All right, Scott, over to you to kick off our, our industry panel. And uh we've got a, we, you know, we got some big shoes to fill from the, from the uh groups this morning. Uh But today I'm excited to uh uh moderate a panel uh with these three steamed uh gentlemen, and uh I want to quickly introduce them and then we'll jump right into some questions. Um So first, uh right to my right, I have uh Nick Sp uh director of public sector presales at Snowflake. Uh next to him, Howard Levinson, uh former general manager area vice president at Data Bricks, um and now advising responsible for strategy for them. And then uh finally over there, Garrett Smith founder and CEO of Reveal A I. So we've heard quite a bit today. Um I think starting with the admiral, you know, he, I think he, he did a really good job sort of giving us a focus on focus on mission. And then, you know, in the later things, I think Mattis did a really good job on tying, you know, data needs back to the mission. And we want to spend our panel a little bit today on the application of some of these data concepts uh specifically around like decentralization and A I and, and those types of things and how they relate uh in, in the mission space. Um So, Garret, I'll start with you. Um So we heard today um that data is the new ammunition or, and you know, data is the future of war, those kinds of uh hyperbolic statements. Um What do you see as the importance of data to decision uh dominance in the battlefield? Yeah. So I'm today, I'm I'm gonna fill the the space on the panel potentially as the voice for uh deploying uh data and very specifically processed data uh to enhance decision making at the tactical edge. Um I I have a strong passion for uh conceptualizing and then developing and deploying technologies that facilitate better, faster decision making right at the point of friction uh in the battle space. So I think that uh the the emergence uh and proliferation of uh relatively high powered data processing capability in the palm of our hands in the form of table smartphones and that thing that sort of thing uh coupled with a uh drastic proliferation of raw data feeds coming from a variety of different sensor platforms. Uh All converge at the human cognition level to uh make decisions. And unless that data reaches a decision maker in a digestible actionable format, uh we're very unlikely to see good fast decisions in contested environments. And, you know, as we were preparing for, for the um the panel, I think you and I had a really good conversation about two weeks ago and you were giving some examples from your time uh in the marines. Uh Are there any examples where you've seen, you know, sort of the need for data and, you know, sort of like matisse shared maybe a lack of data or um you know, having access to the data that you need it at the tactical decision point. Sure. So uh we've had a couple of uh discussions thus far about uh the timeliness of, of data or the right types of information or intelligence getting to decision makers uh allowing for good and fast decision making. Uh at the, at that tactical edge, uh rewind about 10, 12 years ago, I was operating in the southern reaches of uh of Afghanistan. We were the farthest south uh NATO units in the country and we were preparing to conduct a raid on a town called Barra at the very southern end of Helmand province, right on the Pakistan border, we knew uh about a week, maybe two weeks in advance that we were going to be uh uh conducting this raid. Uh So as uh as good marine officers, we generated a bunch of intelligence uh requests that went up the chain to uh processing exploitation, dissemination cells uh and a variety of different intelligence production capabilities at echelons above us where uh intelligence was supposed to come from because we were uh relatively low on the priority stack. Uh You know what we got uh about a day in advance of the raid uh was uh nine year old slant angle uh meter resolution U two imagery that probably didn't accurately represent the area that we were going to operate in. And then uh on the eve of conducting the raid, we had some low resolution Google Earth imagery and one in 50,000 us geologic survey topographic maps, all which were supposed to inform uh safe operating. So when we went over that bem and realized there is a set of canals there and there's actually a village over there and the drug weapons bazaar that we're supposed to be, you know, action is actually over there and not over there. Uh We immediately adopted a huge amount of risk going into that environment. Um I realized this is 10, 12 years ago and in the information age that's eons and eons ago. However, the operating environment really hasn't improved for decentralized edge based tactical operations and the advent of uh uh and proliferation of uh uh intelligence collection capabilities I S R capabilities and data processing capabilities really are going to require applied software development to facilitate uh delivery of uh processed actionable intelligence to people like me 10 years ago. Thanks Garrett. Um So Nick in, in addition to software um and the things that Garrett just described as a need, you know, we've heard today that, you know, data is only really useful if it's quality data. Um what factors, you know, do you feel like going to determining whether or not data is quality enough? And uh you know, sort of if you could, you know, sort of how do you balance the time to make a decision versus the quality of the data? Is that the next question or is that the next panel? Um So I think for, for speaking at data quality specifically, obviously, it does matter and there's a couple of different facets of data quality that we have to consider. And the first is probably accuracy. Um If you're driving