===

Abate De Mey: I’m joined today by Joe Speed. The VP of product at apex AI.

Joe Speed: Great to be here

Abate De Mey: Is Joe Speed your real name?

Joe Speed: It is. I come from a long line of military pilots. So that’s kind of the family business. My dad, his brother, and both grandfathers.

Abate De Mey: Awesome. Yeah. I mean, it’s, it’s a very fitting name for what you’re doing with apex AI. So could you give us a little bit of background on what you guys are doing over there?

Joe Speed: Our special skills taking the open-source and hardening it. So making it deterministic in real-time. Functionally safe certifiable and then safety certifying it to the highest levels. And so our first product is the apex OS which is a ROS2 distro.

And that’s certified to the highest levels of automotive safety, which is ISO 26262 ASIL D. And so we, so that’s what we do. We have a hardened safety-certified ROS2 distro that is being used by carmakers, truck makers, tractor makers, and a few other people as well.

Abate De Mey: So essentially at apex AI, you’re building an operating system that is the core of what cars of the future, especially autonomous cars are going to be using.

Joe Speed: Yeah. And I guess technically you could say it’s a, it’s a Meta OS or a vehicle OS, So what we provide is it runs on top of what you think of as a traditional automotive operating system. Like, you know, an RTOS (a real-time operating system), like a Q and X, a Greenhills a PikeOS, or in the case of some of the infotainment, like a Linux with PREEMPT-RT real-time kernel patch.

And, as far as the application developers, we provide the SDK, the framework, the tools, the middleware for them to quickly develop safe applications that they can then certify for series production vehicles

Abate De Mey: And so what are these applications that they would be building?

Joe Speed: Well, it had started as you know, the co-founders and our company have a deep, rich background in autonomous driving. So, Jan Becker has been doing autonomous driving since the late nineties. Dejan has been doing Autonomous driving and, autonomous agricultural vehicles for very many years. they were some of the very first people to take ROS and use it to do autonomous driving and vehicles.

And they worked with Brian Gerkey and all of the team at open robotics to design, architect, and develop ROS2. And so we’ve been a big contributor to ROS2… We’re on the ROS2 technical steering committee and, do a lot of heavy contributions to the community, both to ROS2. And to the default ROS, middleware and ROS2 Galactic, which is an eclipse cyclone with the built-in Iceoryx zero-copy.

Abate De Mey: Yeah. So before we dive in deeply into what you guys are doing with ROS… The software that you guys are developing, the middleware… Is this targeted towards say the autonomous algorithms for being able to navigate and detect people and then move the car? Or is this something a little bit more general?

Joe Speed: Well, it’s, you know, if you think about how ROS is amazingly flexible and ubiquitous, right? People use ROS for all kinds of crazy things, including lots of things that you wouldn’t normally define as being a robot. […] Our thought around apexOS. And apex middleware is very much around autonomous driving, right?

So we knew that when people are developing autonomous driving, they need the SDK, the framework, the middleware to speed up how easily, how quickly they can develop autonomous driving and have it be safe. The bit that surprised us that was maybe not so expected is…

You know, we have all these automakers and tier-one customers using us for autonomous, but then they started taking our tools and moving them sideways into other automotive domains. So instead of autonomous driving or in addition to autonomous driving, which is more of a “future thought” they’re using it to develop things that go into cars now.

Right. So advanced driver assistance system lane centering, adaptive, cruise control automatic emergency braking, the powertrain cockpit functions telematics all kinds of different things. And so that’s the bit that’s a little bit surprising and wonderful. Is in the same way that ROS is finding its way into all kinds of unexpected corners of the world.

Um, we are finding ourselves getting deployed into unexpected corners of the automotive industry.

Abate De Mey: Yeah. And what’s great about that is that you don’t have to wait until autonomous technology matures or improves. You can immediately start testing out and building the infrastructure

Joe Speed: Well, exactly. Cause you know, there are, you know, pick a number, right? There are, you know, some tens of thousands of autonomous vehicles, but there are tens of millions of regular vehicles.

Abate De Mey: Yeah. And so if you compare it to what the, what the operating system or whatever it is that that is in the current present-day cars, how does this differ from that? And what’s the incentive that is giving automakers and the people we’re developing ADAS and other features to switch over

Joe Speed: Yeah. So there are a few things and we should talk a bit about the architectural technical differences, but one is just the development process. So today – here’s, what happens at automakers, right? Whether they admit it publicly or not, this is the process… somewhere between many and most automakers, use ROS and ROS2, to prototype new features.

