 trillion number. I wonder why that is. Then you look into these interesting reports that show the U.S. is that like a D-grade in terms of its infrastructure, and it costs $2,000 just to keep it there and not to just not to improve it,

but just to keep it there. Maintain it. Yeah, maintain it. And that's, you know, regardless of building new things. So it started with the critical industries and infrastructure, But it was mostly this thought that was, wow, we seemed like we seem like we talk as if we know a lot and have a lot of data about how the built world works and how to make it better. But that's actually very far from true for the physical world, for example. We don't know if a concrete structure or like a bridge is going to be sound and going to be there and how long will it last.

You know, the bridges and infrastructure that we rely on was not built for the kind of like traffic and loads that we currently are demanding on today. So we're stressing the infrastructure on top of it. It was built for, you know, the Golden Gate Bridge was built at a time when a certain amount of vehicles would go over it, a certain amount of weight of those vehicles. And obviously, you know, we've induced a lot more usage of that with a lot heavier vehicles. So maybe you could show us in SportsCast one of these robots doing inspections. And I know that you're not just doing infrastructure. You've got energy defense, manufacturing other robots and other verticals you're flying in. I would love to see what these robots are and then get into the business model,

because it is this week in startups, of how you make money with these robots. Yeah, absolutely. Yeah, when I was in college, looked around, and like I was describing, there seemed to be like this world that, you know, technologists and startups, like didn't really pay that much attention to. It's the world of energy. It's the world of manufacturing. It's the world of defense and public infrastructure.

And, you know, I saw this. like up close and personally with the power sector. And it was just this idea of, man, we don't really have that much data on the built world. And thus, it makes it really hard to know and understand how to predict how it's going to perform. And what you're showing on the screen here is the Golden Gate Bridge, I assume a nuclear reactor, and then it looks like a really either another type of bridge

and inspectors literally repelling up and down them, which is dangerous and I'm sure quite expensive. know what those individuals get paid, but they're getting paid half as much as they should. What does a person get paid to repel off of a nuclear power plant or the Golden Gate Bridge? Where do they make, $100 bucks an hour, $50 an hour, do you know? You must know. Yeah, it's about, it depends on the level, but it's about in between like $30 and $70 an hour. That's it?

Oh, yeah, it's probably, it's even a little lower. $30,000 a year, just times about $2,000. That's competitive. I mean, yeah, over time, you can get a little higher. But yeah, it's exactly right. This is a spherical tank, for example, at an oil and gas refinery. But just like this, you enter like this world, you know, like most, most like folks who are starting technology companies or in robotics or AI, like, have never stepped foot

at a refinery or don't really know the first thing about like structural and material science or what's, what are the hundreds of different types of corrosion instead of steels or instead of concretes. But these all are like really important, not just like to predict and, and, and, you know, and ensure that we're not suffering from some sort of catastrophic failure, which actually has environmental as well as just functional implications, but also how do you actually modulate how you're operating the infrastructure to actually get more out of the, let's say, the power plant or the refinery,

while also reducing the amount of greenhouse emissions that are being released by the company, because whenever there's a catastrophic failure of a pipeline, guess what? a lot of like explosion leads to unfiltered carbon emitted right into the atmosphere. And like the worst environmental, you know, recorded environmental incident was the Nord Stream pipe exploding, for example, or, you know, these deep water horizon events. So, you know, ensuring that you don't have these kinds of like failure is actually really important as you relate to the election at zero and those things. But yeah, so the story was 10 years ago in college and basically came across just a weird

problem of power plants having these shutdowns and someone had died. Listen, a strong sales team can make all the difference for a B2B startup. But if you're going to hire sharks, you need to let them hunt and you can't slow them down with compliance hurdles like SOC2. What is SOC2? Well, any company that stores customer data in the cloud needs to be SOC2 compliant. If you don't have your sock too tight, your sales team can't close major deals. It's that simple.

But thankfully, Vanta makes it really easy to get and renew your SOC2 compliance. average, Vanta customers are compliant in just two to four weeks. Without Vanta, it takes three to five months. Vanta can save you hundreds of hours of work and up to 85% on compliance costs. And Vanta does more than just SOC2. They also automate up to 90% compliance for GDPR, HIPAA, and more. So here's your call to action. Stop slowing your sales team down and use Vanta. Get $1,000 off at vanta.com slash twist. That's vanta.com slash twist for $1,000 off your sock two. A charming backstory if you ever was one. Show us the robot. Okay, all right.

Hey, show us the robot. We want to see this thing in action. Okay. You know, every founder's got a charming story. Do you have a PR team that helped you craft that or that's the authentic story? This is me. No, this is me. The authentic story, okay.

Yeah, so the first robot was one that was climbing up the wall and gathering information, visual and ultrasonic. Basically what we're looking at, what I was looking at these, you know, 10 years ago was, what is the structural integrity of the pressure vessel? and how do you ensure that you're understanding, just like you're doing a cat scan or doing a sonogram, you use high-frequency sound wave to look inside of a material

needing to open it up and destroy the story. So what we're seeing on the screen is a robot that's about the size of a pool cleaner with a tether, and it's zipping up and down some pipes. And it's open. So you can see all the innards of it. It looks like an insect crawling along the pipes.

That's the size of a pool cleaner. Like I said, Am I about right? The robots, yeah, it's a seven size of a briefcase. I'm going to come to a couple different forms, but basically adhering. Yeah, because it's upside down and it's gripping. Is it suction cups, magnets?

What is it doing there? So it's climbing up surfaces, whether it be an outside of a ship or, let's say, some sort of like piping or a dam even. We'll use neodymium rare earth magnets arranged in a hallmark array. and that maximizes pull force into a surface to allow for payloads to be added onto the robots. And they're collecting different kinds of data layers.

One of the data layers, for example, is this ultrasonic data layer that's looking at what's the structural integrity, corrosion, erosion of the surface to get generalized idea. What is the health of this? Just like you would do like a picture of a belly using a sonogram test for pregnancy. Wow. So do humans do this when they're climbing up and down, we saw them repelling? Do they have some device that they do this manually with? Yeah, they do.

So basically our savior today is Joe, Joe and a rope. But basically, it's these guys. These guys are our best defense. The guys who are hanging off of ropes or climbing on scaffolding or on JLGs. And they're armed with single probes that you use some gel. You squirt the gel on a surface, let's say, on kilomeres of pipeline. You'd squirt gel every 10 meters, every one meter depending on the criticality. And then you use the ultrasonic sensor.

and you record the waveform. And then because you know, if you understand the speed of sound through that material, you can actually understand what's the thickness of that material. And then you could use, and then you record that down on a piece of paper or in an Excel sheet. And basically, that's the way that we understand how the fieldwork works. They're taking a sample. But you're taking continuous. So you have the full picture.

