#012 - Trends from Embedded World 2025

Summary

In this episode of the Embedded Frontier, Jacob Beningo discusses the key trends observed at Embedded World 2025, focusing on the integration of commercial tools with modern workflows, the rise of Zephyr as a dominant RTOS, the pervasive influence of AI in embedded systems, the necessity of multi-core processors, and the mainstream adoption of DevOps practices. Beningo emphasizes the importance of adapting to these trends to enhance development efficiency and product reliability.

Takeaways

• Commercial tools are starting to support open source development.
• Zephyr is becoming a dominant player in the embedded systems industry.
• AI technologies are increasingly being integrated into embedded devices.
• Multi-core processors are becoming essential for complex IoT applications.
• DevOps practices are being adopted more widely in embedded development.
• Security and compliance must be automated in development workflows.
• The integration of AI can significantly enhance embedded system capabilities.
• Open source and commercial tools can work together effectively.
• Modern workflows are essential for efficient embedded software development.
• The industry is evolving rapidly, and developers must adapt to stay competitive.

Keywords

embedded systems, trends, Zephyr, AI, multi-core processors, DevOps, modern workflows, commercial tools, open source, Embedded World 2025

Transcript

Jacob Beningo (00:00.514)
Hello and welcome to the Embedded Frontier, the podcast where we dive deep into the ever evolving world of embedded systems. Whether you're an experienced engineer looking to stay ahead of the curve, a newcomer eager to learn, or a director looking to understand the latest industry trends, this podcast is your gateway to understanding the intricate world of embedded systems. I'm Jacob Benigno, CEO and founder of Benigno Embedded Group. We're an embedded software consulting and education company.

dedicated to helping teams and developers modernize their processes and enhance their skills. Our goal? To empower you to succeed in your career and projects by developing affordable, reliable, and secure products efficiently and on time.

That's right. In today's episode, we are once again going to be talking about trends in embedded systems. Now, these trends aren't the overarching trends that I've been seeing or was expecting for 2025. These are going to be the trends that I saw at embedded world 2025. So walking through the show floor, talking with the vendors, talking with other engineers, hearing feedback from people who are attending my course. That's the type of thing that we are talking about today. And

The thing that's very interesting about this is you can actually learn a lot by seeing what's going on at a trade show. And it doesn't so much tell us about what the trends are today in our industry. They really elucidate what the trends of our industry are going to be tomorrow. Okay. For those of you who maybe don't attend a lot of trade shows, you'll often find that what you see on the show floor is often new announcements of things that are being released. But you also see a lot of

you know, uh, smoke and mirrors. Let me put it that way, right? Products that have been put together for the show, um, that they hope are catching, you know, hope will catch on, but you know, may not even be production ready. So now the thing, trends that we're going to be talking about today, we're not going to be talking about the smoke and mirror type. We're going to be talking about the trends that I kind of saw at the conference that I think will become important trends in the back half of 2025 and into 2026 and 27. Okay. And, uh, some of these are going to coincide probably with the trends I put together for 2025, but in any event,

Jacob Beningo (02:12.91)
I hope to give you some details and some ideas of what I saw and where the industry is going. Okay. Now I'm to give you basically my top five trends that I saw. There were a lot of really cool things announced. You can kind of check out Beningo.com on my blog. Some of the things that I saw and stuff like that. But for this particular session or podcast that we're going through, I'm going to give you my top five takeaways. Okay. Now the first one, you know, not terribly surprising, but something that I think is actually really exciting is that

there is a commercial integration with modern workflows occurring. Okay. Now, what do I mean by that? Well, as many of you know, as embedded software developers or software developers and engineers in general, we love our open source software, right? Open source software is a pivotal pillar of how we develop software today. No matter what type of software we're developing, we always go to open source. We go to GitHub, GitLab, whatever. Go to our favorite places to try to find open source software that we can leverage, right?