So they have an idea, they need a proof of concept. And so they use ROS and ROS2, to develop it when they, when they have a good prototype and they like it, then what they do is they throw that over to the fence, to another group that develops a “pilot implementation”. So they throw all that code away and do a new implementation.

And then when they like that, they throw it over the fence to a third group who develops the functionally safe certifiable version that would go into series production. Well, what that means is for every idea for every Feature for every new application you’re doing, you’re developing it three times and this is not weeks or months.

This is years, right? This is why it takes from like, you know, idea conception to showing up in a production vehicle at your local dealer is, you know, on a, in a best-case, you’re talking about like, oh, I don’t know, five years, six years. Right. You know, eight years is very normal in some cases that’s 12 years. And so why do that?

If we could just help you if you’re prototyping in ROS and ROS2 if we can help you get a path to bring that like more directly into production. Okay. We’re going to shave years and years and, people and, many millions of dollars from the cost of each of these things.

Abate De Mey: Yeah. So the people who are developing, let’s talk about like “ADAS” – is that like the Toyotas and the Volkswagens of the world? Or are they subsidiary companies that are designing this and have some sort of robotics background that they decided to use ROS?

Joe Speed: Well, These days most of the companies have some group of people or people within them with some robotics background. Right. So, you know, ADAS I mean, these are robotics domains, whether they call it that or not. Right. You know, they’re doing use cases and things that would look very familiar to roboticists, especially any that have worked on like AMRs right.

Like if you’ve worked on a turtle bot, you probably kind of understand what they are trying to do.

Abate De Mey: Mm.

Joe Speed: The but this work is happening at the automakers themselves. So right in automotive, we talk about there’s automotive OEMs. so that is your that’s your, your Daimler or your Toyota, your Volvo, your Jaguar land Rover, right?

It’s those kinds of companies you have, what’s called your tier ones. So these are the big, main suppliers who do like complete systems for an automaker. So that’s people like continental, ZF right. These kinds of companies. and then there’s a lot of others, right. You know, there are secondary suppliers, there’s ISV there are people who specialize in developing specific kinds of software that go into such systems.

Um, and you’ve got the Silicon makers, you’ve got, you know, there’s an entire ecosystem, right. It, it kind of spiders out. So, you know, for each vehicle that gets made, like, ultimately you’re talking about thousands of suppliers made something that went into that.

Abate De Mey: And so in the vision of apex AI, then are all of these companies, as they’re passing from maybe one company to another in this chain, are they all using the same product that you guys are developing as a sort of middle ground so that they don’t have to rebuild the wheel every time

Joe Speed: So I think yes, to some extent, but you know, I’ll be specific like we’re a software company. So where we fit is, you know, the pieces in a car that runs software. So you’re talking about, you know, ECU’s that, which are, you know Typically, those are, you know, somewhere, you know, kind of for sake of argument, let’s say something like a raspberry PI-ish in terms of compute power.

Right. So, you know, whether we’re talking about, you know, a PI zero or PI four but it’s kind of in that one. You have your MCUs, these are the microcontrollers, right. So the closest analog for your listeners would be like, you know, like you what is it.. the ESP32, right? It’s that kind of class of computing.

Okay. and in a vehicle, you know, it’s pretty normal that you’re going to have. You know, depending on model, manufacturer, you know, a pretty good frame of reference would be 120 to 180 of these things. so think about, you know, imagine if your robot has like 180 different computers in it. And each of them is developed by, you know, they’re not, it doesn’t mean 180 different companies, but you could easily have them coming from a half dozen or a dozen different companies.

And they’re using different tools, different operating systems, and their toolchains are different. Their test methods are different. and now you’re trying to collapse that together so that you can have a smaller number of bigger computers. And, you know, this is hard. So with ROS, you know, ROS was developed to, you know, earlier on everybody was building their own robots and no two robots were the same.

So you’ve got. A lot of emphasis on portability, a lot of emphasis on having some nice hardware and device abstractions. So, you know, if this robot has a different camera than that robot, I can still get it, figure it out and get it to work. I don’t, I don’t have to throw the application away. And that’s one of the things that’s a little crazy in automotive is, you know, you start swapping out some hardware pieces underneath.

You might actually have to scrap and completely rewrite an application because you don’t have these kinds of abstractions, you know, you don’t have the ability to like, just pick up your software. Like, you know, oh, I was running on, you know, ECU A today, but I can’t get it because of the COVID supply chain.

