Leveraging Modeling and Simulation for Combat Effectiveness

Watch the Video

---

Read the Transcript

---

Brig. Gen. Douglas Wickert:

All right, welcome everybody. So this is those that have held on to the very end of AFA. If you’ve made it this far, outstanding to you. So this is leveraging modeling and simulation for combat effectiveness. If you came here kind of expecting to hear about the old legacy modeling and sim, I’m sorry to disappoint you. We’re gonna talk about the 21st century. That was all so 20th century.

Throughout this week, you have heard from our senior leaders a consistent theme about we have a need for speed and there is a promise. You know, it goes by lots of names, modeling and simulation, digital engineering, the digital thread, and there’s this vision of pulling that digital thread all the way around the life cycle. So what we’re gonna do is really kind of explore that, what it looks like, and we’ll do it through four different chapters. And to do that, we’ve got folks that have been doing this in various capacities for a long time.

So to help unpack how we weave that digital thread all the way around, I’d like to introduce and welcome, we’ve got Mike Pedaci. This is from the far your right to your left. Mike Pedaci, he is the president of Ansys Government Initiatives. He leads the delivery of digital mission engineering and simulation software across defense space, Intel, communities, and then recently spearheaded the AGI’s participation, not the AGI, you’re thinking of a different one, into Lockheed Martin’s inaugural AI Fight Club. And they simulate over 114 years of flight test in a single month of synthetic training. So that’s just an example when we talk about being able to do in a single month, 114 years. There is an acceleration factor there.

Next to him is a very old dear friend of mine, Dr. Dan “Animal” Javorsek. He’s the Chief Strategy Officer for Scientific Systems. Animal is a retired Air Force Colonel, command test pilot, over 2,000 flight hours, a bunch of different aircraft types, 50 and counting. Previously, some time at DARPA as a program manager, you may have heard of the Airbot Air Combat Evolution Ace or Alpha Dogfight, that was Animal’s baby.

Next to Animal is Jonathan Hudgins. Duster is his call sign. He’s trying to shake that, but we won’t let him. So, Jonathan is a Strategy and Growth Manager for the Google Public Sector. He focuses on delivering hyperscale cloud and scalable AI to defense workloads. If you have used genai.mil, you have Jonathan to thank. So, obviously that’s bringing in direct warfighter insight, he is a Lieutenant Colonel in the Air Force Reserve and has done a number of different duties across DIU and now R&E.

And then finally, Doug Gill is a Senior Staff Scientist at Flight Safety International. He’s got more than 26 years in the simulation sector, and that’s not 26 years simulated in one month, that’s actually 26 real person years. He’s a leading authority on live virtual constructive, you may have heard of LVC before, how we build those architectures, the sensory simulation, as well as specializing how the visual environments and the physiological factors, such as cyber sickness, impact warfighter training.

So we’re gonna do this, we’re gonna divide it up into four different chapters, we’ll treat these chapters as chapters across the life cycle. Chapter one will be concept design, and that’s kind of an obvious application of modeling and simulation for the early engineering, but then into concept development and how do we actually really accelerate that development. Chapter three will be real-time operational execution, and then chapter four will be sustainment, and then there’s a special chapter five if we actually get there.

So we’ll jump right in, and the first question, let’s start with you, Mike. Multi-domain concept design. So obviously, we fight across multiple domains now, and so in the concept design space, and I really do think you’ve got probably a unique seat at the table with this from Ansys, a set of tools that are used well across all of our industrial partners, and I’m not gonna ask you to name names, but describe how industry uses digital engineering to accelerate their future force design and their concept development, and how all those different vendor models fit together. I think if you just start talking, they’ll eventually turn you hot.

Mike Pedaci:

Great question. I think I can speak on behalf of all of industry saying simply now more than ever. I mean, essentially, we recognize as industry that in order to win the fight of today, the fight of tomorrow, to achieve the department’s mandate of speed and scale, we’ve gotta fight digitally before we fight physically. So I think that’s number one. You know, me being, I’m newer to AGI, was 20 plus years in the defense industrial base at one company, and started here in January. I’ll describe kind of our vision by a question posed by my 12-year-old daughter.