Open source software is fantastic in that it provides us with software that's already been proven, already works, that allows us to bring it into our code bases and basically build upon the knowledge and experience of others, right? Why reinvent the wheel? If I need a file system, I don't want to do that from scratch. I want to go find someone who's done it before and leverage their expertise in order to accelerate my development, right? Well, the thing that's interesting about what I saw at Embedded World is that a lot of the commercial companies that offer commercial tools,

Okay, think compilers. Think of, you know, maybe different types of, you know, file systems and things of that nature, right? Different stacks that you might be using that would normally be commercial related. I'm seeing them actually, you know, oftentimes that those commercial solutions tie you into their ecosystem, right? And if you want to use their, you know, their compiler or their stack analyzer or any of those types of things, you're stuck going with all of their solution or nothing, right? And, you know, that's kind of the way it is. Well,

What I saw was actually an openness of commercial companies to really embrace open source software and modern workflows. Okay. And what is happening with that is that as these commercial tools start to open up, they start to allow us to get a little bit better performance out of our software. They allow us to leverage commercial tools where it makes sense. So I can, as a developer, be leveraging open source, absolutely, you know, be doing modern techniques. And then if I need more

Jacob Beningo (04:40.128)
processing power, let's say, of my system. So I get a new CPU, I can go and get a better compiler, right? Or what I am seeing is that the commercial vendors are actually enabling more modern techniques. So oftentimes what you'll see is that, you know, while in web or mobile application development, we see that they have adopted, you know, test-driven development, DevSecOps, CI-CD, and many other modern techniques. Embedded systems has been slow to adopt that.

But what we're seeing here is that the reason we've been slow is that a lot of times our development tools just, you know, prevent us from being able to leverage those things effectively. All right. Now, one example I do want to point out, because I thought it was actually pretty cool, was that, you know, when we think about commercial vendors, IAR, for example, you know, they produce compilers, code analyzers, security solutions, things like that, right? We often think of them as maybe a company that you would go to for high reliability software.

Right. They have, you know, their tools are often times, you know, certified for safety related specs, you know, 26262 and so on and so forth. Right. Well, the cool thing is that when I went to their booth and kind of checked out what they were doing, they were very much starting to integrate their tools into more of a modern workflow, making it so that their compiler could be used in the cloud with GitHub workspaces, having it so that they could easily integrate in with test driven development tools like CPPU test.

So you could do stuff in the cloud, very easily integrated into your CI-CD pipelines in, for example, GitHub or GitLab, or maybe if you're using Jenkins, for example. And so I found it extremely interesting that these commercial companies are opening up their, almost what you could before consider their isolated silo, and they're opening it up in a way that makes it very seamless for them to be able to interact with modern workflows and also modern technologies that really are going to change the way that we develop embedded software.

And so I found that to be extraordinarily interesting. kind of another thing that I saw there, and I kind of picked IAR specifically out of all the different companies that were over there, specifically because not only did they have this cool integration with the cloud and this new cloud platform that they just launched, but they also just integrated their compiler tool chains with Zephyr, Zephyr RTOS. And so when we think about open source code, Zephyr is probably the

Jacob Beningo (07:05.558)
premier example of open source software development, right? You know, ran by the Linux Foundation, know, hundreds and hundreds of developers who are contributing to the project, 45 plus companies that are contributing to it and involved in its ecosystem. Okay, that's a big open source initiative. And, you know, while mostly, you know, they've been using like GCC, for example, for their compiler, companies that maybe are doing IoT devices are going to be resource constrained, that need more performance, they can actually go and use a commercial...

tool like IAR and it's built now right into the Zephyr tool chain. just from my command line, I can make a couple little adjustments to the switches and compile either with GCC or with IAR. OK. And again, this is a trend that I saw where many commercial companies are starting to open up their ecosystems and very collaboratively work with, you know, enabling themselves to work with modern workflows, but also enabling them to work with open source software. OK.

And in the past, mostly we've been restricted to these standalone environments. For example, with IAR where you had to use their IDE, you were stuck in their compiler, you were maybe stuck on just using Windows and things like that. This ability to now work across the cloud and integrate across different open source platforms and, you know, on Windows and Linux and these different environments, it's really showing this commercial trend of changing and embracing open source software and kind of the way that we as embedded software developers like to develop software today.

