Vibe Coding Works, Vibe Robotics Doesn't

Can you build a robot the same way you vibe code software? Not even close.

In this episode of KP Unpacked, KP Reddy and Nick sit down with Guy German, CEO of Okibo, to unpack why programming motion control got 10x easier but building robots still requires years of field testing. Guy breaks down the three requirements for general-purpose construction robots: physical capability (reach, payload, battery life), tool flexibility (spray guns, rollers, power tools, dust collectors), and intelligence (real-time perception, work plan generation). Humanoids fail all three for construction. Chinese robots require pre-fitted BIM data that doesn't exist in reality. Okibo deploys on messy job sites with no prep, no perfect drawings, just LiDAR and situational awareness.

The conversation moves from why construction has the highest suicide rate (cognitive overload plus physical toll) to why workers retire with permanent damage after 30 years (carpal syndrome, can't bend arms from overhead work). Guy shares a story: a veteran worked with Okibo robots for one week during a pilot. When it ended, he begged to keep the robot. His health improved that much. The insight? This isn't about productivity. It's about safety and empathy to the worker. Then they tackle why VCs forgot the venture part of venture capital. If you're showing a hardware prototype and the VC asks about traction, leave the meeting. They've disqualified themselves.

Key questions answered:

• Can you vibe code a robot the same way you vibe code software?
• What are the three requirements for general-purpose construction robots?
• Why do humanoids fail all three requirements for construction work?
• How is the Chinese construction robotics approach different from Okibo's?
• Why does construction have the highest suicide rate of any industry?
• What happens to workers' bodies after 30 years of overhead drywall work?
• Why did a veteran beg to keep the Okibo robot after a one-week pilot?
• What's Okibo's data advantage from deploying across 3M square feet?
• Why is skilled labor shortage real (and getting worse)?
• What should you do if a VC asks for traction on a hardware prototype?
• Why is the capital stack the biggest impediment to construction robotics?
• Is physical AI the biggest technology wave of our lifetime?

If you're building hardware and getting asked about traction, wondering whether robots can work without perfect BIM models, or trying to understand why safety and worker empathy matter more than productivity metrics, this episode will show you why the physical world is messier than code, and why that's exactly where the opportunity lives.

Listen now.

Transcript

Welcome And The Vibe Robot Question

KP 0:00

Hey! KP uh we have a duck uh yeah somebody uh we get like someone zoom bomb us welcome welcome guy to KP Unpack. This is your first time.

Guy 0:18

Thanks for inviting me. I'm honored. I think uh you might be our first guest.

Nick 0:24

Have we had a guest before? I mean you have you know you have guests when I'm not able to do it for the week, so that that kind of counts. I won't you know pretend to be insulted or anything, but I think you've had other guests. This is the first time I've been on a podcast with you. We've had another guest.

Speaker 4 0:42

I'm super honored right now to be the first guest of uh well while both of you are on board, so well I think it's a it's it's a pretty timely guest because you know I had Dr.

Speaker 1 0:56

Barry Clark on in your absence, Nick, who's built a built a couple of things in robotics and knows a thing or two about robotics. And we did talk about physical AI and VLM and all the the fun things that he likes to talk about that I find interesting. So it's a good segue in terms of like we we talked about a bunch of theoretical things and now we can talk about some actual things. And I posted something. Did you see did you see the post, Nick? You know, because that's all you do is read my LinkedIn post and Substack. And that is not what you do. That's the premise of the show, yes. Which post do you so I I posted this idea, right? Like AI and vibe coding, right? Everybody's a coder, right? Everybody's a coder. Your mom's a coder, right? Everybody's a coder, vibe coding, right? So we've we've taken this thing that we believed to be a highly complicated thing, and AI's made it a little bit more available to the masses, i.e., I can build my own software, I don't have to buy it, I can buy build the software that I need for my for my purpose. And then I might even throw it away, right? I'm gonna, you know, may not use it again. So is there gonna be like vibe robotists, you're like roboticists, right? Are we gonna do vibe robots where we start to think about pieces and components for for those of you that are my age-ish, if you remember Radio Shack, you could go to Radio Shack and buy stuff and put stuff together and build your own whatever? I think first computers, you know, first PCs where you made your own. I think a lot of those were sold through Radio Shack. And I remember getting one of my computers that used a cassette tape as a hard drive back in the day. So is there gonna be like this idea of like just what we've seen with vibe coding, the same with robots, where you think about components. I remember when Barry and I were working together 15 years ago and we started talking about you know coding in Ross. I was like, who can we find that knows how Ross works? Right, that was hard. It took months. And you know, Barry tells me these days, what was taking us three months to do, he can do in under an hour now, in terms of the programmatic aspect of it. So if we think about programmatic aspect of it, that becoming fairly simple. And then the hardware thinking around the hardware becomes more modular. Can you just go build your own robotic name the use case?

Speaker 3:36

And and Guy, as you think about the answer to that long-winded question, introduce yourself and talk about what you do in the in the company you're building so the audience knows the context for why you're specifically answering that question.

Speaker 4 3:52

Sure, sure. So so uh I'm Guy Gelman, CEO and co-founder of Folkibo, and we're doing autonomous robots for construction. Uh and yeah, so so we've we've i've I've been working in this space for quite some time, I think close to 10 years now. Yeah, that's a great question because I mean vibe coding is is the big thing right now, and and you are right, like writing a ROS node today is much, much easier than it was, like even I would say even six months ago, or of course 18 months ago, and and it be it becomes simpler and easier.