So I’ve got three girls, and my 12-year-old looked at me, and she says, “Dad, what do you do?” And I said, “Well, I was going to answer,” whether I’m in meetings a lot, you know, and I do PowerPoint, right? Hopefully, you know, AI will take care of that soon. And then I said to her, I said, “Zoey,” because I get to pick some of the movies because I’m outnumbered with the girls, so I said, “Hey, make them watch the Marvel characters,” right? So I said, “Zoey, remember ‘Endgame’?” I said, “So, you know, remember ‘Endgame’?” I said, “We’re not — our company, Daddy’s company, AGI, is not — you know, it’s not Captain Marvel. It’s not Black Widow. It’s Doctor Strange. And our job is to simulate the fight 14,605 times and pick the one that wins.” And that really — that’s our vision as a company, to deliver to the war fighters, except instead of having a green magic stone, we’re the ninth-largest software company on the planet, and we’re the largest, you know, technical software company. So we take it as our mandate, given that we supply so much of the defense-industrial base with our capabilities. We’re kind of a leader there.

So to get at what are three things that we’re doing right now, sir, you know, and I think all of industry would echo this that’s in the digital engineering space, is number one, we’re aligning our organization for the mission. So what do I mean by that? Field-deployed engineers. So we’re investing in talent that can – because unlike the defense-industrial base, which we serve as a primary tool provider to all the large primes, the mid-primes, all the booths you see here today, right, and this week, but we recognize that our direct-to-government customers don’t necessarily have 300,000, 400,000 engineers. So what we are doing now, more than ever, is forward-deploying our engineers to go into the tent, right? That’s number one. Two is we’re collaborating like never before with the defense-industrial base, left of program, right? So what we’re doing is we’re partnering on weapons systems, on platform development for next-generation capabilities, way left, and even in that sense, embedding capability there. And the third thing is we’re looking at ourselves differently. You mentioned the AI Fight Club, sir. And by the way, the general was a user of one of our products back in the ’90s, STK 1.0. But one of the things we’re doing is instead of looking at ourselves as a physics solver, the highest-fidelity simulation from mission to silicon, we’re looking at ourselves as a synthetic data provider. Because I’ll be the first to say on the panel, since you asked me first, I’ll be the first to use the term AI. And so in order to train AI and ML, you need a massive amount of data, millions and millions of data sets. And you can’t do that in the physical realm. You can’t fly enough or fire enough to get the kind of data sets that you need. So you need synthetic data. And one of the things that we have the advantage of being a commercial software provider is we’ve done some of this in other industries.

So think about self-driving. We enabled self-driving capability through synthetic data for a variety of cars that we all probably drive here. And so we’re starting to transform and see ourselves as a leader in synthetic data generation so that we can train the AI systems that we’re going to need for the fight today and tomorrow, so that we can do, like the secretary said earlier in the conference, he had his own spin on speed and scale. He said adapt and scale. And so those are three tangible things that we’re doing, sir.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Mike. I’m particularly intrigued by the end of PowerPoint. That’s real promising.

Jonathan Hudgins:

I have a solution for that.

Brig. Gen. Douglas Wickert:

Well, actually, Jonathan, I did want to go to you next. So you heard Mike reference fighting digital. And so to actually do that, to fight digital, we really do need to kind of scale up from just isolated single servers to being truly kind of enterprise-wide, an enterprise-wide digital sandbox. So from where you sit, again, kind of a unique place with what Google’s building out for the DOW. How does the commercial cloud enable the Department of the Air Force to run these massive, high-fidelity modeling workloads and do it securely at both IL-5, IL-6 level, enterprise-wide?

Jonathan Hudgins:

Yeah, thanks. No, for real, switch over to Google Slides. Gemin creates all my slides now, so anyway.

Brig. Gen. Douglas Wickert:

You heard it here first.

Jonathan Hudgins:

Yeah, yeah, yeah. So you mentioned data, right, all the data you need. There’s like three key ingredients. There’s the algorithm, the data, and the compute. So what I’m bringing, what Google’s bringing to this equation, are all three, but in particular, the compute and at a hyperscale. What Google has done- obviously, Google has Google Cloud, right? We’ve got massive data center campuses all over the country, all over the world.

