0:00Ryan Aytay, president and COO of Code Metal for about five and a half months, says the company does work of consequence where failure is significant. He previously ran Tableau as CEO after roughly fifteen years at Salesforce.
2:12Ryan Aytay says Code Metal translates any code to any hardware, using formal methods such as mathematical proofs and testing so the result keeps the same intent, is optimized and is provably correct for systems like defense, cars and power grids.
5:49Ryan Aytay cites the CrowdStrike incident, one line of code that took down air traffic control and many flights. He says AI-generated code from tools like Claude Code and Codex isn't safe enough for mission-critical systems at the edge.
7:21Ryan Aytay says AI-generated code is mostly correct but must be perfect, covering every requirement, security scenario and hardware assumption, then physically tested. Code Metal translates, verifies and optimizes, using test generation and fuzzing to prove it matches the original intent.
11:16Ryan Aytay says Code Metal is trusted by the Air Force and built a platform with the military that brings data silos together with code translation and a simulation engine for war gaming, which also applies to logistics or F1.
13:47Ryan Aytay says defense equipment like software defined radios often runs on legacy C++ that isn't memory safe, and Code Metal can modernize a million lines to Rust with a guarantee it works on the device. MATLAB to VHDL is harder.
16:12Ryan Aytay says CEO and founder Peter Morales started at MIT Lincoln Lab and worked on the F-35, then formed Code Metal about two years ago with cofounder Alex around the idea of making hardware as programmable as software.
18:01Sam Nadler says Meta's new glasses appear to cost $200 to $300 while Snap's are about $2,000 and very thick. Ryan Aytay, who owned early Meta Ray-Bans, likes not looking down at his phone but notes privacy concerns.
20:28Ryan Aytay says prediction markets like Kalshi and Polymarket can predict outcomes, sometimes better than legacy media and polls, but anything resembling gambling could influence results, so he favors being open but careful.
21:33Jordan Metzner still finds the camera a little creepy, but notes Meta gave glasses to veterans with vision problems and imagines uses like reading a menu. He says $300 makes them accessible and expects glasses designs to converge.
Provably Correct AI Code Translation: How Code Metal Does It: the key questions from this episode, answered with quotes and timestamps.
Ryan Aytay [0:00]: We like to say we're doing the work of consequence, and, of course, the work the consequence of failure, of course, in this world is pretty significant.
[0:07] Theme song: Built this week, breaking it down. Built this week, we show you how. A fresh idea, a clever tweak, you locked in true.
Sam Nadler [0:17]: Hey, everyone, and welcome to Built This Week, the podcast where we share what we're building, how we're building it, and what it means for the world of AI and startups. I'm Sam Nadler, cofounder here at Ryz Labs. And each and every week, I'm joined by my friend, business partner, and cohost, Jordan Metzner. How you doing today, Jordan?
Jordan Metzner [0:34]: Hey, Sam. Happy to be back. Exciting week in AI world. So much going on. Lots to talk about. And obviously, really excited to have a great guest this week as well.
Sam Nadler [0:44]: Yes. Super excited to introduce our guests. But before I do, please don't forget to like and subscribe. We're 500 subscribers away from 30,000 subscribers on YouTube, a big milestone for us in under one year. New episodes out every Friday. So if you're into AI and startups, please give us a listen. And with that, I'd like to introduce the COO of CodeMetal, Ryan. I'm super excited to dive into your business today. But if you don't mind, give us a quick intro of who you are, and let's jump into high level what CodeMetal does.
Ryan Aytay [1:20]: Yeah. Well, Sam and Jordan, great to be here with you today. And yeah. So I am very excited to be here. I'm about, I think, five and a half months here in at CodeMetal as president and COO, and it's been an amazing journey. Kind of like you said with AI, there's a lot of things happening, and that's ultimately why I'm here. Before this, I spent about four years running the Tableau business. If you've heard of Tableau, great analytics company as part of Salesforce since I was the CEO of that for about three to four years. And then prior to that, spent, roughly fifteen years at Salesforce sort of from the early days in 2006 and 2007 doing a number of different things across the business. And so, yeah, that was it's been a great journey, and, we moved on from that into this new world. And, a lot to talk about, a lot to unpack, and I'm, of course, still learning in my own right.
Sam Nadler [2:11]: So my understanding, and I wanna hear from you because I'm I'm I'm not gonna do it justice, but CodeMetal's real value proposition is its ability to translate code from, one language base to another, but using formal methods. So, institutions, large enterprises can really trust that that translation will work. Is that the core functionality? And I'm sure there's millions of things I'm missing here, but how would you describe it in one sense?