Okay. So, you know, things like DevOps, DevSecOps friendly integration, which is really important. So this does allow us then to work with safety critical systems, run static analysis, compliance checks, and automate security scanning as part of CICD workflows. Okay. So that's really, really cool, I think. Now, on top of that, you know, there's other trends that I saw while I was at Embedded World as well. So this commercialization or this...

commercial tool companies embracing kind of open source and modern workflows. That was a really big trend. Okay. Now the second takeaway I could say that I had or trend that I saw was truthfully Zephyr is going to dominate the embedded systems industry. Okay. Now from time to time I, you know, have occasionally been accused of, you know, kind of showing how, you know, commercial tools are better than open source or people have said that I hate open source. And the truth is I love open source just as much as I like my commercial tools.

Jacob Beningo (09:32.044)
The fact of the matter is that you have to look at the types of projects you're working on and apply open source where it makes sense and apply commercial tools where it makes sense. Okay. You know, as a consultant, as a developer, I very much am for us looking and properly analyzing our projects and choosing what works for our project. Okay. And so oftentimes if you hear me talk against open source, it's usually because I see teams just jump to open source without evaluating the criteria for their projects. Okay. And that is a recipe for disaster.

Okay. I don't want that to happen. I've seen it happen. I've bailed people out when that happens. It's not a fun situation to be in. Okay. Now my suggestion or takeaway of Zephyr dominating the industry. This is coming down to the fact that when you look at real time operating systems in general today, okay. You know, there's a couple, there's lots of different options out there for us. All right. But I think Zephyr is building on the success of the Linux foundation with Linux in general, right?

They know how to create true open source software, how to run true open source projects, right? They know how to get a community to rally around software and really build it out in a way that it solves a lot of problems and can solve, you know, be a good platform for people to leverage in developing their software. I think Zephyr is actually doing a great job of this. You know, oftentimes we talk about overnight successes and we could joke that, you know, Zephyr has become an overnight success, right?

And by overnight, if you're talking about nine years of people working on the project, you know, trying to make it come together, trying to get partners and all that kind of stuff, then I guess you could say it's an overnight success, right? But the fact of the matter is, is that it's not just an RTOS, okay? Yes, there is a kernel at the core of Zephyr RTOS. And that kernel does similar types of things that you'd find in FreeRTOS and ThreadX and NutX and, you know, pick your favorite RTOS, right?

But the thing that I really find interesting is that it's very much more about building an ecosystem of libraries and tools around the RTOS that enable people to develop secure, safe and reliable IOT devices. Or just stand, typical standalone devices. They don't have to be connected, right? But the configurability of Zephyr, the support that it has, I just find it essentially snowballing in the industry.

Jacob Beningo (11:57.74)
Everybody who I've talked to is talking about, you know, Zephyr or they said, yeah, we've adopted Zephyr or they're talking about adopting Zephyr. Okay. And the fact when you have a community with 45 plus big companies who are all contributing to it, you see companies like NXP who with like their MCX series have like almost, I don't want to say they've gone all in with Zephyr, but they've essentially doubled down on Zephyr. A lot of their focus is around making sure that their software and their SDKs all work with Zephyr.

If you actually look at how NXP redesigned their SDKs recently, just a couple of months ago, they released new SDKs and they're using West. Okay. And they're using these ideas that came out of Zephyr to make their own SDKs more componentized and a lot easier to work with. Okay. And so Zephyr isn't just enabling, you know, an RTOS, it's enabling a completely different way of thinking about how we design and build embedded systems. All right. And because of that,

I really see Zephyr coming to dominate the industry. If it isn't already, okay? Zephyr might have already dominated our industry and we just don't realize it yet, right? So my second takeaway from the conference really is that Zephyr is going to dominate everything. And one point I really wanna point out as to why I think Zephyr is also really good. When you look at their most recent, okay, so let me take a step back here.

Many of you know back in September of 2024, I did an RTOS comparison study, okay? A performance study where I compared a commercial RTOS PX5, I compared ThreadX, FreeRTOS and Zephyr together, okay? So in the future, there'll be additional ones I'm trying to add in like RTThread and, you know, NuttX and some other ones, but at least for September, that was all I was able to kind of pull together. And when I actually looked at the comparisons and stuff, I saw that Zephyr, you know, was okay.