What we’ve done for the national security community is we took a different approach from Amazon or AWS. Instead of building a bespoke GovCloud that cost billions of dollars to build and maintain, we got our commercial stack, our CONUS, or Continental United States Commercial Cloud, all 10 regions. Each region has at least three data centers in it. So we’re talking 30-plus data centers accredited to IL-5. That happened in 2022. So a tactical example of that capability today, as the general mentioned, is genai.mil. So genai.mil is running on that infrastructure today. It took us 54 days. We talk about– you talk about speed. We got the call from DOW CDAO this time last year, said we want to do this. We want to launch this. We have this idea called genai.mil. 54 days later, we launched to 3 million users. You can only do that with hyperscale compute. And hyperscale compute that is unbounded by the restraints that current GovCloud has.

We were able– because if– and another tactical example of that is, if you’re a regular user of genai.mil, you probably noticed that the other models, not Gemini- Gemini went live on 9 December 2025. The other models took nine months to go live. They just went live a couple weeks ago. Why? Because they’ve spent the last nine months trying to find GPU capacity in GovCloud. GPU capacity that doesn’t exist, or didn’t exist until they trucked in more GPUs into those data centers. So when we look at modeling and simulation, from the Google’s perspective, we see modeling and simulation from an infrastructure, from a compute infrastructure perspective, very similar to how we look at LLM pre-training. The months– all of the data that has to go into pre-training the next version of Gemini, or Clod, or Grok, all of that work is spinning across parallel computing connected by our infrastructure, by our network, across multiple data center campuses, across multiple regions. And we’re able to turn that up and turn that down. And now we’ve got that accredited to be able to serve national security community.

So that’s available today at I/O 5. That’s what genai.mil is running on. It’s coming– that global infrastructure or hyperscale infrastructure is coming for I/O 6. And later next year, you’re going to see genai.smil, which is an example of that same hyperscale capability for secret workloads.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Jonathan. So we’ve gone from– so we’ve built the infrastructure. And we’ve got the digital sandbox. And now we’ve got these physics-based, physics-derived models that we can do. So we will move from chapter 1, which was all about the concept in the model, vendor model development, into actually accelerating that into actual real things. So Animal, I want to pose this one to you because I think, from your time at DARPA, I mean, you actually did this. You took a model. And you turned that model into autonomy. And now that autonomy is turning into really, no kidding, operational capability. And so talk about that pipeline in chapter 2 here of turning the models into real capability.

Dr. Dan “Animal” Javorsek:

Yeah, you bet. We’ve all heard the saying, all models are wrong. Some are useful. We also know that the synthetic environment is kind of doomed to succeed. And by that, I mean it’s actually quite common to be provided some synthetic environment and a model in it where everything works out fine. And Beaker and I both have a lot of experience getting to go do this in practice and see where the models are actually wrong at. And you said that you used the word pipeline there, just kind of like the discussion we’re having here. It’s actually more of a loop, right? And I think what you find is that when you are going to take these systems and deploy them into the real world, you get a chance to find where all the edge cases and, say, the brittle parts of especially these AI models tend to expose themselves. And they don’t tend to expose them in that synthetic environment, but rather in once we get a chance to see them in the real world.

Obviously, as we’ve discussed before just earlier this afternoon, the scale is certainly different. And the ability to explore a much larger piece of that landscape is what’s enabled by the compute, and the modeling, and simulation, and even the analysis that teams like yours provide. But when it comes down to it, you do have to put it into the real world to really see if it works.

Oftentimes, the philosophy that we’ve used for that is still valid in that I want to build the model. I want to make predictions of that model that we can now go and look for. And we can look for places where the uncertainty is highest or our margin is the least so that we can maybe go back and ultimately update that model. Because the model is able to be far more extensible in that synthetic environment to places that we just fundamentally can’t test.