So I’m moving to ECU B from a different manufacturer. That’s not so easy. but you know, for ROS, these are normal problems. And so the things we do by building upon ROS, using modern language, modern tools, having these kinds of abstraction error, layers, you know, that’s how, like if you visit our headquarters in Palo Alto, there’s an entire farm of.

Automotive computers from all of our customers with different Silicon different operating systems, completely different things. You know, everything from a TI to a Qualcomm, to an Nvidia drive to a Renesas, right. R-car and. And all of these running different things, but for us, they’re all running the same code and those things get exercise and stressed and tested all day every day.

Um, we have a CI farm on AWS graviton2. So something interesting is at the moment, all of my customers have ARM in their cars. So we have a build farm on AWS graviton too. So we’re able to test on arm and then deploy to our arm, automotive VCU, farm, a physical ECU. And that gives the customers the reassurance that they can now take that code and develop it for that target device.

And they’re not going to have issues, right. They’re not figuring it out for the first.

Abate De Mey: Yeah. So are you guys the only ones who are using ROS in real-time on a car? as the, as the infrastructure to control it?

Joe Speed: Absolutely not. So ROS and, more specifically ROS2, is being used by a ton… you know, that is kind of the default platform for anyone doing R&D. Any space, right. You know, AMRs, AGVs forklifts, drones, whatever. Right. What’s different is we’re the only ones who have a safety, real-time, deterministic, functionally safe, and safety certified.

Okay. ROS2 Distro. So one way to think of it. So, you know Everyone uses Linux. Right. And, but when you need to deploy it to like enterprise-scale and have it be hard and, and supported and secure and all of these things, you know, you might do something like you get red hat. So like red hat is a hardened commercial, Linux distro in that same way, what we offer apex iOS is a hardened commercial ROS2 Distro.

Abate De Mey: Yeah. So what is the journey been like getting there? and all the contributions you guys have made to ROS in the process.

Joe Speed: Well, it’s, it’s been hard work, but you gotta remember I’m Johnny come lately. Right? So here’s the thing. I wanted to join apex years ago. Okay. And my wife was like, no, I’ve had enough of your startup nonsense. Like, why don’t you just be there for them? You know, if you love them, you don’t have to join the company to help them.

So just be their friends and help them. And that’s what I’ve been doing for several years now… But then when we got up closer to Thanksgiving, she actually gave me a head nod, and said, “Yeah. Okay. Okay.” And so I called them on Friday and we had details sorted out by Monday.

And here I am. so I know a lot about apex. I’ve been working with them for years, but I’ve only been an employee since November (2021), right. That end of November, the first day of December. And so, so, but yeah, I can answer your questions. So, you know a lot of heavy lifting, you know, you look at what Jan and Dejan did working with Brian Gerkey and the great crew at open robotics around architecting, designing, developing ROS2 all the contributions there.

Um, getting apex as a company off the ground. and we have some really great early investors, you know Some Silicon valley VCs and people like Airbus Toyota research people who believe very early in our mission and we’ve been blessed to get amazing engineers. And the, you know, a lot of people are very passionate about this topic.

A lot of people. Deep skill and experience in robotics and automotive. but like people come from automotive, you know, you can imagine if you’re a developer, it’s a little frustrating that when you invent a new feature, you don’t see it on the dealer floor for many, many years, and they want to change that, right.

They want: “how do we get automotive to move at cloud speed?” And so that’s what we’ve been working on from an SDK framework tools and middleware perspective. others have been working on it from, all the infrastructure around that. Right. So, how do you take cloud-style dev ops and bring that into automotive?

How do you bring cloud-style, virtualization, hypervisors, container, Docker, Kubernetes, all these things and bring that into a car and make it functionally safe and safety certified? So that’s, you know, people like our friends, you know, ARM, AWS, Continental SUSE, Red Hat Bosch, right? They’re all working in this thing called Sophie, the S-O-A-F-E-E scalable open architecture for the embedded edge.

Uh, but the thing that is convenient is it sounds like a girl’s name, right? So you just a Sophie it’s, it’s easy. It’s, it’s hard to spell, but it’s easy to say. And so they’re working on that. So I see that they go together like yin and yang, right. So we’re focused on, you know, how do you develop modern applications that are portable and easily virtualized, and they’re dealing with the

“how do I virtualize? How do I deploy? How do I do an over-the-air update? How do I support mixed-criticality?” So this is a big deal. So it used to be every single function in the car had its own. I think if you’re building a robot like that’s just insane. Right. But this is how it was done. And so now as they collapsed that together with software-defined vehicles, we have multiple domains running on the same physical computer.