It's possible, in fact, probable that the humans are going to miss most. most issues. Am I correct that they're going to miss most or some? They're going to miss a fair amount or there's actually human errors that relates to interpreting the waveforms. But there's other kinds of techniques that you either are or are not using. So visual is one. Just like, hey, this thing looks like it's leaking. That's bad. Or with around like welds, like you have to do certifications of welds on critical pipelines, for example, and you're using x-rays. Interpreting the x-rays actually pretty difficult. and it's also super dangerous because using something that can cause cancer

if you're not appropriately operating it. So we actually will put on the robots something called phased array, which is basically ultrasound for just like hundreds of different soundways going into like a very small area. You can apply basically these different payloads onto the robots to look at erosion, look at cracking, look at generalized erosion, but then you can also add other kinds of information. You can use electromagnetics to look at what's the,

But what's the damage over top of some substrates that you have to remove some sort of insulation? So anyway, what you're trying to solve for the customer is how do you reduce the downtime or the time I'm spending not making my product? And so that's what you're trying to first help the customer understand is how do you ensure that you're solving this problem of ensuring that there's not going to be so catastrophic event, but limiting the amount of time you're not making your product. So the robots are going into these missile silos, for example, or on top of flight decks on destroyers. It's climbing inside of power plants at boilers. It's going on to dams using suction and adhesion there. but basically we've gone from what you just saw in terms of the robot climbing up a wall looking at corrosion and erosion. And we would now combine that into a bunch of different robots, some of which are doing this climbing,

some of which are using drones that are looking at using photogrammetry to understand what is, like, in general, let me do a quick analysis of potential damaged areas over large geographical area or maybe integrating like a walking dog. or and then you can use fixed sensors to continually monitor. You know what this reminds me of? This reminds me of the Pro Novo, a full body scan, which a lot of doctors will say, hey, you don't need it. It's going to cause you to find things, nodules, little things, groats in your body. You're not going to know what they are and you might panic and get anxiety.

And I'm like, well, wait, but what if it is something and you live longer because you found, you know, some, God forbid cancer or tumor early or something? something with your brain. I would much rather have that. Therefore, you are going to inspect these things and have an image in time, and then you can look for the deltas and what's changed between the two imaging. So if you were to do this every year on the Golden Gate Bridge, what would be the frequency that the Golden Gate Bridge or, you know, a submarine should have this done to it? So we're actually working on, so I'm in Pittsburgh, Pennsylvania right now, which is where I, you know, so I started the company, did three and a half years of bootstrapping it, down to like a hundred bucks mega count, ended up choosing to go to YC opposed to an acquisition offer, went out to California, against all of esters like Desires, came back to Pittsburgh, close to customers, was able to grind closer there. In Pittsburgh, though, it's interesting, you know, there's so many bridges. It was where we, you know, 6, 9% of the world steel was like, was built here.

and now it's kind of reinventing itself in terms of like this robotics and AI hub. But what's exciting is actually, it's actually a really great state as it relates to the political support to try and utilize technologies like GECOs to do things like create the most sophisticated bridge evaluation

and infrastructure process. There's like, you know, we're so working with the governor actually on an initiative with bridges. But to answer your question, you want to be able to look at... How often do you got to inspect a bridge? You want to be able to look at

a bridge, you would want to look at it with a deep scan, like, we would do like a full health, like, here's exactly what's going on with the entire bridge. Maybe like one, three, five years. You don't want to, like, look at it every year. You actually, though, like, once you understand the general health, you know, similar to how you would do, like, with a human, you would, you would then use fixed sensors that are enabled by Wi-Fi or 5G. And then those are constantly updating a digital twin.

and that actually like this explain what a digital twin is for people so digital twin is it's represented in software it's three dimensional you can manipulate it but it needs to update itself so it needs to be continually updating with information

whether information is the health of the asset or how the asset is performing so an example for a bridge might be a real world example you've come across might be yes a real world example is a tank so a tank at let's say a pulp and paper manufacturing a place, the place where we all got our toilet paper. So we have a really big contract with this

company that's interested in extending the useful life of their tanks. But what they want to do is instead of, you know, the tank is 20 years old, you have to, you know, it's passage useful life, so we have to build a new one. And we come in and say, actually, you don't need to. We'll take, you know, this tank plus 50 to other tanks that look similar to this. And we'll tell you how to make it last 10 years longer or even 20 years longer. We'll tell you exactly what to repair. And Jason, we're actually, because we now have this information on this health and structure, structural integrity of the world's, some of the world's critical infrastructure, like 500,000 assets.

That's where we use AI. Hey, take a moment. Picture the ultimate all-star team for your startup. You got that image? Maybe you're thinking about the Avengers, huh? Maybe the X-Men? Wolverine, Cyclops, Iron Man. Well, let's be realistic here.

Until you've raised your Series A or B, you're going to need help building this A team, right? You may not have all the resources to do it. Finding talent, managing all these timelines, and ensuring quality is a huge task for anybody building a product. And scouting for the perfect all-in-one 10x developer, it's going to take a lot of time. Well, there's somebody out there who does this for a living. Skip the hassle and power up with DevSquod. Okay? They're like Professor X from the X-Men. They just know where all those great mutant developers are, those really powerful ones. DevSquod offers a complete product team loaded with top tier talent straight from Latin America. Same time zone as us, right? Your team is going to have

six full stack developers, depending on what you need, a technical product manager, keep everything tight, and a bunch of specialists. Maybe you need UI and UX design. Maybe you need DevOps. How about quality assurance? They're all going to be synced up in your time zone and it's going to cost 75% less than the equivalent team in the USA. We all know there's an incredible arbitrage. There's talent everywhere in the world. And now that we all work online, remote, hybrid, whatever it is, you know that you can integrate this amazing dev squad team into your team. If you're looking for seamless collaboration and the flexibility of monthly payments, without long, long-term commitments.

Well, it's time to take action. Head over to devsquad.com slash twist and snag an exclusive 10% off your engagement. That's dev squad, d-E-V-S-Q-U-A-D dot com slash twist. It's time to squad up and transform your startup with the all-star team you deserve.

So let's take a look at a digital twin then. I'm assuming you can show me one or? Yeah, because this is fascinating. You inspect this container, right? And let's say it's got, you didn't give the exact example of what you would continuously modern the digital swim, but I'm assuming maybe there's some area where you think it might get fractured

or be compromised. And so you put a sensor on that permanently. Yeah. That sends a continuous, reading to let you know if it's getting worse. And at one point, you think it's going to explode or crack or fail, which would be the equivalent of like in a human body, just monitoring some, you know, tumor that might be benign or might not be benign. Am I correct? Am I that framing here? That's right, but you also, okay, so this oriented towards a business model, too. It's like, you know, we started building robots and they're really exciting and cool. What we ended up finding was that just building robots and using a robot as a service, we don't actually sell the robots.

We're going out to site with the robots and getting data and then giving it to customer. Got it. What we ended up finding was that that wasn't a model that actually oriented towards value creation. So we were creating outsized returns in some cases like nine and even, like there was even a case of a half a billion value creation because we stopped this like crazy explosion at this refinery, the biggest refinery in the U.S. because they had gotten bolted information from the guy in the rope Joe. What we ended up doing was, you know, charging them like a couple hundred thousand bucks for this. And that was crazy because how outsized the value creation was.