Why Code Got Easy But Robots Did Not

Speaker 4 4:26

But I think software is relatively deterministic. And so if code compiles, you know, all the unit tests and survives QA, you're most of the way there. But in robotics, it has to perform in the physical world, a physical world is messy, so sensors fail, you have friction changes, surfaces behave differently, lighting, dust, obstacles, all sorts of things that happen in in real life, which is very, very difficult to simulate. And and and that that is a very iterative process. And so although like AI and you know, cloud code and vibe coding and all that really helps, and will also maybe help in design uh of hardware, making more modular and so on, it's it's still a challenge. It will be it will be for for the next few years, uh, I I think it will still be a challenge in in working in with uh yeah hardware.

Speaker 5:22

What um when you when you think about so you mentioned being able to to work in raw ROS has gotten easier with you know working with some of the Ross nodes. What would prevent someone from from doing the DIY robotic approach that KP is referring to, where they're you know they're they're vibing something into it into the physical world? Obviously, if they're like I think there's there's a spectrum here, like in order to do you know industrial robotics, you know, on commercial projects, there's safety certifications, there's like a whole host of things you have to think through. But if you're for instance, if you're wanting to paint or do you know some sort of automated construction work around your house and just it's a fun project for you to take on, like what are the limitations? Like, why can't someone go do that?

Speaker 4 6:08

So I mean you can do a lot of things that you uh it was much, much more complicated to do a few years ago, again, even a year ago. But uh there there is a limit. Robotics is you know, they work in in physical world, and you need to test whatever it is you're building. So it can get you up to a certain point very fast, and then still you need to go through all the way of you know getting uh making sure that that your robot actually performs what you're expecting it to perform in the real physical world. And and that is not something that you can you can make shortcuts. You still need to go through the whole process of actually running, running the thing, the physical thing in the physical world, uh testing it, getting back to the lab, fixing what needs to be fixed. Again, you can use AI, you can use you know vibe coding, you can use you know smarter tools that become better and better, and you can get a lot of information much faster, but still it's it's not it's not a magic. It's it's it still requires you know the real experience in in the field. You need to, if it's if it ai is involved, you you need to gather a lot of data, you need the training. So these things still takes takes time and and resources.

Speaker 1 7:29

You know what the the coolest thing about AI and robotics is my Instagram feed of all the AI generated videos of robots doing things that they're not actually doing. It's almost like the prototyping of robots, right? It's like someone posts some video about a robot doing you know, brushing their teeth for them. It's not real, but it's fun. It's it's fun. That'd be that'd be so cool. If uh, but you know what's what's interesting, I think, is you know, they I keep joking around about like buying a unit humanoid for my house. And everybody's like, hey, you know, they're so cheap, but they can't really do a whole lot, they're really more like remote control. And I'm like, yeah, but I need to unload the dishes, take the dog freak. There's some certain use cases where I just don't have time that I needed to do. And someone's like, Oh, you can't train it to do those things, it's that's not what it does, it's kind of like a remote-controlled humanoid. And then my son tells me, like, Dad, just hire someone in India to remote control the unit tree to like do things remotely, like make your bed and put the dishes away, and it'll be human-controlled from India.

Speaker 4 8:39

And I'm like, Yeah, but then it might be or buy a figure figure AI robot that you know they're I mean much smarter than I've I've become a big fan of Molly's show Sorcery.

Speaker 1 8:51

She had a pretty good video of the figure guys, and now I think they're they're kind of moving big production. I saw something on on the Twitter slash X that Nick is very famous on now. I don't know if you know this. Nick is quite famous on the X these days. So yeah, I saw something on there, like I guess they're had a bunch of their uh humanoids. But you know, it's interesting though, is like a lot of the videos I see on Instagram. Um, at least I'm not saying Facebook, Nick, I'm at least somewhat hip. I'm not on the TikToks though. But you know, there's all these like Chinese robots doing all kinds of stuff, right? And I think some of it's real, some of it's fake, some of it's maybe not as autonomous, right? And you see this stuff like tile laying and skimming concrete, you know, it's just all kinds of stuff, right? And I think my general guess is they're not fully autonomous. There's probably someone off camera pushing some buttons, and then you saw that humanoid robot like in some Chinese day parade kick a little kid.

Speaker 4 10:02

Yeah, yeah, I saw that one. Kick him in a very sensitive area, you know? It's like what is happening?

Speaker 1 10:10

It wasn't funny, actually. It was terrifying. It's terrifying, no, it's absolutely terrifying, right? Like that you know, and that might be the evolution of mankind, right? It's like we gotta learn how to fight better because we gotta be able to fight the robots, right? So yeah, note to self enroll in some martial arts classes.

Speaker 4 10:29

Sure. I mean, some of the you know videos out there are fake, of course, or you know, just showing uh a very small part of a successful uh you know work of some robot and not the not the whole picture, but I think like if we're talking about construction robots and so we're looking at China, like they're very impressive on on general scale, or like right, what they're doing in and but for construction robots, Chinese went like the Chinese companies they went to a very particular

Humanoid Hype And What Videos Hide

Speaker 4 11:04

direction, and and this direction is to work with the regulation in China, also in Hong Kong and other places where where Chinese uh construction robots companies work and rely on data that is fitting the robot's team or other digital information that is fitting the robots before the robot actually starts the work. And and it's interesting that like not one or two, but most of the Chinese companies are doing exactly the same thing. And if they don't have digital information, then they still require to do like a very long process and complicated process of doing a 3D modeling or 3D scanning and then offline modeling of the whole area before the robot starts working. And then so it's it's a short, sort of a shortcut to the real thing, which is get a robot to the field, get it into where it needs to work, don't do any preparation, don't fit in anything because most cases the you will not have that, they will not have the exact information that is needed, that is not outdated, that exactly the right format that you need it. So a lot of, for example, a lot of Chinese robotics companies are working very close to tightly to a specific construction company. So not only are they getting they're getting feeds for where the robots are going to work, but they're getting them like in their exact format or digital format that they're expecting the the data to receive. And and so like the real challenge is not that, it's it's like doing the the real stuff is like get a robot out there and with its own sensors and and perception and ability knows how to do the work that is required, and also not just one one kind of uh task, but like different kinds of tasks.