home, mention Google Earth. If you're driving home from work today and Google Maps doesn't quite know where your house is. Maybe it's got the wrong street number. That's an inconvenience. But if we replace, I don't know what you drive will say. It's a Prius. If you drive a Prius and we replace that with a cruise missile, that's a bigger problem. We went from inconvenience and slight annoyance to war crime really quick, right? We don't wanna do that. So accurate data is of utmost importance and I've seen you folks through drive Priuses in Wegmans parking lot, you would definitely drive a cruise missile too. Um But that's the first place we have to start is in the data. Is it accurate? Can we use it to make reliable decisions. That's the first thing, the second thing I would say is accessibility is, is that next sort of layer, it's great if the data is correct. But do I have access to it? Can I use that and combine that with other data sets so that I can gain insights and do something with it? We're gonna say this a lot. I'm sure on the panel, I've heard the same sound bite from my two distinguished colleagues here. Data is a raw resource by itself just by itself. Data is useless. It's when it's combined and contextualized that it gains the ability to be an insight or a fact that we can rely upon. So accessibility helps us get to that point. Can we get to that data? Can we make it useful? And the last thing I would say is applicability. Um I'm going to ask a question of my data, what is at this location in Afghanistan? What is in uh me? What's in Loman? What's in all these provinces? What's there and what can I action against it? I need that data to be in a place where I can gain access to it and then apply it to my problem. A lot of the problems that we see in the Department of Defense with data actually isn't on the tactics side. I don't know if you guys realize, but a lot of companies when cookies went away in advertising, they started looking at other ways of identifying individuals that's called open source intelligence. And we've been doing that in the government for decades, but it's relatively new in the commercial space. So with the technology and the data quality things or concerns other things like tactics, I think are also a highlight. Thanks, Nick, you mentioned sort of the open source intelligence. You know, one of the follow ups I had was, you know, what could we be doing? Especially, you know, this group in the audience is probably made up of uh folks who are analysts, folks who are developers, folks who are, you know, are collecting data, you know, massaging it, data scientists, what can we all be doing to make the data higher quality and get it into the hands of war fighters? But you know, in, in a in a better fashion. Um That's a great question. I think it mostly comes down to a process and accessibility thing again, um Data correlates very well. Context is important for data quality. I mentioned that earlier error has a margin error isn't just a yes or no error can be a little bit off or it could be a lot off the wrong street number is more accurate still than the wrong zip code. So there's levels of integrity in that data and we got to figure out what is applicable to the problems that we're trying to solve. If I'm dropping a bomb with a 10 mile blast radius, five ft left or right. Probably not the most critical thing I should be worrying about. So I think the first thing is understanding the process and what the accuracy of the data needs to be. Um sort of surrounding those processes with good data aggregation and context and processing it in responsible ways. It's also really important. I'm sure we'll talk about that later. Thanks Nick. Yeah, I do remember when I was driving this morning. Um, ways thought I was on like the 66 H O V and I wasn't and I'm like, man, that's sort of annoying cause it's giving me the wrong thing to turn off totally different with the cruise missile. Um, Howard. Um So, uh warfare is evolving to include non kinetic tactics such as cyber information warfare, et cetera. How does, um, sort of data dominance, avert, counter protect against those threats, like the totality of all those threats? Yeah. So when I think about non kinetic uh, data, I think about kind of break it down into three areas, the cyber threat, the uh electronic warfare threat as well as the informa, the, um, uh, information dominance threat. And when you think about cyber, uh our adversaries are trying to steal our data or they're trying to disrupt our communication capabilities. Um Electronic warfare again, communication, intercept, um, or, or uh limit our capabilities and then in the information dominance, like there's so much fake news out there. Just yesterday, I used a image generator called Mid Journey and I said, produce a picture of Obama kissing Trump and make it look realistic. And by Golly, it produced something that you could believe. And so those things have the ability to, you know, disrupt our political process, undermine the credibility of our government. And uh we're seeing these problems all over the world today. I mean, it's uh it's never been more apparent. So I think the way to challenge these artificial intelligence bots that are doing this stuff is to have more sophisticated artificial intelligence to detect and respond to these things. You know, in cyber, we have all of these endpoint detect and response capabilities. I think it's time for a uh or I think there already exist uh artificial intelligence detect and response capabilities out there which are capable of detecting whether text was written by a chat robot or image was generated by a uh automated system. And I also think that um these systems have to be real time, the more time that we allow a uh adversary to live on