So what we ended up also finding was there was the lack of an ability to action on the data to improve how the customer was operating their assets. So let me walk you through that. So what we'll do is we'll have a suite of different robotics that we offer and work with the customer to try and solve for a problem. In this case, it was how do I manage 50 of my most important pressure vessels and tanks for this customer? And about five years ago, we started developing Canaleaver, which is our enterprise software. So when a customer buys Gecko, they're buying enterprise software, and that's called Canaleaver. And what they're getting from that is a solution oriented towards a very specific problem for that customer set. So we actually not only had to become experts on robotics and AI and software,

we also had experts on like our customer's actual problem, both upstream and downstream. I'll take you through an example of a... Which I'm assuming you got by asking them questions in customer interviews and saying, well, what are you going to do with this data? We've now given you the data and they told you, hey, well, we need to make this decision, when to retire this tank. And then it became...

Your business becomes not selling a robot. or selling an inspection, your business is now extending the life of tanks. That is one of the important value outcomes, yes. But it also was, like, in the beginning, like I had to, like, spend all my days and time at the customer sites and just, like, living and understanding their problems, like, better than they could. And did that for not just power plants, but, you know, manufacturing facilities, like, places that are making steel or places that are making aluminum, places that are refining

oil, that are operating a hydroelectric dam. Like, we have to end up going into these industries and understanding exactly what they're trying to produce. And from our first principles, what goes into both the, like, good and poor outcomes. And then also understanding where they get, where they are getting value and can full value, like, from a regulation standpoint. It gets really complicated. So for this customer, we call this technique, by the way, in the business, a bear hugging. So when you have a customer who's like a key customer, you give them that big bear hug, which means you get off. location, you spend time with them.

I learned this from a company where investors in colddensity.io that does people counting. And when you are on site, you will overhear things, you know, and you're going to have the customer, just through the course of hanging out with them, give you insights that you're not going to get in a 20 minute customer interview. You might get them, but in all likelihood, just hanging out at the facility or maybe having even a drink or having lunch with folks, at some point, you're going to have these epiphany moments, yeah?

Then that's what happened for you? Yeah, 100%, but also like you can codify that into a business model. So, you know, one of our, one of our series A investor was Founders Fund. Oh, wow. Trace Stevens is our, as our partner and board member there. But it was interesting is while I was, we went, we took a trip to the UAE in like 2020. It was literally right before COVID. We almost got stuck actually in, in Oman, I think it was.

We ended up just, he ended up just talking through Palantir in the early days and how, you know, he's helped set up the Pellenteer office in the Middle East. We ended up talking deeply about forward-deployed engineering. And I was like, wow, this is so cool that they have taken this approach, forward-deploying as you send out your engineers on deployments as an implementation team of the software you sold. And then you work alongside customers to understand their problems to help create the software modules that are oriented towards the solutions that the customer is

actually trying to solve. Because in reality, when you deal with these industries, They're so complex. These problems are so complex, and they are so hesitant to either communicate or even to talk about the different problems that exist in these, like, you know, Manhattan-sized environments, like the size of refineries, the size of Manhattan. And so these are, then there's, like, so many different things, like, you know, variable frequency drives that, like, need to be, like, you know, looked at and wrench turned in this way

and all these, like, nuances. And this is actually one of the big issues as well is that there's, you know, these people that rely on every single day are completely, they're reaching this point of phasing out, whether they're dying or retiring, and there's a huge knowledge gap. So anyway, they began to think about what if actually we took the early learnings of forward-deploying our roboticists in combination in concert with forward-deploy, software engineers to actually build a vertically integrated stack of data collection of various types and a lot of it.

So we call them data layers and then pull all that into a single source of truth, a data warehouse, and then deliver the modules and software to self-customer problems, but do so located actually alongside customers because, you know, you have to convert someone using a different system. You actually have to help build it alongside of them. Yeah. And the current system was probably pencil and paper, pictures, and stuff scattered across to spread systems, I'm assuming, yeah?

That's right. And inconsistent, like, you know, per site. So, like, Marathon, you know, they might have seven refineries and each of those refineries operates completely differently because they're both producing, you know, $20 billion each or something like that. You didn't show us the actual digital twin. Let's get it. Make sure we show that, yeah.

Of course. You got sidetracked there. Yeah. No, it's so fascinating what you're doing. It's easy in an interview like this to get sidetracked into all the different nuggets of what you're discovering as a founder. But I did want to see the digital twin concept. Let's do it. Yeah. So you start with like, okay, what problem you're trying to solve?

Well, we're trying to solve for, you know, increasing life extension or understanding like how to fix like 50 tanks and manage 50 tanks. All right. Sounds good. So the outcomes we were able to do this solve where I'll just skip that to the end. Basically, customer will say like, okay, customer, I need, I need your metadata as it relates to the structures that we're going to go out and try to evaluate. So they'll send us to the metadata and then we'll incorporate that instead of C candle lever as we build out their profile. And so we're delivering using distrawlings, you know, what is a very rudimentary digital twin. And so this is an example of a 3D representation of a tank using the dimensions of the customer. Then you send out your robot fleet. And so the robots go out there and they're climbing all over these structures and they're trying to evaluate what is the health of this tank.

And doing so as quickly as possible, while the tank is actually an operation, so you know, to shut the thing down. And then you understand what the health of that structure is. this one was pretty bad. Red is good. For example, green is bad. It literally pixelated because as the robot's climbing, it's pulsing the area it's climbing over 100 times every single inch. And then you can either look at what's the mean in terms of how structurally sound

or how healthy that area is, like an inch by inch grid, or you can look at the data in other ways. But you want to label all that data set because it'll be very helpful as you relates to what kind of corrosion is there. for whatever reason here when you're looking at this tank shell, the bottom 25% of the tank is green. Yeah, so the bottom is actually super healthy is what you're seeing. Oh, green is healthy. Red is not healthy.

Yeah. All right. AI is moving really fast. And we're seeing it in all aspects of our lives in business, personally. And you know what? It's okay to be confused by all of this. I found a great podcast for you. That's perfect at helping you navigate all things AI.

It's called The Next Wave. and it's brought to you by the HubSpot Podcast Network. The host Matt Wolf and Nathan Lans both have an insane level of AI expertise because they both founded and invested in AI companies. So think of the next wave as your personal chief AI officer. That's right. Serving up fresh perspectives, industry insights, tools, tips, and tricks on how to use AI to grow your business.

That's the most important part, tactical advice to actually put AI to use in your business. Check out the episode, the large language. model race with Peter Wang, founder of The Noron. Pete covers the latest AI development's trends and research, including a fresh take on where AI and your creativity collide. Its impact on industries like music and entrepreneurship, the balance between innovation and ethics and the rising capabilities of LLMs. So here's your call to action, super easy. I want you to go search for the next wave on YouTube or your favorite podcasting app and subscribe and discover how to leverage AI to unlock new opportunities for your business, brought to you by the HubSpot YouTube Network.