Speaker 1 12:52

Yeah, and it's interesting too, like you know, we've seen robots on job sites, right? We saw Spot the Dog, we've seen all these like layout, like HP industry robotics, and and all these companies. And and I think there's been a high degree of focus on robots on the interiors, on the insides of buildings versus the outside of buildings. And I think sometimes people have this weird like, oh, well, if you if you have to do something outside, the variables are too great for it to be effective.

China’s Data-First Construction Robot Model

Speaker 1 13:27

I'll say maybe, right? I think the yeah, you have conditions, you have weather, you have lighting, especially if you're running something that requires any kind of cameras, right? Like maybe there's issues with that if you're using cameras instead of layar. But what they don't understand is on the interiors, there's a lot of people, like the variable there, you have you might not get rained on, you might have some variation in lighting conviction convey conditions, but jobs get crowded, like jobs get really, really crowded. And and I think it's interesting, but there has been a focus, and I think you guys have focused on you know on interiors, right? The inside of a building. Nick, what was that drone company we saw that's like they're flying drones to like wash the windows? Right.

Speaker 14:16

I thought yeah, it's a couple there's a couple building facade, like clean cleaning drones. There's drones that are doing more inspection-oriented, they're not autonomous, are they? No, I don't think so. Not not not yeah, not not not in most cases. There's a there's a pilot. I think like regulatory-wise, you have to have a pilot.

Speaker 1 14:36

So so what do you I mean I I think you know, we'll we'll get back to like can I vibe code a robot, so to speak? But I think if we if we think about the interiors, are there are some good variables where you're not having to figure out there's a need to go over mud, right, and fall over. So there is something about like the surfaces being fairly stable, not always. But to your point, you know, I just got off a call where we were talking about construction drawings, right? Construction drawings, people think that they kind of just get better over time. Like it's it's almost like a data enrichment thing that with every set of drawings, they just get better and more detail. And and the answer is that's not actually true. They kind of take three steps forward, two steps back, one step forward, two steps. I mean, it's it's not a linear path to precision, right? So, what you what I heard you say about some of the Chinese robots, how they're thinking about it is they're saying we're gonna build telemetry off of a set of artifacts that aren't actually real, right? They're they're drawings or I mean, BIM, there's not one BIM model on the planet, right, that is accurate. There are like really good looking works of fiction, generally speaking, because once you go into the field, things change, right? The precision doesn't exist. And so talk a little bit about like the idea of like, you know, you have the the the ground truth is exactly that. It's on the ground. You know, you don't get to be in an office and push a motion control to your robot based on a floor plan because it's mostly wrong.

Speaker 4 16:15

Yeah, sure. So so so yeah, you're you're right. It's it's it's many times outdated and it can work in a very, very sterile environment and in some very uh you know special edge use cases where exactly what you've built, you've built it to have robots in in the workflow. And and and you mentioned uh Dusty and Spot. And I I think those robots were the first one to be introduced in construction because they were actually the ones that did not disrupt the the existing workflow, like taking materials from point A to point B, for example, on construction sites, or doing surveys, or or even even the layouts is still pretty much you know in line with with current uh workflows. But I think like the next the next wave is going to be much more destructive and also much more valuable, which will be robots that actually perform work, fabricate, manipulate, execute skill trade tasks on site. That's where we're where we come in. So of course it's it's more challenging to build and it needs to you know know its way around, like localize itself indoor in six degrees of freedom, in real time, and execute uh a work plan that is industrial great work plan. It's it's not like you know, just pick and place tasks. So yeah, so so it's it's a big challenge. It's uh it's a lot of fun, and and yeah, and and it requires different kinds of levels in the robotic stack uh in order to actually work.

Speaker 1 17:54

Yeah, and I think it's I think what's interesting about how you how you're approaching it is you know, we think about motion control within the building, right? Yeah, motion control, you know, the movement of the mission uh of the robot and the motion control, you know, of the arms and all that. And a couple things, you know, you you you probably have lots of variability of indefectors, right? So that's that that's I always tell Barry if I was gonna build a side hustle company right now, I'd be in the custom indefector business because I think that could be like if I could 3D print indefectors custom designed for people, that might be a nice little side hustle. Or I'm gonna start roasting coffees, I'm not sure. Coffee beans. I've been thinking about that too, because I don't have enough to do. So either be in the coffee roasting building or in the robotic indefector business. Those are my two choices right now. But it seems like one of the, because you're working with LiDAR, right, your equipment's very self-aware of the actual conditions. Once again, you know, if the building was the drawings, we wouldn't do as built drawings at the end of a project. So they're very different. And then also like situational awareness around people. So, I mean, how do you think about like what else, like what can't, what do you think can't be executed in the interior of a building by a robot?