It was very much on par with FreeRTOS. There were areas where it really didn't necessarily perform well as a kernel. All right. And the thing that kind of I found to be extraordinarily interesting is that the Zephyr project, while they were kind of like, my gosh, like, you know, this doesn't look good necessarily for Zephyr. They didn't complain. They didn't, you know, try to push back and say, well, these results aren't valid or anything like that. What they did was say, OK.

Jacob Beningo (14:18.006)
Here are some results that show that maybe the Zephyr kernel isn't performing as well as it could be. And then they went back and they made it part of their development process or they added to their roadmap to improve the Zephyr kernel. And if you look at the most recent release, which was like two weeks ago, Zephyr 4.1, it had over 700 different individual developers who contributed to that release. But a lot of the adjustments to Zephyr during that timeframe were to improve performance.

So they took the benchmark that I put together, recognized that there were improvements that they can make, and they improved their open source kernel. in areas, it's 50 % faster in some areas than it was before. I think when I looked at the data, it's like 30 to 50 % better, okay? Or faster, I shouldn't say better. That's very subjective, right? But if you go and watch the video that they put together about the Zephyr 4.1 release, it's...

fantastic. It talks about all these new improvements. It shows how they've included the thread metric benchmark and how you can automate all these things to actually compare different RTOS performances under the conditions that you're working on for your own projects. Okay. And so they took the guesswork out of it. They made improvements to their kernel and they made it so that you could easily benchmark your own code. Okay. Now, in my mind, that just shows that they care about resource constrained devices. And it's not just all smoke and mirrors, as you might say, right? Saying, yes, we're fast and we care.

and then ignoring it. Okay, they very much applied it. Okay, and if we look at some of the stats here, you know, I've got them right here. First, they include the benchmark. Their cooperative scheduler improved the performance by 20 % over the 3.7 LTS release, okay, which was the one I used back in September. Preemptive scheduling, you know, this is the scheduling that we use most of the time when we're working with an RTOS, was 46 % faster.

than the 3.7 version. Synchronization primitives like semaphores were 35 % faster. Message processing, 24 % faster. Okay. And then on top of that, they added support for compiling with IAR. Okay. Now, you know, we could say that, you know, maybe I'm biased because I put together the report and they made changes based on my recommendations, essentially, right? Nothing like some free open source consulting to kind of help point out where you need to improve, right? But the fact of the matter is,

Jacob Beningo (16:40.416)
The thing that I think is important is that they cared. They cared about their software. The community rallied around it and they made a bunch of improvements that benefit all of us who are interested in using Zephyr or who are currently going to use it in our embedded products. Okay. And because of that, honestly, and many, many other benefits, I think we're going to see Zephyr take over the world. Okay. Now, another trend that I saw at embedded world 2025, as many of you can imagine, AI is everywhere. Okay.

You could not go more than a booth without seeing powered by AI or AI generated a fuzz tool or AI this or that. Okay. And if it didn't say AI, then it said ML for machine learning, right? AI at the edge, running machine learning models and inferences and things like that. Okay. So AI was everywhere. Okay. Now AI, where exactly was it? Well, Silicon vendors. Okay.

You know, I saw, you know, ST and XP releasing chips that have neural processors on them. Okay. New silicon companies like a Arab semiconductor. You know, they, they had some parts that have, you know, very interesting parts. I think they call them a E series, E one, three, five, and so on like that, you know, single core, multi-core parts, all with single core or multi-core neural processing units. And these aren't just small, like, you know,

80 megahertz units. I mean, we're talking like 800 megahertz gigahertz clocks on some of these parts. OK, so it's very hard to call some of these things resource constrained. All right. But they are essentially microcontroller based devices with neural processors so that the inferencing of AI models can be done much faster at the edge. OK, now one of the cool things is it wasn't at embedded world, but open MV just just announced a Kickstarter project.

where they're using some of these new STM parts, these STM32N6, and then the ARRIF chip that I just mentioned are going into their new OpenMV products, which essentially enable real-time edge detection of devices using basically camera, using a camera module, okay? So at full 30 frames per second. It's extraordinarily cool technology, okay?