A lot of our tests in the United States happens in the desert southwest. And it turns out the South China Sea is just a very different environment, say, environmentally. And so how do I make sure that the systems that I’m planning to work in the desert southwest is going to actually work in that new place? So yeah, we have a lot of experience doing that on the DARPA side. We had several programs that essentially took us from this synthetic environment and started putting it into the real world. The naive assumption here is that the model that you’re going to get and the AI agent that is going to do this is really designed to just try to fly the airplane better and try and win that game.

One project that I explored that was absolutely fascinating in my time at DARPA was an effort we called Game Breaker, where we basically, instead of using the AI agent to simply win the game, we wanted a whole host of- think of it as an entire family or population of agents- to explore the strategy space and find where the game breaks. And that’s probably the most useful tool from our flight test applications. If we can use these AI agents not just to perform better, but also to explore the landscape and find where they’re going to break, that really helps us accelerate the deployment of these sorts of software into the real world.

Brig. Gen. Douglas Wickert:

Yeah, that’s actually a really exciting thought experiment. We still do campaign planning. I mean, it’s a very 20th century approach. The guys and gals stand up at the big map with a grease pencil and, OK, here’s our campaign plan. Here’s our O plan. And we are a quarter of the way into the 21st century. So actually, I love the– I mean, this really bridges kind of nicely into the chapter three. And Doug, I’m going to put this one to you. So we’ve gone from models. We’ve solved that simulation to real kind of problem. And now we have a real capability.

So chapter three is all about using that real capability for real-time decision-making and real-time execution, not only for the campaign planning to kind of what Animal suggested, like using several RL agents, reinforcement learning, to kind of explore, run hundreds of thousands of campaign plans to actually come up with an O plan that maybe we never had thought of, standing there at the map with a grease pencil. But then once the actual– if the war does start, of course, if we’re really good at what we do, then the war never starts. But if it does, using the models that we’ve built for actually making real-time decisions and helping forecast and inform those decisions.

So Doug, I’d like to hear your thoughts. This really fits nicely with your experience with live, virtual, and constructive. During that real-time execution, how do we use the simulation? How do we use the models to build that cognitive resilience and make better decisions?

Doug Gill:

So I think there’s actually two ways this could go. So one is really large-scale analytic models of campaigns and finding out what works, what doesn’t, what the best approaches are. That’s not my particular expertise, so I’m not going to run down that path. I’m more on the side of training the human and making the human operator very skilled and adaptable. And so there’s a foundational set of skills in flying and systems mastery. And a ladder of training systems and experiences to build up that experience and higher fidelity experiences that allow it to be fully integrated and be employed across the lifecycle of the pilot or the operator.

So what we’re talking about here, I think that the real stress on the operator is the extremely complex environments of scenarios that they’re going to be operating in and all of the many phenomena that occur in them- kinetic, and EW, and communications, and lack of communications, and so on. And so from a simulation training perspective, that is the problem of the synthetic environment that, ultimately, many players are training in. And we’re trying to rise up to a true, actually, coalition-level, multi-domain environment with multi-level security. I mean, that’s the aspiration.

There are a lot of gaps between here and there. But I think it’s useful to break it down into its key elements. So the synthetic environment fundamentally consists of Earth. So that’s the representation of Earth’s surface — Google Earth, CDB, all these things. Then there’s atmosphere, and that’s weather, and that’s all the things that transmit through the weather. And then there’s activity. And so that’s all of the things that are going on in the world classically in simulation. This is the domain of DIS, and HLA, and CAFDMO, and MAFDMO, and all those sorts of things.

So what we’re doing kind of at the frontier right now in- I guess a couple more comments about that. So the Earth domain’s really pretty mature. The atmosphere, weather domain actually emerging. What we have is actually very crude approaches, getting closer to the physics are starting to emerge, but we don’t have them yet. And lots of labor in modeling and simulation has been around the activity domain and sort of the nature of working with federations like this and HLA. So the real innovative edge right now is coming out of JSC. And what JSC did is slice off a particular piece of atmosphere phenomenology, really, and flip the model. So instead of distributing the responsibility for behaviors across the federation, it centralizes it in a massive compute complex. And everybody in the federation consults that central resource. And the net result may or may not be better physics, but it is terrified.