Speaker 4 19:14

So I I I think that eventually, eventually, and I like maybe not tomorrow, everything will be automated. Everything will it will be possible to automate all the tasks that are in construction and not only construction, but like the physical world is gonna is gonna be dominated by smart automation. And we're we're actually seeing that happening. But in order to, you know, to build a platform that is you mentioned the antifactors, right? It's like to build a platform that is general purpose platform specifically for construction, uh I think it needs to solve three things at the

The Three Requirements For Real Automation

Speaker 4 19:49

same time. So one has to have physical capability, so right for kinematic-wise, the degrees of freedom, the reach, the you know, payloads that are required in in construction, have enough battery life and and the right form factor, right, needs to fit uh through doorways and move around real job sites and and all that. So that's the physical capability. That's the first thing. And the second thing you mentioned is the tool flexibility. So it's not just the end-of-arm tool, right? If it's uh holding a spray gun or holding a roller or holding uh I don't know a taping box, or whatever, whatever the tool is that's attached to uh sometimes a power tool. It can be a dust collector, it could be a heavy-duty pump. So it also needs to be part of the flexibility of the platform because you can't rely on you know finding the getting all the cables down, you know, uh around and then finding the right um power tool and then connecting it to power socket. And if if if you need an autonomy level machine, it needs to have the the antifactor also needs to have to be able to handle the the power tools. And the third thing that you need, in order to have a general purpose robot that can do actually uh many trades and not you know just not a specific uh application, it has to have uh intelligence, but not just any sort of intelligence and not a human reasoning intelligence, like different kind of intelligence, which is the ability to have the perception of the indoor geometry, environment, to know its way around it, and also to create in real time the the work plans that are required per different kind of tasks. So those those three elements are are are a must. And funny enough, humanoids, for example, they lack all three. Like they they fail all three for construction. Like they're not gonna be able to do that.

Speaker 1 21:48

I think the uh the whole issue with humanoid is really that they're very good at utility of tasks, right? They can do lots of different things, but they're generally slow at all. Of them like when we had our robotics company for one of the use cases, we used gantry machines, they were so fast. I mean, it's like you had to step away from them, they were just so they moved so quick, right? The actuators, the motors just so quick, right? And I think in construction speed, you know, there's a reason why we have all these various tools on a job site because we really do believe that there's specialized tools for specialized purposes, and there's not, you know, that's why we have the master format format divisions, right? There's there's literally numbers across uh, you know, every little thing that gets done. But what do you think? Like one, I think anybody building a robotics company in construction has to be brave and committed. There's so many other industries that might be easier, right? The need is high in construction, but the friction is real, right? The friction is real. So I I never really like bad mouth other robotic companies. I'll badmouth software companies that are not in our portfolio. But robotics companies, I kind of give everybody credit, right? I think it's just really, really hard. Whether, you know, we're whether we're on the cap table or not, they get credit. And so, you know, August Robotics just raised around. And I found it interesting. I'd done it with those guys. I found it interesting. They they were very single purpose, right? Customer focused, single purpose. We're gonna go drill holes in concrete. Was there, you know, now where does it go from there? Well, TBD. But I think this idea, you know, I think Dusty was doing layout first, is that right? Probably. So as we think about like people going in in these, like very narrow, I mean, drilling holes in concrete is a very narrow use case, yeah, right. And you'd ask, well, who needs that? Well, data centers, right? Data centers need a bunch of holes drilled so that they can mount racks, right? So growth market. But and it seems like a good way to get in is under a single use case, but long term, right? Long term, you have to be you the cost of production, right? The way you get costs down and can invest in better in better execution by the robots is scale, right? Which means you can't just do one thing, right? You can't just do one thing. So and then so and then the second part is you can have two mindsets or a hybrid mindset around orchestration. Either robots are gonna be on the job side, they're very self-aware, right? And they're looking out for each other, not running into each other, and there's situational awareness. If you think about a fleet of robots, much like people, right? We have situational awareness. Hey, I'm down here, don't paint on top of me, right?

Speaker 4 24:52

And yet they do.

Speaker 1 24:53

Yet they do, yeah. Or there's some idea, people have this idea like, oh no, there's gonna be an orchestration layer that's you know controlling, right? There's like some antenna in the middle of the building that's controlling whatever all the robots are doing. Like the superintendent of robots, right? I mean, how do you think about those those two ideas? One that there's this master orchestration layer that controls everything, versus the robots having great self-awareness to not paint over each other.

Speaker 4 25:26

Yeah, that's that's a great question. So I mean, I think it's it's it should be eventually a mix of of both. So interoperability between between robots working in the same same job side, like sharing the same map, for example, same 3D map, is very valuable, not only for you know not stepping on each other's toes, but uh, but uh like you know, from data consolidation perspective,

Fleet Orchestration Versus Robot Self-Reliance

Speaker 4 25:51

and you know, the the overall efficiency is is much higher if you do have interoperability between robots. And of course, if if if the platform that is actually performing the different kinds of tasks is the same platform, but only using different different tools, it's it's it it becomes much, much easier. Uh, and and that's definitely part of the things that we're you know designing the platform to support. So like multiple robots doing exactly the same the same tasks, and multiple robots doing different kinds of tasks, but still cooperating through some some shared um infrastructure, for example, the navigation map that they're all all sharing. And and yeah, and so so and and also you you mentioned it is uh every robot needs to know exactly like it can only can't only rely on like a master mastermind, uh like a single point of failure that is controlling all the robot, and it also needs to know its its own task and be self-sufficient. So it needs to have like its own real-time collision avoidance uh capabilities on board. It has to have edge AI, cameras, lidars, or sensors, everything that that is required in order to perform the job. So so so both both things are are important.

Speaker 27:13

Hey, hey guy, this this kind of that question bleeds into deployment and how how customers are adopting your product and other construction robotics companies today. I kind of I'm curious to like even take a step back on that question. You've you you know, you've been in this space over 10 years or close to it. How have you seen the adoption of construction robotics in general change over time during that period? Just the general acceptance, expectations, and you know, introductory calls, like how are those conversations different?