Ryan Aytay [2:39]: And I have a few slides I could pull up and kinda walk you guys through kind of how we look at the world. But I think what you know, when I think about what we do, we're really trying to lower the barrier, if you will, from code to metal. And that's, from software to hardware. That's really where that kind of comes together. And it's really from any code to really any hardware, but I think what you said formal methods, which is a form of verification and using the right, mathematical proofs and the right type of testing to ensure that whatever the source is, wherever that code comes from that it behaves, that it has the same intent, that it is optimized and that is running of course in this new capacity on any hardware just like you would expect it to. And I think, provably correct are the two words that I like to use because it has to work. We're not we focus on mission critical companies that have, a lot of different things and if you don't mind, maybe I'll just see if I can pull up a couple slides. Yeah. Let's do it. So, because who doesn't love slides ultimately.
Ryan Aytay [3:40]: Right? So we gotta have some broad centered conversations. So, as I was mentioning, our goal is really to lower the barrier, as you think about kind of code moving into, from code to hardware environment, that's where the name CodeMetal comes from. We talk about provably correct and I think the really important thing to talk about here is like pretty much everything we depend on, at least I know I depend on in my life, like there's code somewhere. Right? Whether it's, your phone, you depend on that, but that's not as perhaps mission critical, but, like, the military that protects us or the car that you drive in or the, think about, aerospace or the power grid or go down the list of the medical devices that you, if you are dealing with like an MRI machine or like things that you need to make sure that, you can operate in a daily scenario. Like, those systems all depend on code. And so that's great, but it's also a little bit scary because a lot of that code is legacy. Maybe it's, been written years ago, and it's hard to change it. Right?
Ryan Aytay [4:42]: Especially in an AI world, we've all, heard of Tesla and an over the air update. And when you get an over the air update, things change. The hardware gets, it becomes modern. You can bring new technology quickly. And I think that's the type of thing that is really important in this new world. And so, it's great that all these systems depend on code, but the reality is is that, as we go through this world of, like, how do I change things quickly? How do I ship faster? How do I, safety safely translate, old code to new production systems? It's super hard. It's time consuming. It's costly, etcetera. Takes lots of effort, whether it's humans or, AI or whatever it may be to kinda make that happen. And then of course you have to test it. You have to make sure that it's actually something that will work on the said hardware platform that you need to go through.
Jordan Metzner [5:36]: Resumes don't tell you who someone is. Interviews do. Interviews that captures how someone thinks, how they communicate, how they solve problems at scale.
Ryan Aytay [5:49]: We like to say we're doing the work of consequence, and of course, the work the consequence of failure, of course, in this world is pretty significant. Like, you can't have your car fail or your plane you know, like, are just catastrophic scenarios. And an example I like to talk about, even if you go back to something as simple as the CrowdStrike thing that, you may remember took down a lot of the air traffic control and many different flights, that was just like one line of code that ultimately kinda broke. And that wasn't even like a very hard thing, it just was an issue. It was a mistake. And so when we think about like, hey, well, but we're in this new world now and like AI solves everything. Right? I can just go out and start like hammering away and doing some code gen, pick your code gen platform, Claude Code, Codex, but on the list of, folks that are out there. And they're all great and we love them all. But the reality is, like, in a production world, AI generated code actually isn't quite safe enough and really never will be for what we'll call a mission critical system at the edge.
Ryan Aytay [6:48]: Right? So what do I mean by at the edge? Well, it's it's basically the hardware side of things. Right? So it's great if I can vibe code an application, a time and expense app or something like that. That's awesome. All of this happens very quickly and that's why we actually really like those environments. But to make sure that it runs safely, accurately, and just like it used to on the with the old code, but in the new world, in this system sort of mission critical system environment, that's really what we do. And so if you think about like, well, why, tell me more. Right? Well, the reason why sometimes it will fail or why it will fail in general is there's a gap. And that gap is really what we try to solve. But, the gap exists because, AI generated code is mostly correct. Right? We know this, and you can actually go in and ask, said AI platform, like, can you do it in an accurate way? Well, pretty much. But we need it must be correct. Right? It has to be perfect. It has to consider every requirement.