And so you– and so your EW, your communications, your laser range, your line of sight, all these things, it’s truly fair. And it is now possible to create a much more realistic and stressful scenarios in that particular domain. So beyond that, there is — the chasm between what we have in a synthetic environment we have in the real world is still vast. There are a lot of things we can do better. There are a lot of open problems. But that’s the place where we’re really, really making some serious inroads right now.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Doug. You know, the kernel of what you’re really hitting on is that warfighter trust.

Doug Gill:

Yes. And how good are the models- so in the case of autonomy, how good are the models of the autonomous thing that we’re interacting with in simulation? How well is it being driven? And I was sort of coming at this from an operator in an airframe or an operator of an airframe perspective. But there’s a whole other perspective, which is the mission planning and the campaign, which is setting up the whole scenario that that individual is training and operating in. And experiencing it in simulation is- I mean, having those reps in simulation will give them the confidence, the skill, the cool in the real environment.

Brig. Gen. Douglas Wickert:

Yeah, thank you. Animal, I’m going to follow up with this one with you. But anybody is welcome to jump in. So on the concept or on the topic of warfighter trust, some really interesting lessons with alpha dogfight with the warfighter and the progression. That’s really the reason you structured it the way you did, to go from just simulation into actually in-flight simulation. So talk a little bit about the lessons for that. And if anybody else wants to pile onto the trust too, you’re welcome to do so.

Dr. Dan “Animal” Javorsek:

Yeah, I’ll at least jump in. Beaker’s right. It’s funny, when we ended up publishing- because of COVID, actually. We ended up putting everything that we did on YouTube for the alpha dogfight trials, for example. And it very quickly became- everyone latched onto the AI agents and the AI agents flying the jets and beating a human and blah, blah, blah. And that wasn’t actually the intent of the entire program. It’s kind of what consumed all of the public narrative. But really, it was about trust. And it was about warfighter trust and emerging disruptive technologies.

There’s this long history. And I don’t need to consume all the time here. But if you go back to 1939 and you say, how did the mounted cavalry deal with internal combustion engines, you saw that, actually, they were massively reluctant to embrace that. And that’s partly because, if you thought about what it took to be in a part of the army back then, they identified themselves as their ability to ride a horse. And if you took that away, which had to kind of happen in dramatic fashion, when General Herr was defending it– and there’s a long story here. But the point is that, at the time, they were very, very reluctant to give up the horse. Because they had been victorious on the battlefield for 1,000 years or more on that. And you had this very new technology here. And you see a lot of autonomy in AI kind of follows a very similar pattern. In fact, automatic ground collision avoidance and relatively straightforward automatic system doesn’t have any fancy AI and stuff. But the community was very reluctant to embrace that.

I mean, it took 30 years for this to actually get deployed to an all-electric jet, the F-16. And so living through that, Beaker and I both lived through a lot of this. It was kind of foundational that if we’re going to do this, we need pilots kind of on board with skin in the game to do these sorts of things. And fundamentally, that’s what we set out to do with all of these. And so when you think about it, that word “trust” is loaded, as it is in your personal life. It’s contextual, it’s subjective, it’s relational. And it’s very hard to gain and easy to lose. And when I talk about what it’s like to be a test pilot, it’s really about, how do I quantify properly calibrated trust in the warfighters that I’m sending our kit to? Because I don’t want blind faith, and I don’t want this, say, catastrophic pessimism or skepticism in the technology that they don’t adopt it. I need that properly calibrated trust in the system. And that’s fundamental, and that’s something that we’ve learned.

We have fighter pilots really participating in the deployment of this in a variety of different programs right now that I just would say we did not see when remotely piloted aircraft and RPAs and all that sort of stuff. So yeah, trust is exceptionally critical in what we do.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Animal. So in the interest of time, I do want to make sure we get to the- it’s not the sexiest chapter, but it’s actually probably the most important one. So we’ll assume that we’ve made the models, we’ve translated them into real capability, and we’ve used them for all of their glorious things, and it’s doing great work.