Adoption, Product Fit, And Field Learning

Speaker 27:48

And then also, you know, once you do get a yes from a customer, there's a there's been a ton of focus from the robotics community on just getting more and more robots out into the market and just getting deployed, right? Like that's maybe the best way to have an advantage is to collect the data live in the field. It's not it's not synthetic data, it's not a world model, it's not a video language model. Like it's real hard data that is better than any any other piece of data out there today. And you've done over two million square feet of interiors. And so like you have you you have what I think is a real uh a real data advantage because you've deployed so much. So like I'm I'm curious, like, yeah, secondarily to dig into what have you learned from those deployments? What are you focused on today? Where where are you where are you seeing new and interesting things to try to get customers up to speed and just faster deployed and on a job site?

Speaker 4 28:47

You know, we we hear we heard it from from first day, like construction industry is very conservative. They don't like to change things, they don't like to adopt new technologies, and not only from from robotics companies trying to do you know provide solutions to this industry, also from like so software companies providing I don't know, SaaS uh solutions. So it's they're very conservative. So I don't think so. I don't think it's they're more conservative than any other industry. I think if you have a great product, they'll adopt very, very fast. If if the product provides a positive ROI, if it's easy to operate, if it's easy, you know, it it it it makes sense to use it, they'll use it. It's not like they're it's not I don't think that this specific industry is more conservative than any other industry. Everyone likes great products, but you need to have a great product, and when you're a startup and when you're only starting, you don't have a great product yet. And so this industry, and also other industries, they don't like to, you know, they don't like to teach you how to shave on their own faces, right? So it will be difficult. So every every startup has this, like the beginning is is is pretty hard. You need to convince people, you know, to waste their time and effort and resources on on getting you, you know, you know, getting your product better. And and so so you you need to find exact partners that are willing to do that. But eventually, once you do have, once you found the product market fit, once you have a product that provides value, I don't think there's there's a problem in this industry. On the contrary, they're looking for solutions. Just just provide them with the good solutions and and they'll adopt. And and we saw some like in in some verticals in this industry, for example, the off-site manufacturing, the sales cycle there was amazingly fast, like a few weeks from the first call until they're they're actually you know ordering three or even five robots, like three, four weeks between the first call and and sale. So so that that's super fast. And the reason is like they really need it, and they really have a big problem. And if you if there is a solution, they'll they'll pay for it. And and yeah, so so it's all it's all about and yeah, it's getting to a great product. We're we're not in a great product yet, so like we're on the way there, but getting to a great product takes time, and and AI and hardware and robotics, that's that's pretty pretty intense, and and it takes even longer, it takes resources as well. So yeah, but it's a very, very fun journey.

Speaker 31:29

And then and then yeah, talk to us. So yeah, cover the deployments you've done. Am I right? So roughly two million square feet you've deployed on.

Speaker 4 31:38

Yeah, I think, I think it's it's closer to three million by now. Like we're it's it's in the in the deck, it's it's two million, but you know, it keeps it keeps uh adding adding out more and more projects. Yeah.

Speaker 31:47

And then what are the what are the key, if you're advising a younger founder in this industry, what are the key lessons that you've learned from from deployment? I mean, it is like you talk, you talk, you talk, I mean, what I heard you describing the adoption curve a little bit is there's a much higher bar in our industry because like you're expected from day one to to perform just as well or even outperform the current, you know, the current frontline workers on the job site. They've been doing this for who knows how many years. That's a super tall ask. The industry is efficient, right? Like the the workers are efficient. I was doing some calculation on, and when you look at cost of materials and the the labor, if you if you put construction into Elon's idiot index, which is like here's the raw material that that's still that's delivered versus the the finished finished product, like you want a very high in order to see a lot of disruption, you usually want a very high idiot index. Like rockets famously were like a 50x idiot index because the complexity of the manufacturing process to get to a finished component for a rocket, there was just loads of complexity and super high paid engineers and all that, you know, all that jazz. In construction, the idiot index is 2x. So like there's literally no, like it's already it's already so efficient that your your product has to be better than from day one than that than that existing workforce. And so when you're deploy, and I know you've learned that, like that that lesson is a it's a hard lesson to learn. And I think I I see a lot of startups fall victim to that. They put the product out maybe too early, but it's the fine balance of like deploying, you know, deploying as early as possible while still pleasing the customer and and earning their trust. So, like, yeah, what lessons would you give a founder just about deployment? Sure.

Speaker 4 33:38

So so first of all, of course, I mentioned when you're just starting, when when you you have like your first prototype or second prototype, doesn't that, and and like it's you know, you know, it's it's early, it's early days, and and you're not gonna provide like a lot of value from from you need to find the right partner. So it's it's not not every partner, not every customer is is suitable for that. So you need to find someone that has some vision or someone that is starting to feel like real pain, that is not around uh like even the efficiency, because for example, the this the skilled labor shortage, that that's a real thing. So in in in many places that people talk about labor shortage, it's not so much as labor shortage because you can always find people looking for jobs, but it's skilled labor shortage. And and and and that's that's a real thing. So so when when when serious people and serious companies look ahead like five, ten years from now, they they understand that they're they're they're gonna be in a big problem if they're not gonna look for solutions today. And the younger generation is not following the previous generation into trades at the same rate, and honestly, for a good reason. Like many of these jobs are physically demanding, repetitive, dangerous, make you know, long-term, tall on the body, and and and so that's a big problem. And if you find that the customers appreciate that, you know, wants to improve safety of workers, so you don't need to be like more efficient than day one for those kind of customers. So those those will be the first one who adopt and try out new things because they're visionary and they they they see ahead, or they have like innovation departments and budgets just for that. So like thinking about tomorrow, not just not about this project, and and and you know, that they're not paying you from the from the budget of the project, but from uh some some other pocket, and so it makes sense to them. Uh so those are gonna be those are gonna be the first customers, but eventually, eventually this is it it's not gonna hold just the fact that you know it's safer or or solving like 10 10 other problems, like or providing data, and you have a dashboard and all sorts of other stakeholders, you know, get visibility into the project. You still have to be more efficient, you still need to be positive ROI. And so the robot needs to perform better and and and on a lower cost. It's still need like the economics still needs needs to work, but that that's gonna be later on, and and so it's so it's yeah, it's a journey.