Ryan Aytay [7:46]: It has to make sure that it tests every possible issue or security scenario or hardware assumption, and then it has to be physically tested. Right? And so this is a lot of, like, what we do at CodeMetal. And what I like to think about it is kinda like a promise. It's like, okay. So we don't just do code generation. That in some ways is getting easier and easier. We are all about, we translate, we verify, and we also optimize to make sure that we're doing it in the most efficient way because we know that in this world, there are certain systems that just can't afford to fail. It's kinda like if I'm trying to, change the electrical in a city, like, can't just do it I don't want the lights to go out, but I want it to change. Like, these are the scenarios you have to think about. And so, just because it's always fun to show a little bit of, like, what does this look like? This is a little picture into something that we actually haven't even fully brought out yet, but we will, which is, bit of our own platform, which is, code in, code out. This is a bit of our test generation suite of how things look. And if I think about, like, where do these things go from here, a lot of the things that we talk about at CodeMetal are formal methods, and within that, there's different types of testing. There's, can call it testing and fuzzing and, go down the list of things that we believe are very important to make sure that we're able to look back at what was the intent and the behavior and the origin of the code and how do we use that as a way to kind of go through and do a proof to say like, is this right?
Ryan Aytay [9:16]: Does this work? Will it run on said hardware? So we have this belief, right, where, generating code actually, the price is going down. It doesn't really cost anything, but if you get it wrong, it's pretty substantial. And so that's a bit of where we are. Sorry. That's a very long answer about like what is CodeMetal, but really I just it's a it's it's a very nuanced thing to talk about, but very important and most people don't realize, especially I didn't realize before I joined here is how critical this really is. And I think AI helps us, but it doesn't solve every single problem.
Jordan Metzner [9:51]: Okay. Awesome intro. I've been building software and hardware for years and have tried to stay away from mission critical softwares. And that's always been, hey, guys, nothing we're doing here. We're not flying planes, you know. It's always been kind of a safety net for building, web SaaS and mobile apps and things like that. But, when you have to pay the brass tax of, human life is on the line, obviously, the requirements and the seriousness of the product and software and hardware go to another level, which, historically in my career, I've tried to stay away from because obviously this is a highly risky area and, it takes a lot of time to make sure it's it's gonna go right. And I think on top of that, you've got a lot of historical systems written in legacy code bases that, many engineers don't even know anymore. Right? And they have to kinda like go back to learn. And probably they'd never, but without AI, they probably would never wouldn't, have the ability to be rewritten and, more efficiently and run on, lower cost hardware and things like that. So tell us a little bit, obviously, we see like, huge major, defense contractors on your website and obviously, we don't wanna discuss anything confidential. But, tell us, like, a good use case of, where a company came in and, they applied CodeMetal and they were able to kind of, take old legacy systems and implement, them in a new way.
Ryan Aytay [11:16]: I think there's maybe two things I would talk about. The first would be, we work with different types of, we'll call it the Department of Defense or Department of War, pick your term for that. But, in certain cases there are of course a lot of and I have to be a little bit careful so I don't kinda go into too much detail. But you can imagine, like we're trusted by the Air Force, which I think is a fantastic thing. But ultimately, know in these environments there's a lot of different data. There's different data silos that have been set up. There's data that's been around for years. There's newer data. But ultimately, when we're in a world where there is sort of this concept of we'll call modeling and simulation of like what actually happens, if let's say scenario a happens and this other scenario happens, like it's a war gaming effort, right, over time. It's a modeling and simulation effort.
Ryan Aytay [12:10]: And, we have worked closely with the military to basically build what we call a platform. It's called war matrix, which ultimately is something that kinda brings in the different, data modules or call them silos, if you will, and bring that data together. Of course, that brings some of our, code translation, modernization, and integration together and then build a simulation engine on that, of course, which is powered by many different things. But it allows, I think, for a pretty robust experience so that you can go through and do quite accurate and real life scenarios of like, well, and again, don't wanna go too deep, but like, if this happens, then this happens. Right? And again, I can't give you like specific examples quite at this point. You wouldn't want me to anyways. But really that, so modeling and simulation is relevant, of course, for a military environment. It could be relevant for, logistics environment. It could be, relevant for even like a sporting environment, like an F1 scenario.
Ryan Aytay [13:12]: There are many different scenarios where modeling and simulation become important, but you can't really do it well if you don't bring the data together and you don't do it in a way that makes it, fully translated and kind of all in one experience. So that's that's one thing we do. I would say the other thing we do if we go back to like the translation, modernization and optimization component where we verify things using formal methods and a number of the different, test suite functions that we have. And we work with a lot of, call it defense companies. Right? And I won't name them, but they're pretty obvious. There aren't that many of them. They support the US government. When you think about a company that, let's say does software defined radios or a company that does like, surveillance equipment or a company that does, critical communications equipment out in the field, they are in many cases built with legacy code. Let's say it's C++ And that code base ultimately, it's not safe, meaning memory safe.