our network. In fact, the, the, the data generally doesn't exfiltrate a network till six months after it's been attacked. The longer that we allow them to infiltrate our network and stay within our network, the more uh problems that occur. The same thing is true for the information warfare. If there are information that's being shared on social media, the longer it perturbate in the network, the more likely it, it gains credibility. So I think you need uh artifi intelligence systems that have high compute power that are able to detect these things that are real time in nature so that we're detecting these things as they occur. And um I think uh we're building those systems but it's a, it's a cat and mouse game and you're always going to be trying to stay ahead of the adversaries. Thanks, Howard. Yeah. Um I, I do like in, in your example of the kissing, sure that you, you know, said to make it realistic, I know having built systems. Um I got a lot of requirements back in the uh you know, olden days and I never got the requirement, the system shall work like, you know, so we always built like, hey, it was red, it had this button, it had this button, but whenever, you know, we always got beat up for, you know, the system shall work. I like this, you know, it should be a real realistic picture. Um Garrett, um You know, Howard mentioned quite a few different, you know, uh areas there. Um and, and Vince um um discussion earlier today with key, he mentioned connectivity. Um And so, you know, I think in order to do some of the things that Howard just mentioned, um how do we ensure that these, all these different domains of these different systems? Talk to each other? And share data amongst them because, and, and what are the values of that? Yeah. So i it's incredibly important that systems uh disparate systems, in particular from a variety of different vendors maintained by different systems. Integrators and mission command system. Integrators are able to communicate effectively together. And I think beyond that in the uh battlefield of the present and near future and long future, uh we are likely to see contested operating zones like operating within the weapons engagement zone in the Indo Pacific uh contemplating a China or North Korea threat and under conditions where our systems are being uh jammed, uh attacked and uh and otherwise disrupted by an enemy. Uh I I very strongly believe in the idea of decentralizing uh capabilities so that even distributed highly decentralized uh operators, be they robots or humans or human machine teams uh are able to effectively uh achieve decision dominance. Uh e even without uh uh consistent high grade connectivity back to some centralized uh processing or exploitation or Intelligence center. And in order to do that, you need to develop systems and capabilities that are flexible and perhaps opportunistically connected but not necessarily requiring connectivity in order to effective uh effectively, you know, give decision dominance capability to distributed uh units. There's something I'd like to add to that like historically, applications have own data. If you write a uh if you use Microsoft word, it puts the data in doc format, it owns the data. I think we need to move to a place where the data is open in any application can access the data. And I really think that the driving factors need to be open data formats. Um so often uh you know, somebody will create some data for you and you basically have to pay them to get your data out of that system. So open data formats that are accessible by virtually any the application where the application doesn't own the data. But the data is, is is stored in a lake house somewhere and then you can, anybody can process it um will dramatically improve the collaboration that happens and the enrichment that the data sets can receive and that kind of goes to the data quality issue as well as you enrich more, you generally will find poor quality. And that's a good segue to my next question before I jump into the next question. Um I just wanted to prep everyone we, you know in about, you know, 10 or 15 minutes, we'll be able to take questions from, from the group. Um So please start prepping questions. I've got a list of about 68 questions. Uh But uh I think they'd rather hear from you all. Um So, so Nick actually to that point around the data decentralization and sort of like breaking data a away from systems. Um I wanted to ask you actually about data ownership um and data sharing. Um Eileen Vidrine, who's the CD A O for the Air Force recently said it's not your data, it's the Air Force's data. Um Why like, and I think Howard hit on this a little bit but why is she saying that? What does she mean by that? And, and, and what is the solution moving forward? Yeah, I, so to Howard's point to everybody's point up here and everybody's point in the audience, we built silos, We had no choice for decades. We had no choice but to build data silos. Some of it was procurement challenges. I brought a new sensor package to whatever base on whatever platform. And so I ordered the entire stack. The joke I used to make at Langley Air Force Base was I bought 42 racks. I never bought a server, right? Because the entire thing is gonna be an entire pod. We had to build those because of performance constraints that we no longer feel in the cloud. There's a lot of complications about on premise data, a lot of complications in the procurement process and constant refreshes that mandated the use of silos and one-time purchases. We've moved wonderfully beyond that these days, we can start breaking down those silos and connecting that data. I had the same, I faced the same when I was down range as well where I've got an aircraft flying overhead that I can't talk to. I'm on the