Speaker 1 36:13

Yeah, it's it's interesting. You hit on like several points there, but you know, our friend and partner, Paul Kwan of our general catalyst, he's deep in all the hardware tech and all that. And one of the things I've heard him say in several meetings is to focus on the empathy of the have empathy to the worker and focus on that empathy to the worker. And it kind of solves all these other economic things that we think of, right? Because if you align with the empathy of the worker, then it kind of works everything out. And one of the things I think people just absolutely

Safety, Empathy, And Skilled Labor Shortage

Speaker 1 36:47

forget about the construction industry is safety, right? It's the industry with the highest suicide rate, right? And and you say, why? Why? Because the cognitive overload of just so many decisions, so many risks coming at you on any moment of any day. You have the physical, like it's tiring. You have things like heat, unless you talk to Tiffany over at Stia, and she can get you cooling equipment, cooling, cooling apparel.

Speaker 37:16

Thanks to Stia for sponsoring this podcast.

Speaker 1 37:21

But but but it's the one, it's it's one of these industries. And I think people forget the safety thing. And you know, everybody thinks safety is like, oh, someone died. I'm like, of course, that's that's a binary outcome. It's it's the injuries, it's the fatigue, it's all of those things are also safety issues. And I think we're at a at some conference together, and and and you guys had your platform, which has a reach of how many feet?

Speaker 4 37:51

The high-reach platform can reach 24 feet, and in in in uh in another version, we'll also be able to reach 30 feet.

Speaker 1 37:58

Right, so 30 feet. And then there's little right, and then there was another platform next to you. We won't say their name, give them airtime, but their reach was like what, Nick, like 13 feet or something.

Speaker 38:10

13, yeah.

Speaker 1 38:11

Yeah, 13. And so, like, you know, I was like, ha ha, you're only 13 feet. We can go to the I can be very loyal that way, but you know, the besides the the form factor and reach and productivity, right? There's a massive safety component, right? Yeah, so if we take productivity out of the way just for a second, right? Obviously, it's more productive, right? If you don't have to do scaffolding and do all these things, right? Yeah, but when we think about like the safety of people, I was at a job site like three, four weeks ago, and just watching a guy on scaffolding, it was on like the the the ones that have wheels on them, right? So they're moving them around. He's tying off. There's a guy on the ground kind of pushing him around, yeah, spotting and pushing him around, right? He has to keep tying off on different things. I was like, this is insane. This guy's job all day is to push this other guy around.

Speaker 3 39:08

Yeah, yeah.

Speaker 4 39:12

You pay for a person to do nothing, just if just look at another person. If maybe if something happens, he's there, but uh but what's he gonna do?

Speaker 1 39:20

He's gonna like put his body in front of him and catch it. Like, what is cold 911? Cold 911, really. I didn't understand.

Speaker 39:27

911 after he after he collapses, yeah.

Speaker 1 39:29

Right, like uh okay, I don't know. Anyway, uh, I'm not an ocean expert, but I think when we think about like safety and I know one of your clients that you know they talk about like the fact that you're you're one model that can paint and do you know, the one end fighter that does painting and all that, like like there people don't want to if if he can keep his people from having to go up high, they're happier. Yeah, right, they're happier.

Speaker 4 39:57

So there are two things, right? Of course, if if we eventually get all the people off airlifts back to the ground, we we save lives, actual actual lives. But I think it's it's even it's even sadder than this because like it's not only about the injuries and and and traumatic injuries and and all that. What we see is that workers that spend decades with their you know arms overhead and neck band and wrists under constant stress, what they they they reach after 30 years of doing exactly the same thing, uh retirement, with with damages that are you know for good. And we talked to some people who like have like they they can bend their arms, like they have all sorts of carpal syndrome, and totally real. It's terrible, and like you know, this kind of job is not for for people for the long long term. Like, we don't see people painting cars in car factories, right? Zero people painting cars in car factories, all robots, and eventually we won't see them doing lots of the trades they're doing right now because it's just not meant for people, not not efficiency, not not you know, not that for health reasons and safety. It's just not work meant for us. We we're we need to go up the the value chain higher in the value chain and do do do do more important, significant work, but also that doesn't affect our bodies in in in in this uh horrible way.

Speaker 1 41:31

Yeah, I I I I have this thing that I tell people, especially my wife, that you know, I'm never gonna retire because it's not like I drive a truck, right? It's not like I'm in construction that as long as my brain is functioning, I could like why would I stop working? Right? Why would I stop working? Yeah, because wisdom only compounds, right? I think I'm pretty smart now at 55. I think at 75, I'll be intolerable to my friends, right?

Speaker 4 41:56

I'll just yeah, and in construction, some people retire because they can't do this, right?