Ryan Aytay [14:15]: It could be subject to bad actors. It could, a number of different issues. It may not run at the performance level that, the customer wants it to run at. And so what we do in that case is we can go through and say like, hey, we'll modernize, a million lines of C++ to Rust just as an example in a very fast way, in a very efficient way, and we'll guarantee that it will work on the device that you actually need it to, like the ones I mentioned before. And so I think that's a, that's also in our sweet spot of what we do. There are many I think what I'm finding in this environment is there are many legacy problems that are defined by code similar to that slide I showed, which are mission critical in nature that need, modernization, that needs translation, that need optimization. Now some are easier than others. Right? So you could say actually C++ to Rust may be slightly easier to tackle than let's say like a MATLAB to VHDL type of thing. Those are a little more difficult, but again, we can do these things.
Ryan Aytay [15:15]: It's just a matter of like how does the technology progress. And of course, I also I think it's important to call that we have many great you know, when I think about things like formal methods, we have a team of people almost like a research department that we've brought some of the best people together, folks that come from places like NASA or Stanford or, AWS or, think of some of the great companies out there in great academic institutions. And so we're building an expertise, around formal methods and around verification and validation that I think allow us to make sure that provability is something that we will stand by. And it's really when you think about CodeMetal, we are in many ways focused on safety but also focused on trust. And I think you need in this world to be able to do that. And so lots of use cases. We're just getting started, but so far so good and a lot of great traction.
Jordan Metzner [16:11]: Okay. Awesome. Tell us just maybe one more question on just like kinda how the company got started and found its niche in the space a little bit.
Ryan Aytay [16:19]: So Peter Morales, who's our CEO and founder, worked in a number of different places. Started his career with MIT Lincoln Lab, worked in a number of different things, worked on the F-35 and some simulation environments. I believe worked at Microsoft and a couple other scenarios. And I think found himself, doing a lot of academia and, teaching of some of the things that we've talked about here today. And I think really just had a concept and an idea of, like, this concept of mission critical and, how do I make even, like, hardware as programmable as software? It was kind of the original thesis or North Star vision. And I think, CodeMetal was formed from that like roughly two years ago. And so, ever since then, I think he's done a really fantastic job along with the rest of the team and his cofounder Alex to ultimately bring, a lot of these things to life and put them into use with really important, mission critical scenarios like, wargaming. Like, we know that we need to be thinking about these things. It's really just a planning and preparation exercise that, of course, you would want your military to be focused on.
Ryan Aytay [17:22]: And then, it becomes relevant. Think about it's relevant automotive, medical device, financial systems, like a lot of these things have become fragile over time. And you can't just assume that, hey, I'm just gonna release, said AI platform to go solve it for me because as we know, sometimes these things have, hallucinations or other problems that occur.
Jordan Metzner [17:43]: Yeah. Awesome background. Great story. I can't believe how much progress the company has made with, these types of industries that obviously historically move slow and take a long time to work with that, you've gotten such progress in such a short period of time. So that's that's really impressive. Great little history, Sam. Should we jump into the news?
Sam Nadler [18:01]: Yeah. Let's do it. A lot of news lately. A lot of news in AI.
Jordan Metzner [18:06]: I mean, it's nonstop. I don't know. Where do you wanna go, Sam?
Sam Nadler [18:10]: Well, unfortunately, I didn't choose a very AI themed news. I did choose more of a meta themed news, but I thought it was fun. Meta debuts their new cheaper smart glasses. I don't know if you guys knew this, but about a week ago, Snap unveil Oh,
Ryan Aytay [18:28]: They're they're
Sam Nadler [18:29]: Very thick. They're super thick. And they're also, like, $2,000, I think, while these appear to be, $2.02 200, $300. And then just maybe we can wrap this up into one kinda high level thing. Meta also plans to do a prediction market. Sounds like it will be arcade point based, and that if those points translate to dollars, I don't know. Ryan, for context, one of our best episodes was about building a bot for prediction market. Would love to, a, is Meta have the better glass strategy or Snap? And what do you guys think of this prediction market plan?