ground, I'm doing a kinetic operation and I can't talk to the airplane overhead with a bird's eye view. I get a product from someone for a pattern of life issue. And I'm trying to ask a question and I ask, can I talk to the person who made this? And the answer is no, those kinds of silos have got to be broken down. I think technology goes a long way to doing that. But policy is also a major player here. Everybody is into data. Now, we used to say people policy and pipes were the three avenues. We had to, we had to correct people are fixed. My mom showed me daily COVID dashboards for three years. She's never logged onto her ipad before now. She knows exactly what she's doing. People are no longer the problem. Data literacy has, has exploded in this country and globally as well. I hate giving COVID credit for anything but that might be something to give it credit for. But the technology exists. The cloud has broken up a lot of those limitations and, and put platforms, Softwares and service offerings that are very good at breaking down those silos and bringing the applications to the data. The last thing we need is the policy piece. How do we enforce and get that going? I think that's where Eileen was headed the data in our environment. The Air Force's data is going to be used by the Air Force. If your unit generated it, that makes you the expert, maybe not necessarily the owner. There are some responsibilities around security and governance obviously, but our main focus should be to share and collaborate on data as much as possible and rely on data from experts instead of trying to generate it ourselves. I don't walk outside every day. Lick my finger and go. Which way is the wind blowing? I just, normally I ask Alexa and then she goes on for 25 minutes, but I also can use the weather app on my phone. There are companies and, and entities out there that know the weather. So go to the experts, get the best weather, combine it with your data and make better decisions. I think that's where Eileen was headed. You and uh fully agreed Nick. And part of this conversation is about uh whether uh flexible dynamic startup. Uh uh and young companies can access data curated, customer owned or otherwise in order to facilitate better product development. And so, uh one of the things that we saw with the rise of companies like Teer and the propriety of data was that they then were able to use that data to build better products. Um Something that, you know, speaking as a young company is very, very difficult to access uh even uh nonclassic data held in government repositories in order to build products for the government. And so there, there's a, there's a friction point here where uh it, there's a necessity to recognize that this data is owned by the government or the service or whomever and that it might live or reside in some classified repository. Uh However, if it's not uh flexibly exposed to the innovation community, uh you're going to slow down the innovation process that results in delivery of better capability to uh war fighting organizations. And so there, there's a contest point in there. Yeah. Um Hold on, I'll get to you in a second. I think actually that's a great point, Garret. And I know that our final speaker after us, um Miss Margie Palmieri from CD A O, you know, I don't know if she'll hit on this in her talk, but I know that the CD A O has been putting a lot of effort into like A I scaffolding and, and trying to figure out how do we create quality data for the Department of Defense? Get that out to innovators because, you know, where, you know, the, the government has done a really good job of generating a lot of data and what it needs help with, you know, I think from industry and from integrators is how do they, how do we take that data, apply solutions and products on those, on that data for them and back in it's sort of what he was talking about with data as a product. Um uh So Howard, did you have something you wanted to add to? Yeah, I was just gonna say you know, the CD A O has come up with this uh notion of data managers and um we're moving from this experience where it's a need to know what to a need to share. I think that part of the responsibility of the data manager is to promote the use of that data, to promote it so that everybody who could potentially benefit from that data in the Department of Defense has knows that it's there knows the quality of it, knows how to access it. And that guy can then determine whether they have, they should have authorization. So the whole idea of this data mash with the Federated governance really plays a uh plays nicely into that. But promotion of the data, I think is really important. Data is hidden away in all these different uh you know, uh islands and nobody knows what's there. And uh as a data owner, you would be responsible for making sure that everybody that could benefit from your data knows that it's there and it's accessible. Yeah, and I uh uh sorry, I'll, I'll bring up Eileen again. I'm gonna owe her royalties on using her name so much on this. But again, I, I was at a talk from her a couple of months ago and she mentioned uh a title job title, I'd never heard before, a data product manager. And it, it's the data manager who owns the data where it lives, the system that it's on, that's one thing but somebody who is deriving products on purpose from that data and advertising that and, and putting a catalog together of those products. I think new job and career opportunities are opening up content, not just data, scientists and engineers anymore. There's a lot more of that, that middle ground that we had to fill in. And I think Eileen's ahead of the curve on that too. Yeah. And I think, you know, hopefully in our, in our final speech, you know, we'll hear