Jacob Beningo (19:06.324)
Even if we're in a very embedded type of environment, there's places where at the edge, enabling this intelligence by running machine learning models at the edge are extraordinarily powerful and important. Okay. Now, when we look outside of those types of applications, where else is AI actually being leveraged? Well, there was a company that I saw there and these are all just companies that I thought had cool stuff, right? I'm not trying to promote anybody in my podcast here.

I'm trying to kind of point you towards interesting uses of AI technology, right? There were several companies out there that had technologies to enable AI where you could give them your system specs or the things you want to do at a system level, and it would convert your architecture to schematics, okay? And there was one company called CircuitMind that, know, their big banner was architecture to schematics in 60 seconds, right? Now, as many of us know, that could be kind of scary, right?

Did you put in all your things correctly from a system standpoint? But there are companies out there doing really cool stuff like that. And when you went and saw the demos and kind of what some of these different companies were doing, you realize that the way that we develop embedded systems is on the cusp of changing. And it's going to change in a way that's completely dramatic and unlike anything we've done before. OK, so, you know, are these technologies available today for use? You know, you have to go and check them out and see if they fit your needs.

But those were a couple of things I saw that were really cool. I saw some automated fuzz testing with AI powered devices, AI automated fuzz testing. That was kind cool from a security standpoint. It'd be great to have AI help us, right? As embed software developers, I'm going be honest. One, I don't think we have the interest in security to care enough to actually secure our devices. And I say that based on experience where whenever I've partnered with people and we've tried to do security training, security webinars, things like that.

I'm be honest, embedded folks just don't show up. Those end up being the least attended sessions. And I'm not blaming anybody, right? Security, we have lots of stuff on our minds. And let's be honest, security isn't our forte. It's embedded systems. And security is kind of a tough topic to figure out and it's always evolving. So if we can leverage AI to help us in those areas, let's do it. Let's absolutely do it.

Jacob Beningo (21:31.084)
Now, one other thing that I saw that I thought was super cool was at the ST Microelectronics booth, they did actually have a foosball table where you could actually go and play against AI. Okay? And you'll have to check out my blog on beningo.com forward slash blog to actually see the picture. But again, super cool where they were using AI and edge technology and obviously their own processors to enable control to be able to play against a human player.

And I'm going be honest, it was actually pretty good at what it was doing. All right. So that's probably enough about AI. That's probably what you hear the most about when you open up your news app and things like that. But again, I think it's really important to stress the fact that AI is coming. It's actually here today, depending on what you are doing as an embedded developer. And to be completely honest, if you are going, well, there's no reason for me to use AI. I don't know how I'd use AI for developing embedded systems. I'd actually encourage you in May.

to attend the embedded online conference. I'm actually going to be giving a two hour workshop that walks you through different ideas and concepts of how you can leverage artificial intelligence to improve your workflows and develop embedded systems more efficiently and effectively. Okay. And these aren't going to be, you know, these aren't techniques where we let, you know, we just kind of let the machine go and generate a whole bunch of code for us across our fingers and hope that things work for us. Right. These are going to be things where I show you how to use

AI in a pair programming type of methodology, where I share tools, okay, that you can use to improve the way that you develop software, where we talk through different pieces of your workflow and where you can augment those workflows with artificial intelligence. Okay. So, you know, if that's something you're interested in learning a little bit more about, that's what my workshop is going to be about at the InvenOnline conference. And, you know, to be honest, you can, you can save 50 bucks if you use Beningo 25 as a code when you register.

Okay. It's not expensive. The conference like full price is only like $300, which, you know, works out to be something like $25 a month. If you broke it down, very inexpensive. Okay. When we look at the fact that as embed developers, you know, our actual overhead is probably over 150,000 us dollars per year, if not higher. Okay. So very inexpensive to have access to 35 experts in the industry and learn some new stuff. Okay. So I'll get off my soapbox about that.