Speaker 1 42:00

They simply can't. It's crazy. And so, you know, I it I I think it's one of those wild things that you know, we say all the words, and I think construction companies I think have gotten to be much more empathetic and thinking about their staff and employees and their long-term and and all that. So, you know, I think if if the first thing you think about, I think it's like similar to AI, right? If if the first thing you look at AI and say, oh, it's gonna make me more productive, it's about productivity. You're actually thinking about it the wrong way, right? That that that's not what the goal is, right? It's the goal isn't about productive, it's it's really to say I can do a lot, I can do different things, I can have aptitude in different things, and I get to kind of focus on the things I want to do, right? I look at my AI productivity, which is very high, but it's product, it's productive because it does all the things that I don't want to do. I make it do, right? I still keep the things I want to do. And so I think even robotics in many ways, you know, there if if I'm you know a drywaller, like I am sure there's things on my in my day that I absolutely hate doing, but I have to, right?

Speaker 4 43:12

But he had a guy, a veteran, like he he had all sorts of you know, injuries from when when he was in service and and he worked with our with our robots for one week and then the pilot ended, and he was like, He's like, please, please, can I just keep the robot? What where do I need to to move or shift? You know, just to like he his life was in this pilot where he operated a robot was so much easier for him. Like he is his it was life changing experience. Uh and like that was so satisfying to see. Uh, someone that we we knew that while he worked with the robot, like his his health improved.

Speaker 1 43:55

Yeah, I was I was telling someone the other day, they were talking about you know, innovation. And what are we going to be left to do? Right. I was like, well, why don't we take mops away from everyone and make them get down on their hands and knees and scrub the floors, right?

unknown 44:12

I guess.

Speaker 1 44:13

Because we'll take the we'll take the innovation of a mop away and let them do that. Let's see how that goes over, right? So there's just a little bit of like, I'm not saying someone still doesn't have to mop. There's somebody out there that has to push a mop and a broom and all the things, right? But it's not like we're going to take it away from them so they can use spend more hours at work. I mean, that doesn't make any sense. So I think what's interesting, and I want to segue a little bit in a second and let Nick talk. But I think my conclusion is we're at a rate of change of technology and like the platform Okibo is providing and the ability to do various things, right, within the building. And even as you think about doing different proof of concepts around different use cases, it doesn't feel like I think we're starting to attract the right talent. You're not from the construction industry. You you chose this, right? I was born into this, I didn't really have a choice. But I think we're attracting smart talent like you guys, right? To come into this industry, versus before you might be like, why would I go into construction? So that's happening. So we're attracting the right talent. The technology is getting better and better and easier and easier, and the market's starting to adopt it. There's comfort, like, yeah, like let's let's try it out. Let's see what happens, right? The thing that I think is the biggest impediment to innovation in robotics in our industry is the capital stack.

Speaker 3 45:36

Yeah.

Speaker 1 45:37

So Nick, your turn.

Speaker 45:40

Yeah. So the capital stack, I mean, just to give our audience some context for what that is look, what that would look like for a robotics startup, you have to raise equity, right? Which equity is highly dependent on your traction, which we just said is somewhat difficult to get early on, right? Because no one, you know, there's no

The Capital Stack Problem In Robotics

Speaker 46:03

use cases, there's no data, no one trusts your product yet. So it's like kind of a chicken or the egg thing. Maybe sort like storytelling is a thing that can help you raise equity, like you know, showing credibility through the way that you talk to investors and customers and building trust that way, and telling and sharing a compelling vision that, you know, shows a category-defining company potentially. But then, you know, as you so outside of equity, you have to manufacture the machines, right? And so that's the thing with hardware that I think most people are probably aware of with, you know, just basic hardware devices, but it's true of robotics too. And it's just it's more expensive with robotics to get the robotics and the hardware that you need ordered in order to deploy at the right time while not having that consistent stream of customers yet. So it's it's there's this chicken of the egg thing of like, how do you structure debt? How do you structure equity? How do you structure the ordering of inventory? How do you think about investing in your supply chain, vertical integration, all of this, you know, all of the all of like the key strategy decisions for the business are kind of dependent on how you think about structuring capital out of the gate. So yeah, if you just kind of want to riff on your experience with the right capital stack for robotics companies, challenges, and what the ideal you know state is for a company like Okibo, I think that would be really helpful.

Speaker 4 47:27

Sure. So I mean, of course, the the opportunity is is massive, but the bottleneck is is still uh you're right, it's it's the capital stack, and and so you have to be you know creative and uh more flexible. And so there is, of course, venture capital, but there's also equipment financing and like you mentioned debt and customer back backed uh deployments, leasing, and strategic industry partners that can also help in financing. So that you need to find different ways to finance different different different parts of you know this journey, which requires, of course, a lot of resources. It's it's it's um uh it's not it's not simple building a robotics company. Uh but I I the sentiment that we see in the in the last few months is is is starting to to change. And and and one of the things is uh like there are two things, right? Is first is the defensibility of many SaaS businesses, which became a little bit questionable now with uh cloud AI, cloud code where everyone can you know just uh write an application in in over a weekend, and so the mode of those companies diminishes a little bit. And and then the the second thing is is which is much bigger in my in my opinion is that the physical AI is going to be, I believe, the largest technology waves of our lifetime, maybe you know, bigger than internet and and smartphones, and it's going to be applied to the physical world and only construction, manufacturing, logistic, agriculture, healthcare, everything. And this is a massive, massive opportunity. And it is it's you know cannot be accomplished by two people in a garage and uh ship next week market. So it's a marathon. And the winners are going to be companies that combine breakthrough AI with real machines, real customers, and years of execution in in the field. And so, in order to do that, you need to you need to survive and you need to do everything you need to do in terms of financing. And and yeah, it's it's interesting times for physical AI companies.