So the base is healthy, but the top 75% for some reason is really weathered. I don't know if that's gross by wind or the ground is painted or covered. What would cause that weird pattern? What we found is like it's very correlated with how you operate the tank. And so this is a really important tank. And they were typically filling their tank halfway up. And this is like a very common method, you know, in fuel storage tanks, for example. You're filling a fuel storage tanks if you're Exxon like 40% of the way up because you just don't want everything to like collapse and you expug, like a bunch of oil into a river somewhere, which is actually not good because you're actually destroying the asset in a certain way.

And you're reducing the capacity you can run at and how much oil you can store. So they were scared of a disaster. So they made a decision that increased the likelihood of a disaster. So good intent, bad outcome. That's right. But also, you know, for one of our customers, Adnock, which is the UAE's national oil company, and we just send a $30 million deal with them. But the big thing they're trying to solve for is how do I produce more oil per day, barrels per day,

go from $3.5 million to $5 million barrels per day. But what we're trying to also show is that you can reduce the potential carbon emissions while also increasing your throughput if you just operate your assets more intelligently. So we'll get into exactly what that means. But you don't have to actually build new things necessarily, which is a huge deal. Then you send in a robot that can do evaluations, again, while the tank is in operations in a submerged way. And so, again, you're looking at what's the structural integrity of the floor, because the floor is actually, you know, one of the most compromised areas. And so the green, again, is good, the red is bad, and you want to try and evaluate where to fix things.

And then you use LIDAR to look at what's the depression of the tank. because it's really heavy. And so it'll begin to depress in certain locations, which can also lead to a bunch of issues. So LIDAR is really important, and there's different kinds of rules and regulations that different bodies set out. And then we'll create these repair plans.

And so what that's trying to do is help the customer understand how much capital deploy, and then like how many years you get from the life extension. So what we'll do is we'll use the data that we're collecting from this tank, the other 50 tanks at that site, and then also the thousands of other assets that look just like this to try and evaluate, where is the areas that you want to fix right now to extend the useful life for the asset. But then once you do that, you have to send it back to the real world.

So it goes physical, digital, physical. So it has to be an output where there's an action being taken from the insights, and the digital twin is actually helping to be used by folks that are welding and doing repairs, for example, or trying to make an actual functional decision around how to operate this. asset in this case the tank. But then we'll install fixed sensors that are pinging the digital twin every day. These are those black little circles around the compromised area. One of the rings of this, you think of it like a barrel, is compromise.

So you're putting sensors on it that tell you what? It'll look at the structural integrity. So it'll ping you every day because corrosion actually is, funny enough, doesn't happen linearly. It happens typically in these like weird moments of large decay over short years of time. Typically, that's related to like what kind of chemical is in the, let's say, into the tank that may be out abnormal or maybe there's some sort of, you know, it's really windy and rainy like that month or there's a lot of sodium in the air and that's like causing

a lot of like increased corrosion as you're like by a gulf, for example. It also will also- Like people say, how did it happen slowly than all at once? It's slowly degrading and then some event happens and all the once it gets compromised is generally aviation, bankruptcy and structural failures all seem to go in that direction. Suddenly, slowly then all at once. That's exactly right. And there are some root causes that you can begin to understand. So you might get a shipment of like new fuel or a new chemical.

And that, there might be actually a compromise in the quality of that. Yeah. This is actually an issue as well with you manufacture new things. Windmills are falling over in Germany right now because of four steel quality. So, like, you know, you end up having this, like, issue where you might be getting, like, really inefficient process for some reason. But you can actually tell when that inefficiency is occurring or when there's, like, imbalance of chemicals in your, you know, in your processing batch that you want to be able to react to. And that's, like, something that's not predictable. It's reactive.

People started coming to you saying, hey, we were installing this new thing. we want to have a day one inspection, so we have a benchmark. And so if it was installed improperly, we can, you know, before we make the final payment to the construction group, we want you to do the inspection of the work done. Has that started to happen? It has. And actually has happened with the $132 billion Columbia class nuclear sub program, as well as other sub-work. So we're working with the Navy actually on.

new builds and manufacturing. And so what's happening is... So inspecting them while they're in Dryda. So inspecting them while they're being built, actually. So what ended up happening was... So on the government side, the defense side, schedule adherence is like a really big problem, as well as like if you're paying $132 billion for new subs,

12 new ones, you want to make sure your, you know, your tech dollars are actually being used for building good things with high integrity. So we're actually building out digital twins of the sub as it's being constructed and looking at the quality of welds. because that just recently caused like a six-month delay in a process where they had to take sections of this other part. So anyway, this is like priority one right now on the manufacturing of the subside. And then we also do some work with the Navy. We just got a pretty large.

You have a digital twin of the sub? You have a digital twin of the battleship? Not that I can show you, but we're doing this as well with actually, it's an interesting program. I can show you real quick. but we're beginning to use the same kind of technology as it relates to concrete. So what we're doing for the US Air Force is we're sending our robot, a different form factor robot with leveraging the stack that we developed to climb up nuclear missile silos. So there's 450 in the U.S. And what's going on right now with the Sentinel program is about $125 billion billion are being deployed to upgrade the Cold War era nuclear deterrentices.

system of these ICBMs in silo. But what's happening is the concrete's been into decay and crumble. It's actually causing these. In Oklahoma, there was this oxidation explosion, actually, one of the ICBM missile silos, which you don't want oxidation explosions inside of a nuclear chamber. So generally speaking, explosions plus nuclear weapons, not a good combination. None of your combination. So we got sole sourced on work.

It'll be about $250 million project. project. But what you're looking at is we want to understand there's a steel liner and then a concrete lineer, like five feet of concrete. You want to understand what's the structural integrity of the concrete and like where all the damage, where all the issues are occurring. And then what's a structural integrity of the steel. And then you can figure out, okay, now I can create a plan to fix all this stuff. Amazing. So anyway, there's, there's like these interesting applications. There's like new builds history. And you're, why not? Yeah. No, it's incredible. And I know you have. the deck, the destroyer as well. And I guess looking at the hull of that, is all of this going to culminate in permanent robotics and permanent sensors being put onto these things? Or is that cost prohibitive in some way or just too bulky and too much maintenance in and of itself? Because based on what you're learning, maybe the sensors should just be bulkened everything in the same

way. I know this is like a minor analogy here, but, you know, like air tags, the act of finding your stuff is going to be built into other devices. I think like the Apple remote controls have air tags built into them essentially. And so it does seem to me that based on all you're learning and the stuff, man, they should just be putting these sensors in a lot of different places permanently or have these robots permanently installed because the robots currently have an inspector working with them, correct? They have to be supervised. They're not like, we're not at the point where like these robots just exist in a little cabin and go up every week and inspect and go back like a droid in Star Wars, right? We're not at that level yet. No, and I don't know if you need to be, actually. You want to do

what you said, which is once you build something, you want to understand what's the health of that structure. So, Jayquil, what I believe is in the next five to seven years, like, you won't be able to build new things without personishing the health of that asset on its construction. And you create a digital twin that is able to be updated as well with sensors that you build into these structures, especially structures that are really important, like the nuclear sub. And what you want to do is instead of the sub, saying it's going to last 40 years, it's going to last 50 years, that actually can be doubled. So you want to be able to, you know, you want to make sure that you can create something that can be updated every time that you're doing some turnaround. And then eventually, maybe even don't

need to spend 18 months in Dry Dock to do an evaluation of the health of the structure, which is currently the state of a lot of our Navy. So a third of our Navy is currently in Dry Dock, trying to do its maintenance cycles, which means that a third of our Navy is not out there patrolling the seas and ensuring that conflict is being deterred. So it's actually a pretty large problem that Secretary of Navy, Deltour and I have talked about a lot, actually, is just the schedule adherence and also understanding what is the state of the structures as we build them, and how do we ensure that we're having, incorporating these living models of these glass sets? You could cut that dry dock time in all these cases down by 50.