We have to sustain what we’ve created. So chapter 4 here is really about the real promise, maybe, of having the digital twins and the digital thread that’s fully around the lifecycle is actually on the sustainment side. I won’t belabor the mission-capable rates across the fleet. They are certainly not what we want them to be, and there’s really exciting, promising things that we can do with predictive sustainment, predictive maintenance, better supply chain, optimized supply chains.

So Mike, I’d like both you first and then Jonathan to kind of address this. How, as a department, how do we use the digital model, the digital engineering, into the sustainment side of things, as well as what’s necessary to really kind of hyperscale that data, because there’s potentially lots and lots and lots of data, really, to get the mission-capable rates and everything else that we need. So Mike, you first.

Mike Pedaci:

Sure. I’ll take a stab at it. I think the department does a fantastic job at knowing what’s going on, the hyperscaling of the data. I think the Air Force, in particular, has done a fantastic job there. I think the next question, the next layer, is why is it happening? And that’s where, when you are kicking off a program, that is the time to put in the contract language that essentially pushes industry to create digital twins. Because right now, we say, yeah, you need to use model-based system engineering. However you get there, get there. So if we, as a department, were to go around to all the PM shops and say, where’s your digital twin? Change the culture where we’re really driving behaviors around no-kidding digital models at the beginning. It’s harder to do retrofit, right? And we were just talking to one of your colleagues about that, which we’re happy to do. I mean, we think we need to do, but it’s much easier if we can bake that into the front of the acquisition through contract language.

And it’s a journey to go on. It’s kind of like how you got musicians to give up all their music rights, right, with iTunes back in the day, right? So part of the challenge is getting industry to play. So one of the conversations we’ve been having, because we do so much in the defense industrial base, is remote execution of models. So you keep your data, you keep all your secrets off, but we can — you’re soft, but we can go into the force, into the mission planning, and we can pull in models of airframes, of subsystems, of thermal, and from multiple different suppliers. And the best way to do that is to make that part of, I think, the language. That’ll drive the change in the culture in industry.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Mike. Jonathan, your thoughts?

Jonathan Hudgins:

Yeah, I mean, representing Google, I’m going to go back to the compute, the infrastructure, that perspective. This was a stat I looked up recently as I was just thinking about the magnitude of the hyperscale compute that Google is building, investing in. So my answer is economic. This year alone, so in 2026, and this is public information, Google is investing $200 billion in its compute and network infrastructure just this year. That’s a billion with a B. So I was like, OK, what does that mean to something I can relate to? So then I was like, oh, what’s the FY ’26 NDAA? So that’s 22% of the defense budget is going to compute in Google. The Defense Department will never match that, never. You’ve got to buy F-35s and stuff. You’re not going to match that. But you can use it. You can leverage that investment.

And so that, I think, is from where I sit. When you think about sustainment, how do we, as algorithms grow, as we move more towards digital, that is going to require more compute. For these guys to be able to do the things that they’re expert at– you said hyperscale a few times. You’re going to need a hyperscaler. And that $200 billion, that’s coming from your YouTube subscriptions. That’s not coming from taxpayer money. So leverage that would be my suggestion.

Brig. Gen. Douglas Wickert:

All right. So we’ve got about five minutes left. Just enough time to really explore this. So some of our adversaries, I won’t name them, have this forced civil-military fusion. You can use your imagination. No respect for intellectual property. We obviously respect intellectual property. But having closed gardens where we don’t share things and don’t have interoperability across this digital landscape that we’ve been talking about is not going to win the war for us.

So to each of our panelists, in about a minute or so, we’d just like your thoughts on, is the industry ready, or how are we going to all get together? Nations go to war, not armies. So how do we ensure that we have open, interoperable, so that we don’t end up with vendor-locked models that don’t talk to each other? So we’ll just go right across. Mike, we’ll start with you, and we’ll go down the line.

Mike Pedaci:

I mean, we’re a very large supplier, but we’re not the only one. We don’t want to be the only one. I mean, we have tons of APIs across our hundreds of products that we have opened up and will continue to open up. We use SysML wherever we can, OpenUSD. I mean, we’re all about open standards. And I think, in partnership with the department, to create some of that language that gets to the trust, where companies who are spending their IRAD, Internal Research and Development, can feel like, well, I’m not going to lose my IP here.