Well, not all domains are created equal. Like one domain is controlling. Car radio or my navigation, a different domain is steering, braking, and accelerator, which of these is kind of more important for keeping people alive, right. And not injuring pedestrians either. And so, you know, having these kinds of different critic mixed these different workloads that have different levels, degrees of criticality and putting them on the same computer like that’s really cool.

Yeah. And taking all of these different computers that were developed in different ways. If you, if you could have well, you know, in robots today, you oftentimes will have many computers in a larger robot, but they’re all running the same software, right? They’re all running ROS or ROS2. So it’s, if I need to get a bigger computer, a different computer, I’m moving from Intel to arm to something else.

ROS handles that and it gives you those abstraction layers where I can collapse these into a bigger computer. I can virtualize. I can port it from one hardware to different hardware, right. That’s kinda just accepted as normal and easy in robotics, but in Automotive, these things are hard.

And so they, how do you bring this kind of ease that we accept is just for granted in the robotics community and the speed at which we prototype and deploy in the robotics community, right? Like, you know, when people like fetch in clearpath, develop a new feature, like, you know, you’re not waiting until 2026 to get it in your next robot.

Right. It’s coming in the next update. It’s coming next week. So, that’s what we’re trying to help them with. And it’s been going neat. You know, we’ve got some really great investors, not all of our customers are public. Some of our customers are very public. Like, you know, Zed F continental are extremely public about how ambitious they are, the things they’re doing with us. Others are not, but if you look at who our investors are, it gives you kind of a clue of what’s happening in the industry. Right? You’ve got, your Toyota research, your Continentals, ZF Jaguar land Rover Volvo AGCO, AGCO that’s agricultural vehicles. That’s tractors. That’s like how people get fed.

Like that’s pretty important. truck maker that can’t be named the. hella, which is another big supplier, and God I’d be embarrassed if I don’t think of someone else. And then we have a community of you know, technology, investors, VCs who appreciate what, what we do and, and help get us off the ground very early.

So God bless to

them.

Abate De Mey: Yeah. And so one of the interesting things is that now, as you not only work across multiple companies in the same industry, you’re working across different industries and agriculture and trucks. So does this mean that they all then get to share from the same learnings and then whatever software that you’re developing for this platform now, is now going to be shared by anybody who could jump on the platform as a future customer?

Joe Speed: So yeah, I’ll say absolutely. Yes. With one big caveat. So we respect our customers’ intellectual property. So. And so working with them, you know, we learn, we improve the product, the improving, the product, making more capable, flexible, more performance, lower latency, lower jitter, the ease of getting these kinds of performance gains, like the things that are in the new apexOS, executer.

Is just a shocking improvement compared to what’s in the open-source. Right. It’s really kind of incredible. the latency, the jitter, and the very low CPU cost that it provides.

Abate De Mey: Can you say that again? What was that?

Joe Speed: well in ROS2, you have a thing called the executer, right? So executer decides, you know, what things get done when and in what order. And so we’ve developed one that is real-time deterministic and functionally safe. And it does some rather clever things like. If, if I have a graph, right? So here are the notes, here are the things that happen.

Um, we can take those and collapse them, collapse those down into, into a thread. So they get executed in sequence within the same thread. So you never even get context switching of switching out of the thread. And we get radical improvement in the jitter and latency and CPU cost for that. And I’ll send you, I’ll send you a paper.

It’s actually in a blog that we published about our new product release. And so that’s some publicly available information. It’s kind of. And then, yeah. And then the things that we’ve been doing around the iceoryx zero-copy and the cyclone DDS, these are eclipse foundation projects that we contribute to.

And things like, you know, with DDS UDP, a four-megabyte camera on something that’s like two-thirds as powerful as a raspberry PI three, right? One of these automotive computers, you know, it’s 25 milliseconds. You’re like, no. Okay. Sounds about right. Well, we can do those same four-megabyte camera images, pub-sub between interprocess, intraprocess in 60 microseconds.

And we can do it at 60 microseconds, no matter if it’s a one-kilobyte message or a four-megabyte message. So this fixed latency very, very low jitter, fixed CPU costs regardless of message size that’s kind of big. And for the automakers, you know, when you’re doing functional safety, you have this budget, you have a time budget of, you know, I have to complete this task within this very small time budget or else I put someone’s life at risk.