percent, you think ultimately, 90 percent? Well, the goal should be actually, like, don't spend any time if you can help it. Oh, continuous monitoring. So, if you're inspecting a battleship or a submarine, you could have it at the surface. It was a submarine. Obviously, the ship is already at the surface. You could have underwater robots inspecting the hull while it's out of the ocean. Yeah, you can, I mean, there's like a 2% gain in efficiency.

If you can, like, scrub a hull while you're like, while a ship is going from like one place to the other, just because of, like, barnacle and build up on the whole of a ship. There's a bunch of things you can do to improve the efficiencies of existing critical infrastructure. But I think the big thing we should be orienting to is, like, how do you not be so reactive and incentivize a model, which is currently incentivized for time of materials, right?

So whenever you're doing, whenever you have like these large maintenance primes, they're incentivized to have the maintenance cycle lasts as long as it can possibly last. Right. Show me an incentive. I'll show you the outcome. Exactly. The longer it's in DryDoc, the more the meter is running, you are a big threat to maintenance companies because you'll tell them you only need to do maintenance on this 20%, the other 80% is fine. Well, I think it's not a threat.

It's like it's orienting the outcome towards like improved performance. And so it's a how do you actually, you know, it's just like power by the hour was like the old Rolls-Royce model. It's like how do you get paid for the amount of uptime you're producing? That should be the orientation. It should be like, you know. That's what should be the incentive is how do you keep this thing in service, not how much service do you do? And that's really hard to do because then you have an incentive the other way.

Hey, we got to keep this thing flying and maybe you put something up there in the air that shouldn't be flying and should be in dry talk and should be inspection. And what you're trying to do is get to the truth. And the truth shall make you free. If you actually have the truth, you don't need to. If you can get to ground truth here, first principles, you're going to not have to try to game an incentive. But also, yeah, exactly.

But when you build new things, like think about it, this is the way, what I get excited about. When you build new things, you know, you want to be able to learn from the experience of the, you know, billions of iterations of that thing being, you know, in use every single day. We don't do that right now. We don't, you know, we can model as much as we want about how to build the best sub or how to build the best destroy or how to build the best refinery or new hydrogen conversion, you know, power plant. but we haven't learned from what's the impact of the equipment in operations and use. That's what we have to figure out because you can't build new infrastructure unless you're learning from how the old ones are working. And this is why it's so important to, we get, there's so much sex appeal to building new

things. And like in 10 years, we'll have this really cool new autonomous thing, drone, walking, you know, humanoid that's going to solve all of these problems. But the problem is, there's a lot of issues going on today. And so the approach to solving and using, you know, specific robots for specific jobs is actually just to earn the right to begin building really cool robots that are, you know, that are able to do, like, you know, more interesting things. But you got the business model right. And the business model has to incentivize and, and make a CEO or a CFO give a fuck about, you know, how useful these industry, you know, 4.0 principles and tools are. Because right now that's not true.

And you can hire, I see this time and time again. end where like, I won't name like the AI companies, but like these AI companies come in and say, well, like, you know, completely turn on your head the way you're operating, you know, your entire business. And they'll come in for some contract that ends up expiring because it just did not produce. And that's the problem. It's like you, you think you have all the information and data, but you're building your AI and your solutions off of ground truth. It's actually not ground truth. There's actually a low amount of integrity if you're not, if you're not interrogating the data all the way to the ground level.

And so for us, like, we are building AI and software, but off of like data sets that robots and smart sensors are actually collecting, in order to affect some large business outcome, EBITDA and cash flow is what we orient to, or it could be scheduled adherence or it can be environmental impacts. But you have to be to interrogate the impact from the solutions all the way down to, like, what's actually causing the change. And for us, it's very clear, like, if you can start with a core

of what is the health of everything, of my built structures. Then what we've actually found is that we don't have to ask our customers for datasets, they'll give it to us. And so the end of that case study I was going to show you was those 50 tanks, we actually were able to extend the useful life of that one tank by 10 years and scrap an $8 million capax expense. And we were able to, across the 50 assets, the site said, and did an analysis that predicted because of a modulation.

in Phil Heights, we were able to actually impact gross margin by about 4%. And so there's like these, it has to be oriented towards the outcome. So the customer has to buy that outcome. They can't buy the robots. And we'll be happy to use and integrate. And we do other sorts of robots because I don't want to build all the robots. But it has to be again oriented towards the big outcome and problem for the customer. Otherwise, like, you know, it's not going to get funded.

Yeah. I mean, there's there's a use for. doing deep tech, and there's a use for just trying to make things work in the world, but at a certain point to have to solve a problem. And I think that's, I think, what you learned was, you know, the robot was one way to solve the problem. But the sensors is another way to do it, right? And those sensors being on there. And yeah, wow, it was incredible progress you've made. Let's talk a little bit about AI. We'll open the aperture here. As you collect all this data and over the next 10 years, you'll have systems fail. You'll have Ving's.

you got right. You'll have things you got wrong. You know, weird things will happen. Random things will occur. Ships will lose power and run into bridges. All kinds of events are going to occur. And you're going to be collecting all this data about these things. And then AI will be able to process all that and maybe give us some insights. When do you think you'll start having insights powered by AI a human wouldn't have gotten to in a reasonable amount of time? And what do you think the insights might look like? What might you figure out? collecting all this data and then, you know, running algorithms, machine learning, other things against you.

We've collected now and own data sets on the health and structural integrity of over 500,000 other worlds most critical assets. And what we're doing is we're capturing this immense amount of information as it relates to what is going on as you relates to why do things, where are things damage, why are they, what kind of damage mechanism is occurring. And also building out machine learning to interpret what is a, a, sound wave attenuation indicative of what kind of like issues. And so we've been able to train actually on what is causing certain sort of damage

mechanisms because we've been labeling for the past 11 years. And so it's like it's the AI that we've, we believe they're very much in is like, let's be masters and very excellent at being the best in the world at understanding why things are damaged, what kinds of materials, what kinds of repair techniques, what kinds of inputs as it relates to,

what kind of variables, you know, are leading to certain kind of damage mechanisms to be able to begin to inform and inform what to expect as it relates to how to predict when something would fail. What are the, and then also how do you increase the efficiency, maybe a thermal efficiency or that's throughput or even, you know, like a motor efficiency detecting when a motor is about to fail and how to make it make sure you're adjusting it to be optimum. But basically, when you want to create efficiencies off of this, like, poor information and data sets that we have that no one else does, which is... So you'll be able to go back to the people who manufacture these tanks, the integration firms that actually install them in the construction companies and say, you know what? What we've seen when you're, you know, within 100 miles or 50 miles of the coastline saltwater is X, Y, and Z, these tanks should be built in the following way. Or this is what happens in extreme heat. This is what happens in extreme cold. this is what happens from sun damage.