I think we can get there together, right? And we see that as a responsibility, as a company, to work across industry, because we do business with so many folks in the defense industrial base. But I totally agree. And I think it’s something that we’re in it for the long run. And I think I speak for the rest of industry as well, that we’re getting a lot of receptive folks on the other end when we kick off these conversations about remote execution. So that’s it.

Brig. Gen. Douglas Wickert:

Animal.

Dr. Dan “Animal” Javorsek:

Yeah, I think when you look at historically the way that we’ve tended to acquire our systems, it’s always been kind of vertically integrated. And it’s not really served us well in the 21st century. I think the fact that the department is leaning into separating the hardware from the software and allowing this best of breed was something that we, I would say, have experience in in our daily lives, certainly in our domestic one, all the time.

And it’s something that we’ve kind of taken for granted there that just has been slow to be adopted on the Defense Department. I’ll also add, from a small software company side of the house, the only way we survive as a company and as a business is with precisely this. Because if every contract was a winner-take-all piece of hardware that had to have this, then we really wouldn’t be here. And I think America and the warfighter and the taxpayer benefit from the democratization of precisely this.

It’s really hard to democratize the hardware piece, but the software one, if we’ve seen anything in the 21st century, it’s that that is where you get best of breed. It allows– you have to have an environment where they can compete, a place where they can offer those services up. And that’s what I think is the path of the 21st century. It allows us to get back in front of the tempo limitations that sometimes the hardware chains, just supply chains, don’t afford us to do.

Brig. Gen. Douglas Wickert:

Yeah, thanks, Animal. Duster.

Jonathan Hudgins:

Yeah, open source is in Google’s DNA, right? I mean, go all the way back to Gmail, Kubernetes, everything that everybody’s using, from hyperscale to the edge. That was a Google development, Google open source. TensorFlow, JAX, all of these frameworks that everybody’s using is Google. Inside baseball, how we- I talked about that compute architecture, how our data centers are linked, and then how, basically, the global network.

So Google’s global network serves 25% of the world’s internet traffic. Each network line has four shards, so quadruple redundancy within it. Each shard is built with an open framework so that Google can rapidly switch out suppliers. So if there’s a supply chain issue, new technology comes online, one company looks better than another, we can swap it out. And each shard has a different supplier within that shard.

So if there is a fault or a cybersecurity incident or whatever, we can shut down that shard. There’s three other shards running, so the traffic keeps moving. So it’s absolutely embedded in our DNA and how we operate, and it’s how the buyer should operate.

Brig. Gen. Douglas Wickert:

Thank you. Doug, you get the final word.

Doug Gill:

Oh, amazing. Okay, so FlightSafety. We are probably one of the three largest operators on the planet of simulation. And so maintaining cyber-physical systems at scale across decades is what we do. And so we have an in-depth modularity that crosses time and technology cycles and supply chain disruptions. So that’s all kind of built in. The legacy is actually kind of bespoke, and our full flight simulators have been poster child for legacy monoliths, which they are not. They are not legacy monoliths. They’re actually highly modular things. And in terms of embracing the standards so that we can all play together, it’s, I mean, we’re full embrace. And particularly on the synthetic environment side, there is no, there’s absolutely no other way to do what we need to do other than full implementation of open standards in terms of the simulation platform and the avionics models and flight models and so forth.

Again, fully open modularity. The Air Force is doing a lot of great things in terms of nudging us towards meaningful models, putting some significant semantic frameworks out there in the case of government reference architectures and kind of uptaking and integrating all of that stuff. And the push towards open and open standards, one of the things that’s happening organically inside of our company is, okay, these are actually better ways of doing systems and modularity than the old bespoke approach. And okay, so one of the evolutions of the system is an evolution to those open standard-based modular approaches, which will then play with this ecosystem.

Brig. Gen. Douglas Wickert:

Outstanding, thank you, Doug. Well, I’m excited. I hope you are as excited as I am, everybody. This is probably the defining characteristic of the 21st century. It’s here, and I’m confident that we’re gonna get that right. So please join me in a round of applause for our panelists, Mike, Animal, Jonathan, Doug. Thank you very much.