Jacob Beningo (23:55.097)
So it's something that's been on my mind a lot and I'm really excited about. Let's get back to my 2025 embedded world takeaways. OK. Now, another thing that I saw, which I've kind of actually already alluded to, is growing support for multi-core. Multi-core processors, they're not a niche feature anymore. OK. They've actually become a necessity. People are building such complex IoT and edge devices that you have to have multiple cores.

You either end up with a single core and secure your device using a technology like TRUSO, or you use a multi-core processor so that you can have a secure processing environment and then kind of your normal control code as well. All right? The fact that so much AI is its way to the edge through machine learning and inferences, you are going to find that having even a single core microcontroller with a neural processor is something that is becoming a necessity as well.

Okay. I've actually, I mean, for part of my courses through the embedded online embedded software academy, we use the NXP MCXN947 board. Okay. This is actually a really cool little freedom board that I've been working with for about a year. Absolutely love it. It's got a little neural processor on there. Okay. It's also a multi-core processor. So, know, Cortex M33, very powerful devices.

I've got examples of ST parts around and other ones as well, but trying to share multiple different technologies with you, try not to favor any particular vendor or anything like that. But multi-core support, it was everywhere. It's very much a clear, very clear to me that this is our industry eventually, everything is going to be multi-core. It's the way it's going. And as an example, as I was walking around,

I mentioned I saw it in the microcontroller booths. I even saw it at tool vendor booths as well. So for example, when I went over to the Sager booth, we all know of Sager and love them for their flash tools, right? Things like JTrace and the JLinks and everything else. You can actually see one right here. I've got my UltraPlus that I use on a daily basis. I've been using this for probably since they released it. It's one of my favorite debugging tools. But anyways.

Jacob Beningo (26:16.677)
What I saw in their booth was that their system view tool, which many developers that I know of use this, they've expanded to actually support multi-core tracing now. So what you can do now is if you have a multi-core processor, you can use system view to actually trace multiple cores and see what your system is doing. Okay? So you can actually view into both cores, amongst memory and many other things, to be able to actually figure out what your system is doing. Okay?

I always find that to be absolutely important because a big challenge for embedded developers like us who are starting to work with multi-core is trying to figure out how the heck do we debug and optimize this thing for the application I'm working on, right? And we need to be able to visualize what the system is doing. And this is one example of a tool that you can use to do just that, okay? Maybe a couple of other tools that I also like. didn't see them at embedded world, but Percival Trace-A-Lizer and some of their tools.

great tools to be able to do this type of thing. And then also MemFault has some great tools to be able to you know, crash logs, observability and things like that as well. Okay. So I'll mention some of those. Those are like my three top favorite for observability and, and, know, debugging of systems. Okay. And they're not necessarily exclusive. I'll be completely honest. I'll use all three based on what it is I'm trying to do or what type of problem it is that I'm solving. Okay. So keep that in mind. All right. Lots of cool things that are going on there.

you know, to check out and see. Now, you know, kind of going back, I saw parts like the SEM32N6 that has a gigahertz NPU. I mentioned it earlier. You know, this is just, you know, it's kind of cool. 30 plus frames per second. You can actually perform H264 encoding, you know, in real time on the fly. Just insane. Okay. I mentioned earlier a little bit about the Elif. It's not the E series. It's actually the ensemble is the part.

And they actually have an E1, which is a single core and E3, which is a dual core and E5 with, which is a triple core part with a fusion processor. Okay. So we're looking at here at like ethos U55 chips now. Um, and then, you know, even their E7, which is a quad core part. Okay. So when we're talking multi-core, we're not just saying dual core here anymore, folks. We're talking literal multi-core. All right. It won't be long before we start seeing eight 16. And the really cool thing about this is I actually

Jacob Beningo (28:41.773)
mentioned this as a trend several years ago of this multiple cores kind of starting to explode in the embedded space. And eventually what we'll see is we'll have enough low power cores that will have dedicated application zones. Essentially, you could almost think of them as containers, but this will be hardware related where, you know, you'll have certain features that are on individual cores. You may power up a core just to run specific features and to completely isolate that feature from other things going on on the processor. Right. If I have a 16 core part, I could have 16 different tasks running.