some more about that because, I mean, that's a heavy thing in the CD A O as well. Um So we talked a little bit about, you know, a lot of different systems and silos, you know, uh having siloed systems, you know, I was recently in uh one of the combatant commands and they were prepping for some of the discussion that um the, the admiral um was talking about sort of in Taiwan and, and China and, and one of the, the conversations that we had was around contested logistics. And um you know, one of the concerns is, you know, in the past you've always had like if I'm the air force or, or, well, you know, I've got to use my air force cargo, you know, planes and, or if I'm, you know, the navy, I'm gonna use those, like when we have now these combatant command sort of environments, there's like six or seven different systems that we have to get access to, to see what things are in theater. Um, how can the discussions that, you know, that we've been talking about? Gary, I'll start with you. Like the, the topics that we've been talking about, how can we solve that problem and help those combatant command leaders get access to all of the different systems today that, so that they don't have to have a different logins and you know, what, what are some of your thoughts on, on either what's commercially available or what's being done in the government? Yeah, so I'll, I'll lean on Howard's Point a minute ago about um it being incumbent on data owners and system owners to promote the integration, interoperability and accessibility of their various platforms and uh and, and, and quality also the quality line again, you know, and, and access to quality data sets. I'm not well positioned to describe uh a, a potential future of high interoperability between systems other than to say it's got to happen and that uh we've got to do better at joint. Um I come from a Marine Corps background uh intolerably un joint uh is my service. Um I think I've got some marines out in the audience, certainly in the, in the, in the online environment uh uh audience as well. Um Simple things like uh adopting uh common communication protocols and common data transfer protocols and data storage, uh protocols and systems and ones that are open and not uh siloed in the historical sense or highly proprietary, which is the tendency for uh many technology companies and vendors uh out there. Um There needs to, we need to promote a dynamic and flexible and relatively open set of architectures that allows for high interoperability. Um And then I think also pushing capability again, I'm gonna foot stop here or clap or something around this idea of decentralizing capability. And if that means that we narrow down and simplify some of the machine learning models that are deployed to the edge and that may not be as sophisticated as the highly centralized high compute environment models um to use one machine learn example. Um So be it uh so long as you're pushing that capability to the edge, because we know that in some of these contested environments, uh communications are going to be denied. Um Certain forms of metadata typically derived from an I S R platform are not going to be available or they're going to be spoofed or they're going to be tainted in some way. And so we need an a way to uh intelligently promote uh approach that problem and, and we believe very strongly in pushing capability to the edge to facilitate that and it's probably going to uh lean in large part on software. Perfect. Thanks here. Uh I I wanted to break up and see if we had any questions from the audience before I jumped into, you know, a couple of last questions, it looks like we have one over there. Do we have all the microphones with us? Um You're talking about the importance of computer. Um So the question I have is if we prepare data and we get the wrong data at the, a lot of the ETSS we work with, they're not really going to be transformed on it. How do you see the future of edge and edge transformation happening If you actually made an actual decision space as an operator of analysis? A great question. Um I'm not sure that everybody could. I think first of all, he called me young marine on the edge. I, I'm not young. Uh I or at least I don't feel as young as I used to. Uh So thank you for that. Um uh I think the question was how do we deal with the idea of uh uh edge based decision making in an environment where uh the data collected or the data processing capability or the the software that is otherwise uh intended for use in a higher order architecture like a cloud or a server farm uh is not perhaps as performing to the edge. Should I get that? OK. Cool. So um uh I think software and technologies of the future need to be built and optimized for the edge. Um Full stop. I, I don't think we can always live in this uh world where we can depend at all times on a cloud architecture to deliver our edge to to deliver on an edge requirement. I think that uh we very much need to be thinking about how to optimize and build for edge first in the same way that 5, 10, 15 years ago as mobile telephony came online and you started seeing software developments and web access being uh built from the ground up as a mobile first experience. Same thing we need to do that for the edge edge first and then allow for opportunistic connectivity and sharing and, and uh uh re centralizing that data from the edge back to a central repository. But it isn't always about, you know, a general or an admiral at some distant command center, uh fact checking the data before it goes out to Lance Corporal Smith at the tactical edge. Um We need to empower the edge uh dwelling tactical operators be they humans or machines with the ability to make timely good decisions. They may not be perfect decisions, but that's OK to, yeah, I