And if we can make it faster and more efficient and lower jitter, that gives back time budget and CPU cycles for more interesting things, like your algorithms, right?

Abate De Mey: Yeah. And you guys also recently got some publicity for doing, participating in the autonomous race challenge. Could you dive into that a bit?

Joe Speed: Sure. It’s a, you know, as you could see the… haha… It’s a topic we love very much. So apex is on every single Indy autonomous challenge, race car. So all the universities but apex contributed code and even Indy autonomous challenge specific contributions that we have made to ROS2, and the ROS middleware that ROS2 galactic default ROS Metaware, which is eclipse cyclone DDS with the built-in.

Iceoryx zero-copy is used by every team. So when the Indy autonomous challenge started, all the teams were using commercial software, right, commercial DDS. but by the time they got to Indy, all of them had switched to ROS2 Foxy with the eclipse cyclone DDS with, ISE-specific contributions from us and friends.

So people like, you know Robotec AI, ADlink Bosch tier four, open robotics, and friends but the team that won was TUM. And so there’s a couple of things going on there. So one is. TUM upgraded to ROS2 galactic. So they got the very latest of all the improvements we had made for the Indy Autonomous Challenge. Also our co-founder day on went to TUM. So we have a kind of a soft spot in our heart for TUM. And so we supported and helped all the teams, but gave some extra personal attention and assistance to TUM and TUM won the million dollars. Right. I’m not saying that they did, they’re super talented.

Right. And they wrote great algorithms. They’re amazingly well organized. So I’m not saying they won because of us. I’m just saying we helped them and they want WON.

Abate De Mey: Yeah. And you, you know, you mentioned a couple of the changes. Oh, go ahead.

Joe Speed: that’s the TUM car.

Abate De Mey: What were some of the changes that you guys had to make the Indy autonomous challenge I’m sure, you know, this is definitely an edge case compared to the overall product.

Joe Speed: so yes and no. So in terms of the improvements, I’ll just tell you. rosbag2 in ROS2 Foxy is broken. Okay. It is. So this vehicle, what are you talking about? You’ve got six cameras at up to 155 frames per second, three flash LIDARs at… Depending on how you set them up, you know, either 20 Hertz or 30 Hertz.

Um, so 120 degrees flash LIDARs, on these three radars. it’s not four radars because it turns out that a rear-facing radar of one car will interfere with a forward-facing radar, of a different car. Cause they’re all running on the same frequency. Oops. So three radar.

Abate De Mey: So what two in the front and …?

Joe Speed: Yeah,

so oh, so Narrow field of view, long-range radar forward-facing, and then two short-range, wide-angle radars left right port starboard, right?

Two GNSS with IMUs and HDNSS has two antennas and they did something clever here, which is. one GNSS has the antenna front and rear and the other GNSS has its antennas left and right. And from that, we can get enough granularity to not just know where the car is, but when the car drifts through the corners, we know how many degrees of drift and the steering can compensate.

Abate De Mey: Yeah. Even with the amount of error that is inherent to GNSS.

Joe Speed: Well, and it’s GNSS with RTK to be clear that helps the and then there’s, there’s a drive-by-wire system from our friends at new Eagle, safety MCU. Front-ending the Schaeffler pair of van drive by wire, you know, so there’s a lot of amazing technology in these vehicles. Right. So, you know, you’ve got your, your Luminari, your Aaptiv radar, your, allied vision, Maaco cameras, ADlink computer.

Um, what else? autonomous stuff. hexagon, hexagon, NovAtel the GNSS. there’s, a lot is going on. Like these are million-dollar robots, right. And two weeks before CES Paulie moves the Italian team, they took theirs out to apple. Hey. Apple owns a proven driving proven ground.

Okay. For autonomous driving. And it has a five-mile-long high-speed oval track. So poly move, the Italians, took it out to the proving ground and spun it up to 176 miles per hour, like 273 kilometers per hour, which I think is like, the world’s fastest. I wrote the robo race. People will complain. Cause they’ll say like, well, you did it different.

We did ours on the runway and you did yours on an oval track. But so I, whatever they’re, they’re not Paulie move is not done. Setting new speed records. They’re going to go do some more But then like two weeks later, they went out to the Las Vegas Motor Speedway, which is not that big of an oval.

Right. I think it’s like a one and a half-mile oval and they did a hundred, they hit 173 miles an hour while passing TUM to win the race.

Abate De Mey: Oh, wow. Wow. Have there been any accidents?