Ryan Aytay [19:12]: Well, hey. Look. I'm an optimist generally, but when I think about, I had the early pair of the Meta, like Ray Bans. And I remember I took them to Yosemite, and I kept trying to, like, ask, like, what am I looking at? And this is, like, I think two year roughly two years ago, right when they came out originally. And, for me personally, I've I've always enjoyed the concept of, like, hey. I don't wanna have to be, like, looking down at my phone. I think that's the thing that, glasses like this. It basically gives you a new interface and a new UI to kind of not only look at, what am I seeing, but ask questions and I think get a different perspective. That said, I feel like there's a bit of, there's probably the concern, like privacy concerns that exist out there. Like, am I being watched? Am I being filmed? One of the things I always like at least to tell my kids because I have a 12 year old and a 14 year old is like, you're probably always being watched at this point, and so that's why you need to be very careful about what you do in life. But look, I think the form factor's interesting. I think we're early.
Ryan Aytay [20:11]: The more competition in this space, the better. And I think making these things accessible to other people is great. And, will it be the form factor that, is is vision, a form factor? Probably. In fact, again, I bought them, so I'm I'm an easy test case. But I think there's also voice as another form factor and have, many other companies like OpenAI and things are working on various things to come. So I think it's an exciting time. And, yeah, I think on the prediction market one is like a whole bee's nest of a topic to go through, but I feel like it's exciting and perhaps maybe a way to see, like, more accurate results because it's, less influenced. But I do I have a slight concern, like, anything you turn into, like, a gambling scenario or, like, the ability where someone can make a certain amount of money, then that could influence it. But I you know, if there's a way to, like, somehow moderately regulate it I mean, it's obvious why more people are getting into the space because I think, like, the what is it?
Ryan Aytay [21:13]: The Kalshis and the what's the other one? Polymarket. Right? They're everywhere. Right? And they're, potential good predictors of what's gonna happen in some ways better than, legacy media scenarios. So I think and polls. So I think it's a new world. And I think, being open but being careful is kinda how I look at these things.
Jordan Metzner [21:33]: I think from my perspective, I mean, the Meta Glasses are super interesting. They brought the price point down, obviously, that's really I mean, and then they brought in Kylie, which I think is obviously interesting. I personally think still think like the camera's a little bit creepy. I think if they made like a no camera version
Ryan Aytay [21:50]: Two things right here.
Jordan Metzner [21:51]: Well, they're just like, people in certain scenarios don't want a camera on them like that. Well, I agree that like you're being recorded, part of that is kind of the cameras get obfuscated in your normal day to day life, so you try to, like, kind of act normal. But, know, I don't know if you heard, but, they gave away some of these meta glasses to veterans who have vision problems. And so it seems like, this is like an optimal use case, which is you have vision problems and these glasses can help you. And even if you don't have, like you're not blind, I can just you know, my parents used to like always have to put on glasses when they go to a restaurant and, they show the use case of, like, read the menu for me and see what I like the best. And so it is an interesting, type of new user experience. And obviously, at $300, it makes it pretty accessible. I don't know what they're gonna do with specs at $2,200 plus the, like, cauliflower ear Yeah.
Ryan Aytay [22:47]: The fashion statement.
Jordan Metzner [22:49]: Yeah. Yeah. But it does seem like early days. And obviously, know Apple's working on it, of course. Jony Ive, Yucky mentioned at OpenAI.
Sam Nadler [23:00]: Did Jony Ive confirm he was working on glasses? I thought it was still a mystery what the device I don't know. Was going to be.
Jordan Metzner [23:06]: But I mean, market convergence against like, interfaces. Right? And so, like, it's the same reason why like all the iPhone looks the same as the Samsung phone, you know. I think like we're gonna find convergence on glasses as well. So anyway, hey, Ryan. It was really great to meet. Thanks for joining us today. That was super cool to learn about CodeMetal and just like a whole area where, we don't even think about basically, like six five nines of, confidence and making sure systems will work and, you know, everybody gets back to their family home at night. So that was really awesome.
Sam Nadler [23:38]: Yeah. Ryan, where can people find you?
Ryan Aytay [23:42]: So I we actually just on the website, LinkedIn, I do a lot of posting lately. We've got a couple exciting announcements this week as well, so we'll be getting those out this week and next week and codemetal.ai. So we look forward to chatting with you guys again.
Jordan Metzner [23:59]: Perfect. Thanks, Ryan. Alright. Thanks, Ryan.
Ryan Aytay [24:01]: Have a nice Alright.
Jordan Metzner [24:02]: See you guys.