There might be silly things like a certain code of some sealant in one area might solve problems. And you can even start AB testing this, right? You could tell this person with 50 tests, hey, we're a multivariate test it. We want to have 10 layers put on these, five layers put on these, three layers per on these and want these other ones to be in the shade. I'm coming up with stupid ideas here. But just there's no bad ideas when it comes to testing, extending the life of critical

infrastructure, right? Yeah, that's going to be super powerful. Have you started to give manufacturers like notes or installation people notes on how to do things better from the get? We have not opened up that products or services as relates to helping

improve the OEM process and what materials to choose otherwise. But you're correct in assuming that's where our heads are at, as well as assuming who cares a lot about this stuff while insurance companies do. Because they're insuring all these assets, they're insuring the downtime from these assets.

And so these data sets are actually quite interesting as you relates to the carrot and stick of adopting these kinds of tools because the insights of the, how well is my billion dollars of infrastructure to be taken care of? Right now it's being informed by Joe in a row. So, yeah, there's a lot of...

If you can measure it and you can manage it and you can insure it. Everybody says if you measure it, you can manage it. Managing it, in a lot of cases, means ensuring it. Actually, Freiburg, who I think was the one who brought you guys up on a recent All-In Pod, which is why I invited you.

He did Metro Mile, which was also measuring, hey, how many miles are you doing? We should only charge you for that. And then you start thinking about Tesla's, they have a driving score. I don't know if it's still in the app, but, you know,

one of the people who was driving my car is quite an aggressive driver at times. And it was like, whoa, you're driving pretty close with the person in front of you. It has to distance. It has to speed.

You know, zip, zip. And you could just make insurance for people who are zippy in their car. and people who are slow mows in the right-hand lane, and you can just right-size insurance. What you're saying is, hey, with these tanks, if we're inspecting them and we're doing, you know, this remediation, or maybe we should have a different insurance profile

than somebody who does none of that. And if we're putting these sensors on here, boom, we should have a different level of insurance. That's exactly what happens in journalism, by the way. When I first started my first magazine, they were like, do you do fact-checking? Do you check quotes with the folks who did it? do you record your calls, you know, and they went through all this stuff.

And I was like, oh, wow, this is really interesting. I'm like, why does this matter? Like, well, we're going to make different levels of insurance based on your fact checking. So media insurance, people don't know this. You know, if you're, I don't know, and I don't even know if some people, like Alex Jones, take like an extreme example who like does conspiracy theories and whatever, like, yeah, uninsurable. And then you go to people like, I don't know, New York Times. And if you've ever been in a New York Times story, like, do they check the facts? did they call you and confirm the quotes?

No. New Yorker, at least in my experience, I haven't had a New York Times fact check or check, but I have had the New Yorker check, have had Vanity Fair chat. So Condi Ness does a really good job with that. And the insurance,

I think, works out being proportional to the effort you put into getting your facts correct. And here it's, you know, the effort you put into getting your census correct. Has, have insurance companies started collaborating with yet or no? They've reached out and they've come inbound,

but we basically just held to the approach. which have, look, we're very focused on, like, helping improve the state of our customers' largest problems and our interest in value creation. We'll be more interested in those kinds of models that we're talking about as relates to OEM and insurance at some point in the future. But right now, it's just, you know, we want to build out the infrastructure and a good architecture to begin implementing this, like, this industry 4.0 type of, like, talk. and in a pragmatic way, that's also trying to meet the customers where they're at. I mean, a lot of these customers have a hard time and are very adverse to technologists,

software and robotics like folks coming in because they just have not seen the impact towards helping them fight every, the fires that they fight every single day. And so they're not actually that willing to give you much information to help you build a good product stack. I think this is like, this is why the death of so many,

you know, drone companies or robotics companies or software companies occur in this sector is because one, venture capitalists have no idea about these sectors and what they're talking about. And so, like, you know, we were very much of Lack Sheep because we were just like Pittsburgh robotics company focused on energy. These are all the wrong things back in 2016 when I went through IC. In 2016, yes. And in 2024, now everybody's got the bug, right?

After they've seen what's happened with Tesla and SpaceX, and that opened the wedge up to, Hey, somebody's boring industries or, you know, real world industries might be worth going after. And it was also Uber and Airbnb were also real world businesses. I remember when they were raising their funding, people were like, I don't want to be in a real world business. It's too dangerous. What if somebody trashes your apartment? It's like, well, people trash hotels every weekend. Yeah.

Kind of what hotels are for, at least amongst addicts and rock stars is for trashing them. And like hotels have figured out how to deal with a trashed hotel room. They just throw everything in. They charge the person money for trashing it at the end. Yeah. It's part of the game. Again, here in Pittsburgh, there's like, there's so many robotics companies that like start and die all the time.

And it's began because it's not because they, they are really dumb at building like great robots and solutions. They're actually like really smart. But the problem is what are the robots like useful for? Yes. And that's like the big issue. And that's why we spend so, that's why I spend so much time,

we spend so much time like trying to dig, in with the customers in an embedded way. Because if you aren't embedded... The bear hug is so critical. If a founder gets anything out of our hour together, it's the bear hug works, being on location. And it was a famous story, I think Paul Graham told of telling, or Joe Jebia told it on this podcast,

the co-founder of Airbnb, he said, you know, all the customers were in New York. And Paul Graham told him, go to New York. And he said, you know, all the places with good pictures get rented, the places without pictures don't get rented. He said, go to New York and take pictures. and get a great camera. And they literally bought a digital SR and started taking great pictures, I think.

Yeah. Literally, Ryan and Joe took the pictures themselves as the co-firmers. And this is like the closer you get to the customers, the closer you get to the truth. It should seem obvious, but it's scary to talk to customers for some introverted builders, engineers, whatever. You just got to be right there at their desk sitting side by side with them, solving the problem together. And I think that's what you learned.

And some customers don't want that, right? but you only need one or two to say yes, and then they get the benefit. So if you're on the customer side of this, if you let a startup in bed with you, not in bed, embed with an A, if you embed a startup in your company, you get all the gains years before your competitors.

So if you're in a big company, embed those startups and take a risk with them and help them build the future with you. That's what you were able to do, which is just so brilliant. I think the thing that it's important for listeners to understand, too, is like, in a very regulated environment where especially there's a lot of monopolies at play,

whether it be the government sector or it be the energy sector, like changes very disincentivized. You don't want to change the way you're maintaining something that could go boom and kill people and take down a refinery that's making $50 million a day. So that's actually not that intuitive or accepted for folks to say, hey, kid, come on and give your best shot. That word that was effective was actually in the power sector and specifically the fossil fuel power sector, which were just like, hell, like, I need help because, you know, I've got less funding. I've got less people. I've got less expertise. And my demand is actually like pretty high still. And like I'm having shutdowns of my power plants 50% of the year because pressure vessels just keep exploding. Another way to look at this is what's at stake? You know, I always tell founders like, how much is at stake?