Speaker 1 49:42

Yeah, and I'll I'll I'll say something out loud that a lot of people don't like to say. I feel like most venture capitalists have forgotten the venture part of venture capital, right? If you look at early days, fair file, fairfield's semiconductor, you know, that world, Intel, right? VC is rooted in hardware, right? Silicon Valley is Silicon Valley, right? I want to say it's it doesn't say like software valley, it's Silicon Valley, right? It's rooted in hardware. And I think a lot of VCs, because of the SaaS era, the the comp the computation of value and traction and all these things, right? Because it's easy to get traction with SaaS, right? Freebium, whatever, whatever, they've forgotten that the signal, I think just the signal of traction just makes you a bad VC. I think it just makes you a terrible, it makes it makes you a private equity person. Private equity people ask those questions, right? And I think it may be because I'm a little bit older, whatever, right? But I think that's one of the missing things that that people are just missing. It's sign, it's kind of like if a VC, if you're a hardware startup, right? You're there's what's the what's your MVP, right? What does that even look like, right? It's called a prototype. There's no MVP, it's a prototype, right? And if you're showing the a VC a prototype and they ask you about your traction, you should just leave the meeting. You should just leave the meeting. You should just say, like, hey, like this is clearly not a fit for you. Because if you ask for traction, you have these you have dequalified yourself from being in the deal, right? Because I think the venture part of venture capital, it's coming back, right? It's coming back. I, you know, I think you're seeing it with hard tech and all this other stuff. I mean, Nick and I, I mean, I I mean, Nick are pretty much all we're looking at is hardware these days, these days.

Speaker 51:46

Pretty much, yeah. I mean, I don't think the bar for software is three X higher than where it was two years ago, as it should be, frankly.

Speaker 1 51:55

Right. And and even some of the software stuff that we get interested in, it's not because of traction, it's because they actually have to do some work to build something. There's like some real defense. There's some real defense.

Speaker 52:09

Built into the company. Yeah, in fact, in fact, I was talking to one of our companies today that is a software company, but basically they've hitched on to an industry that requires like a lot of deep domain knowledge, maybe some regulatory some regulatory knowledge, basically some some something that a you know, a startup and uh a software engineer can't, great at AI engineer can't go vibecode and and build. Like it's just it's more of a there's more of an in-person and domain moat. And so yeah, I think that's really true. And KP to your point on venture capital and and hardware, I mean, it's showing up in the data. VC, VCs are investing in hardware again. They were kind of forced to. And so I think, you know, as we look at our portfolio and we look at companies like Okibo, that first part of the story that was actually really hard the last four or five years in terms of raising initial equity with maybe just that prototype and selling a vision about antiquated industry and replacing labor and you know, adoption of a work, replacing a workflow and this age of automation. Now there's like a lot of there's there's one been a lot of progress, but two, there's there's a lot more reason to invest in that because it does have a real moat for all the reasons that we've talked about today. It's hard, it's it's difficult. You have a lot more considerations. There's there's you know, there's much more energy and investment into other areas outside of just purely the the you know the software, the computation element, even the even the supply chain relationships. Like it is a like most most hardware businesses we look at today have like a fully vertically integrated business building approach that's just embedded into the nature of what they do. And so yeah, I think like, I mean, that's our observation guy, but has that felt has that felt like the market? Is it felt like the market has shifted? Like, do you do you feel that from when you talk to investors?

Speaker 4 54:06

Like we're feeling this sentiment, yeah, for sure. Like uh people are talking the talk, now they need to walk the walk in that regard, but uh a lot of VCs there they're they're you know they're starting to realize that they need to to look in into like the next big thing. So not it's so it's two things like mode and right, and also like where's the where's the this industry, like where the tech industry is going to, and I think physical AI is is a is uh is a sure bet today. So yeah.

Speaker 1 54:42

Yeah, it's interesting. You know, my first check was a company called icon 3D, 3D printed houses, hardware, materials, software, single family residential, right? All the things that you should walk away running and screaming, right? But what was interesting is it was impressive, right? It was impressive in terms of what it could do and where they were going. And of course, there was there's no MVP, there was nothing, you know, they they didn't even have a real, I would say, a plan like how they were

Physical AI Bets And Closing Thoughts

Speaker 1 55:16

gonna go to market, right? Because, you know, one of the things I think sometimes is, and and yeah, I think you've you know you've learned this too. However, you think you're going to market in this industry, whatever you think it is, you're wrong. You're wrong. You'll find out what right looks like, but on day one, you're wrong, you're always wrong. And and the market shapes you, you don't get to shape the market in it on any lots of iterations, that's for sure.

Speaker 4 55:43

You do a lot of inner, you go and go back and you go back to the drawing board, and yeah, a lot.

Speaker 1 55:49

So I I think it's interesting. I mean, like I said, I I think there are gonna be some VCs that will be more venture, like actual venture, and not ask silly questions like traction, right? But actually start to understand like where does it go? How big is the market? How are you gonna dominate? Like, what is this gonna look like in the future? And then work backwards from that, right? It's like, what does this future look like in 10 years and then work backwards to see, hey, how how do you think you're gonna get to that future? So if there's any VCs, I don't think any VCs listen to this podcast, guys.

Speaker 56:28

I do not know the answer to that question. I would I would assume no.

Speaker 1 56:32

Probably not. Probably not. They don't want to, they they get all they get enough of me in real life.

Speaker 56:36

There could be some espionage going on. I don't I don't know. I don't know. Appreciate all your insights, guys.

Speaker 4 56:42

Yeah, I appreciate you inviting me, and um yeah, it was great, great fun.