use intertask communication to kind of coordinate the movement of data and signals and stuff like that. And I might be able to do that actually in a way that's maybe lower power, depending on how I have my architecture set up. Again, some of this is, that's very, very future, right? But at least today you can see the trend is moving towards more processing power as well. Now, with that said, I certainly don't...

want to elude the fact that everything in the future is going to be these high end parts. OK, there are a lot of things that were at Embedded World where there were eight 16 bit parts. Right. These little parts are still kind of the workhorses of our industry. There's more devices out there running these little cores with a little bit of intelligence than there are probably these big 32 bit parts that are running AI edge and all these types of technologies. OK, so let's not forget there are still eight 16 bit parts on there.

My personal focus is always around ARM 32-bit parts. So I tend to focus a little bit more on that area of our industry. So don't think, you know, those things are going away. If you go on a DigiKey or a Mouser, you'll find there are just as many 8-bit and 16-bit parts as there are 32-bit parts out there as well. Okay. And then the last takeaway that I kind of have for Embedded World 2025, and you know, there could be plenty of other ones, but we'd be sitting here talking about them all day long, is...

DevOps. Okay. DevOps has become mainstream. Okay. So this is a trend that is here today. Um, I have seen more and more teams saying that they have adopted DevOps. Now, what is their sophistication? The level of adoption, to be completely honest, I don't think it's that sophisticated. Okay. I've talked with many, many engineers who said, Oh yeah, we do DevOps, right? We got that. You know, we know how to do that. Uh, and then I've talked with people who are DevOps experts and we talk.

Jacob Beningo (31:09.871)
together with some embedded folks and they kind of come back and they're like, yeah, no, no, no, these folks, don't really, they're not really doing DevOps or fully doing DevOps, right? But the fact of the matter is, is as a trend as an industry, a lot of us are adopting pipelines. We are starting to get good at CI, CD. Okay. You know, in the past, we've been kind of reluctant to embrace DevOps just because of the added complexities is outside our normal day-to-day way of building embedded systems. But honestly, those barriers have been.

fading quite rapidly for several years now. Maybe one of my favorite things that I saw at embedded world related to DevOps and that type of adoption was DevSecOps. Okay. And integrating DevSecOps testing into pipelines. And I actually saw that at the GitLab. Well, I saw it at the IAR booth, but it was a talk given by GitLab. Okay. And they were showing how they have taken code secure.

and they've built it in pipeline so that you can actually do ISO 26262, is that too many? 26262 compliance, all automated through a DevOps pipeline. And traditionally, this has been something that's been tough to do, right? Functional safety and security in areas like automotive, medical, it's generally been a slow manual process. But the cool thing is that they were showing that with DevOps,

we can actually increase the efficiency of this. We can speed up that process, okay? We can actually have formal verification. We can do our static analysis. We can do vulnerability detection directly in the CI-CD pipeline, okay? And when do that, you can automate those compliance checks, and suddenly that becomes part of our day-to-day CI-CD workflow development cycle, okay? Super cool, all right?

So as part of that, can imagine you can do your Misra checks, you can verify other types of functional safety related requirements, scan for those security vulnerabilities, and obviously plenty of other things here. But the fact of matter is that more embedded developers and teams are at least adopting DevOps, whether they're using GitHub Actions or whether they're using GitLab or maybe even Jenkins. I'll be honest, I've not heard as much about Jenkins lately. To be completely honest, a lot of customers I'm working with are using GitLab.

Jacob Beningo (33:33.977)
To be honest, not to push GitHub aside. The folks that are here using GitHub are really the big, the very large companies are using GitHub. And I suspect that's because they have an IT department that feels very comfortable with Microsoft and Microsoft products, right? They know how to work with those. Whereas kind of the smaller, the startups, small businesses, medium sized companies I work with, we're all working with GitLab, okay? Which one works best for you? You figure that out, okay? I'm not here to...

to tell you, I'm just here to share what I see happening in the industry. Okay. And hoping that that then gives you the tools that you need to go out and investigate a little bit further to see, Hey, are you doing these things? What is the best tool that works for you? Because honestly, it's going to depend. Okay. So if there was anything that was clear from the DevOps things I saw at embedded world, security and compliance are not optional anymore. Okay. We can build these things into our CI CD pipelines.

we can become a lot more secure and we can comply with safety and security regulations around the world. That security and compliance, it has to be automated. If you want to keep up in today's world, you have to automate with that. The other thing I saw was that there is a lot more adoption of the deployment side of things. Now there's no need to necessarily deploy. So out to your customer four times a day or faster.