And if you're doing the family trip planning app every time we get pitches in, like every 100th one or every 200th one is, I'm making an app that takes your group chat and lets you plan a trip in an app. And you're like, not a lot at stake. And the solution of doing it in I message or whatever WhatsApp you're into, it works out just fine. It's like enough. Like there's not much at stake here, like splitting the bill.

It's like it's $100 in Mexican food. you got to split it four ways. Nobody cares. It's not enough at stake. Then you start looking at, hey, getting to space, putting stuff in space, SpaceX, a lot at stake, self-driving, you know, getting from point A to point B. Those is actually a lot at stake in, you know, Uber's business or Airbnbs. Like, I'm going off vacation.

I need a place to stay. There's a lot of stake there. And what you found is like, man, if one of these things fails, that's a half billion dollars. And that's insurance companies. People lose their jobs at the company. People get sued. I mean, that's a lot at stake. And as you said, in that one example, you extend that one tank, you save $8 million,

and you probably made, what, $800,000 off that customer or $80,000 off that customer? Yeah, a little bit more than that. But yeah, it's a little more than $800 or a little more than $800. A little more than $800. Okay, so essentially, if you made a little more, you were 15% of the cost of the other reality. So they got 85% of the benefit. you got 15. Pretty happy.

Ballpark. Yeah. That's where, like, I think technology is at best when the customer gets the bulk of the gain and the company gets a small portion of the gain, it makes it a no-brainer. Yeah. And also, like, you have to understand that these sectors are trying like hell to figure out how to adopt technology and not be sold like a bag of goods that is false. And so, like, there has to, what ends up, you end up have to do is create a model that

Very clearly, you can backtrack into where is the value creation happening. But also, how do I sort through the 10 to 20 different options for robotics and drones and AI companies? That's really tough for these large organizations. And they really just want someone to come in and solve a bunch of their problems. And so if you can create an environment where you can bring in and vet technology, you can vet different kinds of robotics tools, fix sensors that are enabled by some smart technology, you end up putting together the different pieces that make up some large outcome that you're trying to solve for the customer. Packaged, though, in a software that, you know, helps to centralize decision making.

And very clearly articulates where the value creation is coming from. And you can interrogate how those decisions were made and what inputs led to the improved outcome. So you actually need to help, you know, in order to have a lot more startups enter the sector, you need to actually create a model that very clearly articulates, like, what the, the product market fit needs to be or what the problem you have to solve needs to be, with the data layer that needs to be added to the stack, like needs to end up looking like or what kind of information you end up collecting that's not currently out there.

And so that model, that's actually like is pretty, we're going to do this with now half a dozen like other robotics companies where we're like, hey, come under our contracts and we really love the solutions that you're building. You can come under our contracts and add these different solutions. Wonderful. Yeah.

Hey, you've got a great drone, a walking drone, underwater drone. We don't have it. Yeah, we'll plug it in. Here's the API. Let's rock.

And this is where I see the humanoid robots going. It's like, these are really complicated problems to solve. And the data that robots collect in the real world is interesting, but it's not actually super valuable to some customer that's trying to solve. How do I increase the efficiency of my, you know, of my like batch process of making a role of steel?

So, like, the robots can do interesting tasks and can actually observe interesting data in the real world and get information. that's not previously available. However, what is the use of that information as it relates to solving some large outcome for a client? So that's how I'm excited about humanoid and walking dog robots because that offers different data layers,

but they're one of a couple different data layers that you need. There'll be a thousand flowers are going to bloom in robotics. I mean, these little ones to carry your burritos from point A to point B. I mean, if you just watch Star Wars or any modern science fiction, you're going to see a range of robots. And, you know, science fiction authors, and directors and creatives, they really do think about human use cases.

And sure enough, these little robots that would scurry past Darth Vader's feet look just like the ones that are delivering burritos in a lot of major cities. And sure, we'll have a C3PO, we'll have an R2D2, we'll have everything in between. And the bomb,

the build of materials on your robots is a little bit high because you have some, I think, some really intense sensors. Yeah. So those look like those could be tens of that. thousands of dollars, I assume in terms of the bomb. Yeah, it's like upper, it's a close to six figure is about where it is.

But it's not, we're not optimizing for the bomb. But yeah, that's right. But when you look at the general robot, optimist or figure or some of these, the bomb on those is going to be what do you think? If you had to pick a number five years from now, what's the build of materials? And then, you know, we can extrapolate pricing of consumer for consumers after that. What do you think like a functional robot that could walk your dog or, or I don't know, do your dishes or, I don't know, tidy up around the house or work in a factory.

What do you think the, without the specialized sensors, what do I think the bomb on one of those is going to be? It's going to be interesting because you also like, think about like what kind of certifications, like the robot has to like have or come under. But yeah, I think I think it'll end up being, that it's going to be hard for me to imagine it's below 40 in five years. I think it's actually a lot higher than that. Because those are early on, but ultimately you think a $40,000 bomb?

Yeah, but ultimately, I think a $40,000 bomb makes. sense in the next like seven years and then 20 or seven in the next seven years I think it'll end up going down basically just based on what kind of like volume so I'm not assuming like in seven years a lot of volume if there's a lot of volume then I'd probably estimate it's you know it sits closer like to the 20k I think it'll get to 10 and it'll get cheaper than that coming out of China absolutely yeah so 40 when they launch 10 ultimately when they're commoditized everything in between and what what are the major cost you think in that robot but would be one of the top two or three costs that you're going to need?

The actuators are the big thing. I think the compute is also going to be expensive. I'm not sure how that will be dealt with. Yeah, does it have like the equivalent of, you know, an H-100 powering it? Or does it have like a MacBook and it's connected to the net? You know, it's like a very interesting question. Yeah, it'll be, there'll be a lot of robots that like, you know, there's certain robots that won't be able to go into certain environments that's like certified for explosion proof. And it's like, those are expensive.

Yeah. to both to get the certification and to ensure that they won't like combust, for example. Yeah. Battery life comes to mind. But the actuators are what make them move, their arms, but the equivalent of your joints essentially and the pulley system to move things around. Those are not cheap right now. Yeah, and like find dexterities. Like those are really tricky. The hands, yeah. We had a company, Root AI that was picking strawberries with the MIT hand. Oh, yeah. Some of these MIT hands are so incredible what they are capable to do. And then we have Cafe X picking up coffee cups and making lattes and putting foam on them.

And we thought it would crush the cup. And how does it do it? And it's like, oh, no, cups are easy. Like, we're working with berries. Really, you're working with berries? Yeah, we're pulling strawberries and off and raspberries. Like, we're talking about fragile berries off of stems. It's not an easy task when you think about it.