Joe Speed: Oh, plenty. But that’s the thing that’s so genius about the Indy Autonomous Challenge. There is zero risk. Okay. And when I say risk, I talk about something important. Like people getting injured. Property damage? Oh yeah. We’ve wrecked plenty of these, but it’s okay. We repair them. We just build more of them. Like, that’s the thing that’s so awesome about racing.

Like when I explained to my friends in Japan about the Indy autonomous challenge, you know, they say, well, Joe, you say that has no risk, but we don’t understand these words that you’re saying. Right. Because if you have a fender bender, right, like. If you have some little minor thing in your autonomous vehicle program, it’s a great shame.

It’s a great loss of face, but in motorsports, like, let me tell you a scenario. It’s Sunday, you put a bunch of race cars out on the track and they go out and they race, and some of them crash. Is that unusual or is that just a normal Sunday? Like in Motorsports that’s normal. You expect it. It’s part of the excitement.

Okay. The, and that’s true here. The difference here is when these cars crash, I cry a little because I’ve been working with these kids for two years now and like, I feel their pain. So when they laugh, I laugh. When they cry, I cry when they’re happy and jumping up and down, I’m happy and jumping up.

Abate De Mey: Yeah, not to mention it’s a million-dollar car. So.

Joe Speed: Well, there

is that. So I’m not saying I’m, I’m happy when they crash, but you know, it’s, you know, it’s, it’s like Steve Rogers, you know, we have the technology, we can, we can build a better, stronger, faster than before

Abate De Mey: Yeah. And it actually, proves to be a really good place to test things and to be a little bit aggressive without fear of failure.

Joe Speed: Steve Austin. Sorry. Get my Steve’s mixed up.

Abate De Mey: Yeah, no, it gives them a place that they can test without fear of failure, which is

Joe Speed: Yeah

Abate De Mey: important

Joe Speed: Yeah, absolutely.

And they do, they do a ton in simulation and that really helps. And I think we’re going to put a lot of effort into that. try to get the simulation better and better do virtual simulation. So even new universities can join the program and get involved. And so the simulation is a key thing.

You know you don’t want a situation where. You crashed the car the first time you put it on the track. something else is fun though, is just getting the vehicle around and collecting data. You can remote control it using. So what does every kid know how to use? What piece of gear? An X-Box game controller.

Okay. So it’s kind of a treat to see the car doing laps, being followed by an SUV. And in the passenger seat is a kid with a laptop and an Xbox controller driving this million-dollar robot as it does laps around the track, collecting point clouds, collecting imagery, and building data sets that they can then use to train their algorithm.

Abate De Mey: Yeah. Yeah. And so, yeah, you mentioned simulations. Do you guys also offer a simulation package or do you guys use a simulation package to test out

Joe Speed: So for. For my company for apex, it works with the normal ones, basically, any simulation that works with ROS and ROS2 also works with what we do. our customers also integrate them with their, you know, very expensive automotive-grade simulation packages that they use in their development.

for the Indy autonomous challenge, you know, they had started with the ANSYS simulator for the first half of last year. mid-year lot of the teams pivoted to SVL SIM, SVL simulator from LG Silicon valley labs. and with all the plugins developed by Gaia and contributed by some engineers from the blue origin we got that tour.

Perfect digital twin, like just really dialed in and then LG decided we’re not in the simulation business anymore. And so all the code is still there in public. Everyone can still use it but LG is no longer contributing to it. So

Abate De Mey: Yeah, definitely. So what’s next that apex AI.

Joe Speed: more things, more vehicles, more cars, trucks, and tractors. And, and you’ll start to see us showing up in other industries that have similar requirements. Right. So, you know, the. You know, w AMR’,s AGVs things that drive outdoors things that go off-road, things that go indoors all of that.

And hopefully, space. So you know, NASA and blue origin issued an RFI for space ROS, which we replied to. And in space ROS, they were saying. “You know, what we really want is just if somebody would fork ROS2 make it deterministic real-time, functionally safe and safety certifiable to the highest level.”

And you know, we’re kind of like we, Hey, over here we did. And so we’re talking with our friend’s picnic so picnic or just awesome Dave Coleman and crew over there, and they do moveit, moveit2eventually move it three. They do things for NASA. They have things on the international space station and they have customers have requirements right now today for real-time deterministic.

And so, you know, I’m hoping to put all the pieces together.

Abate De Mey: Awesome. you. Thank you very much.

Joe Speed: Happy to, Hey, real pleasure.