But there might be a need to do that with hardware in loop testing, right? Where I deploy when I commit to a test rig that is maybe on-premise or somewhere else that then runs through tests, verify that my system, my code works on hardware. And then eventually maybe once a month or once a quarter, we do a push, you know, by OTA to the, you know, to our actual product and our customers. Okay. All right. So with that DevOps, you know,

It really isn't a buzzword anymore. It's something that has become embedded in our workflows. Okay. And we are seeing that our workflows are becoming more sophisticated as time goes on. So, all right. So with that in mind, those are just a couple of the things that I took away from embedded world 2025. There were lots of other really cool things going on there, but we obviously don't have time to cover all of those things. Okay. So to, for me to summarize for you, what I saw,

Jacob Beningo (35:59.501)
I saw commercial tools from providers like IAR, example, starting to support open source development, enabling their tools to be more widely available to the industry, but also in environments that are a little bit more modern, like in the cloud. Okay. So what this allows us to do, it allows us to balance the best of both worlds, right? We typically have commercial software and we have open source.

And we've generally kept them apart where open source says commercial is evil and commercial says, open source is evil. We're starting to see those two actually work together and say, Hey, you know what? Open source is good at these things. Commercial is good at these other things. Let's allow them to kind of work together where it makes sense for your development effort. And so that these tools actually work seamlessly together. Okay. That's the first takeaway. The second Zephyr is taking over. All right. So if you don't know Zephyr yet, you know, join my embedded software Academy, take some Zephyr training from, from me.

Join the embedded online conference, which is going to be covering a lot of different Zephyr topics. Okay. This isn't me selling to you. This is me saying, Hey, this is coming. I can help you get to the skillset that you need. Okay. But I really think Zephyr is going to take over the world. The third takeaway AI is everywhere. Okay. If you're not working with AI today, again, you're going to have to start looking at these techniques and technologies because they're not just coming to change the way we

work and change our workflows, they are going to be actually embedded in our devices running inferences at the edge in a very short period of time. Okay. So be aware of that. Fourth, multi-core is the new normal. All right. There's no such, know, yeah, you might be working with a single core part today, but before you know it, you may find that the only thing you can get maybe is dual core parts or multi-core, right? Or eventually multi-core may become cheaper than your single core parts. I don't know. These are all guesses, right? But

trends that are I'm seeing in the industry. And then finally, DevOps is mainstream. It's just a matter of you tuning in and adopting and automating more of your workflows so that more stuff is done for you automatically. Okay. So these are the takeaways. Hopefully these give you some insights on things for you to investigate to improve in your own embedded software development workflows and the way that you develop embedded software. So with that in mind, thank you for your time and attention. I really look forward to talking to you again soon.

Jacob Beningo (38:22.295)
And again, if you are interested in joining and attending the embedded online conference, that's going to be taking place in mid May. You can go over to the embeddedonlineconference.com website and register using the code Beningo25 and that's going to save you $50. Okay. If you do it before the end of March, I think you'll actually save like a hundred dollars off the price. Okay. Come April 1st, I think it goes up $50. So, but you'll still save $50 no matter when you actually register.

Okay. And not to, you know, pitch anything, but it is one of my favorite conferences. I know I'm biased cause I'm a co-founder of it, but we go out and we try to get a, you know, some of the best speakers out there in our industry and really come together and share best practices. Okay. So this is really a conference that designed for engineers to try to share knowledge, best practices, and help get you the skills that you need in order to modernize and effectively develop embedded systems.

in a very dynamic and changing world. all right, with that in mind, greatly appreciate your time and attention, and I look forward to talking with you next time on the Embedded Frontier.