But I guess in some ways it is. And then you could, you could actually see these being rented for $10 a day, 20 bucks a day, you know. 10 bucks a day is $3,600 a year, $20 a day is what people spend on lunch now. So 20 bucks a day to have a robot's pretty dope in my mind. I think maybe less about like the commercial, the B2C implications. Most is because like the amount you have to spend on making, like getting that last 10% for robotics and the amount of time. Edge cases. Yeah.

Yeah, the edge cases are just so hard and expensive. So in my opinion, it's more aligned to like what kind of value you're creating from the robots. And then if you can create a lot of value and charge a lot, then you can justify like large amounts of spend onto making some really cool robotics. So I think that's the key that most roboticists haven't actually solved for is what is the value creating and how much can you extract the value create? Once you do that, then you have a vicious cycle of being able to optimize those robotics to, do some really cool things. And I think that's, that's the,

at least the way that I'm approaching it because I don't have, you know, $5 billion to spend an idea. You seem good at picking markets. Where would you send the first humanoid robotics to, to maximize the business model? Soldiers.

Welders. I mean, soldiers come to mind. I mean, think about how much money we put into a soldier. I mean, I had a friend who was a green beret. He was like, I'm like a five,

he told me he was like a $5 million asset. He said the seals are like a $20 million. asset each, you know, based on a cumulative training, you know? I think robotics will not get used in warfare unless there's like some large conflict. And then it'll be like, then we have bigger problems than robotics.

End of humanity problems, yeah. Yeah, but I think it's like you can't send like robotics into some village because like the edge cases, right? It's like there's so much potential issue. And then like then you're dealing with like a large PR problem. A lot of the large government, right? I think it's ultimately going to be where the cost of with a human exposure and the cost of potentially having like a large issue

because of like some of OSHA violation or something like that. So I really, my mind just goes to like what is the most in, deep sea welding is the most dangerous. Deep sea welding is like the one actually in my head I was going to because it's, you know, that's the most one of the most dangerous jobs. Your life expectancy is like four years. And, um, is it four years? Wow, that's wild.

It's like, it's like people get paid like hundreds of thousands of dollars to go to that job. I think it's like 500K was like, a going rate for like an undersea welder. But like your life's legacy is like... Yeah, yeah, trees are dangerous. Very dangerous. Roofers. Construction workers, truck drivers, miners.

I don't think roofers just because it's, again, I don't know anything. Firefighters also very dangerous running into burning buildings. I think, but I think it's like what's the, what is the cost? Like what is the value you can create like from the information that the robots are collecting? I think the, the,

which I think makes sense. But I think I'm more interested in like what kind of... That's where chat GBT's mind went. I just did a chat GPT. What's the most dangerous professions? Logging's up there. I mean, you think about it, logs falling everywhere and like heavy-duty machinery

with chainsaws and blades and... Yeah. Man, if you ever seen those logs like rolled down to hell, man. You're dead if you get hit by one of those, man. I think it'll end up just being oriented towards... Yeah, maybe on the labor side of them too well-de, but I actually think it's more just like walk-downs at refineries. It's, you know, it's like,

welding and perfecting the weld, speeding up the time of the weld. But it's more oriented towards like what kind of new information am I getting in a way that helps to improve the overall state of a, let's say, of the organism that is like making a role of steel or paper products or refining petroleum or making power. It's those set of things. Backer work comes to mind as a good first step too, right? I think that's what Elon's thinking is. If I get them to work in the Tesla factory on repetitive tasks in a controlled zone,

so I'm not sure humans to get run into. Right. What I'm thinking more of is like what kind of information is being collected by the robot that helps improve an overall process that a human, you know, is not constantly streaming data somewhere, right? It's constantly streaming to your head. And that never gets... The best example of that. Best example of that.

So you're walk, let's say like you're walking down a refinery and you're looking at and trying to listen for different kinds of noises that might be indicative of some, like, leak somewhere. Or, like, you're looking at, you know, you're trying to, like, look at, like, temperature transmitters and see if there's, like, some kind of inconsistency of temperature that, like, could lead to something going boom. You can use, like, things like thermal cameras. You can use things like, um, you can use things like LIDARs as well to, like, constantly update. Like, what is the, what's the process of refining petroleum? Um, you can begin to, like, incorporate different kinds of pieces of

information that can tell you how efficient is your facility operating at, and then update whatever model you're using and change the way that you're actually operating the facility. Because ultimately... It's a lessons they would learn in the field that could be incorporated into a better process that you're saying. Basically, it's like we know very little about what's going on in the real world. And so, like, robots or like cars that are going around with lighters spinning all the time, they're like very interesting information and data.

Oh, yeah. monetize in like interesting ways that we don't even know here about or talk about. But it's that same sort of application. They know how popular Broadway is in your town at one o'clock on a Sunday. Measuring emissions is a big thing too. It's like if you're like walking, if you have a robot walking around and doing tests of how much, you know, how much CO2 is coming out of my stack. It's like these, these are like interesting, you know, different kinds of information that can drive,

you know, certain sorts of large outcomes. Maybe it's like some sort of premium you can get from the inflation reduction act or something like that. It's fascinating. I think it's going to be like a brave new world. Can't wait for these things to come out. All right. Listen, Gecko Robotics. Jake, another overnight success, 11 years in the making.

Congratulations, keeping us safe. I mean, I was just thinking about that building in Miami. Remember the pool and that building collapsed in Miami? Yeah. And they had just been spent. And they kind of knew that it was messed up, but they just didn't take. it seriously, they didn't expect it. Man, you get a couple of those happening.

And there are other countries where the building standards are not like the U.S. And that happened in the U.S. Man, I don't know what developing nations now are getting rich and have a lot of buildings that were built, maybe when they weren't as rich and there weren't as much regulation, going back and figuring out, hey, these buildings built in, you know, I'm thinking of emerging countries that are now, we don't use the term first and third world anymore, but frontier markets turning into emerging markets, turning into, primary markets, they're going to need to inspect some of that previous infrastructure and make sure it's

tighter. Yeah, there's interesting stats. There's 700 or 17,500 or 17, 550 bridges in New York. And I think six was the latest or not in need of immediate repairs. It's like this stuff's old and, um, and I just remember the various out and arrows and it was rusted and gross. And that's like one of the premier bridges in New York. It's like, it's really sad to see.

And then also just like, you don't. think about it in the U.S., you can reduce, Roe does this interesting study, 18% is the reduction in U.S. emissions by 2030. If you can stop critical assets from failing and exploding within the

oil and gas manufacturing sector. So, like, these are like pretty interesting, you know, connection points into how important it is to understand the health of the built world that most people don't think about. Yeah, and it's, that's a hard one to sell on unless there's just been something terrible that's happened on an infrastructure basis and people are

highlighted to it because people don't want to talk about the reality of another BP oil spill in the Gulf or another bridge collapsing. It's just, it's dark to think about it, but great that there are people like you out there solving these problems so the rest of us can feel safer. Great job, Jay. Thanks. Wish you continued success and we'll see you all next time on this weekend startups. Bye-bye.