I think generally speaking, we, we see the same problem set in a lot of places and ems is one, it's a, a kind of a an adjacent example in the States. If I send an ambulance out to a certain place, I'm going to examine the treatment plan, what they did, how they did it, how long it took them to get there, the route they used, all of that will be examined after the fact and not real time, pinch that guy's artery cause blood's coming out of. It is a much more imperative thing that probably can't wait for a cloud process to happen away from the edge. So I think the, the, the mentality that a lot of folks are taking here and I'm really interested in it. It's kind of an emerging data science thing is to think about data as a data fingerprint. Uh oil rigs came up with this decades ago. They've been using it very successfully for a long time. If vibration is getting out of, out of place, if heat is getting out of place, shut down this drill, something's about to happen. That is bad. We know the fingerprint of an explosion in our data. And when something starts to look like that, we take immediate corrective action, whatever that action might be, I think this is a place where edge computing really began, begins to solve the problem. And there was a gap and giving us, we know what it looks like when things start to go wrong. Here's how we can very quickly in a very short period of time, take some kind of action to prevent a disaster. I think it has to go there. And yeah, I would just say, um I, I think the model is going to be trained centrally and pushed to the edge. And I look at things like uh GP T four, I think had four billion trillion parameters that it was trained on. Uh my employer data bricks just basically created their own la large language model based on five billion parameters. So uh two orders of magnitude less data, another company took what we had done and uh compressed the model quantize the model and now have the ability to basically run a large language model in your browser. So the ability to take these large language or any type of machine learning compress it down and, and run it in some, you know, device that is in the uh in the hands of a war fighter is really the way it's gonna be. And then we're gonna continue to do all the training on these big systems. Um Yeah, William, they got back from comments here. Some would argue that we're at a time of, of, of of space where we're at a digital uh overload that we've almost become paralysis of decision making because we can't sort through all of what we have access to. So from your perspective, where are we, what's the limitation on being able to do predictive analysis at echelon? So that, that the data that the squad leader needs or platoon leader needs may be the same data that the comma commander needs or the president needs. But we're not waiting on the the analysis throughput or the reach back to sanctuary that we're doing predictive analysis that the systems are enabling us to focus for echelon for that human to make the right decision on the things that the human needs to make in the decision to cut through all the chap of, of, of the excess data that's not needed. I, I would just say right now, the I, I think the biggest challenge across the federal government and particularly the dod is just the number of islands of data there. It's not centralized, it's, it's not cataloged or uh the, you know, the there's no management of the data and the government's, you know, the dod is trying to modernize and trying to uh do all that. And I don't say this as a negative. Like I, I look at the dod, they were the first users of computers, you know, decades ago when a lot of the companies that, you know, I work with today are born in the cloud. They didn't have all of these problems with the legacy of having all of these islands of data out there. But you look at a system like Vanna, I think today they're integrating 500 data sources on a continuing basis. But think about the how many different data sources the dod has, that's just AAA tiny bump on the log in terms of the overall. So we've got to get all the data consolidated or at least knowledge of the data and, and Federated together so that we can start doing training these models to do predictive analytics. It's gonna take us a while to get there. Unfortunately, and I think we're uh as, as you look at the data problem itself, it's not necessarily that we have a ton of data, it's that we have a ton of data that we don't know how it relates to the rest of the ton of data that we have. Right. So we have this massive amount of data. How do we connect that with everything else we have and build stories out of it or or build answers out of it? I think that's a major problem we have to solve. The other thing that I know is being improved on greatly. But the best sensors in the world I worked in cybersecurity for a long time and the best sensor we ever had for a cybersecurity attack was the people. And I think the best sensor that the dod has is a soldier, an airman, a marine, a guardian. Um I, I think that's really where things start to get interesting is when we're starting to capture the observations of soldiers and putting that into one big thing. Blue Force tracker was a massive objective years ago to get some kind of command and control in theater. An expansion of that I think is the next area and how we relate those together and then maintaining those relationships between data so we can make better decisions more quickly. One last thing sorry Scott, I know you want to move on Scott wants to move on. Uh Right. Uh William, I really appreciate this question and I'm gonna approach it. Uh, hopefully briefly, Scott, sorry, uh kick, kick me or have power kick me if you need. Um So I, I'm gonna throw a, a vignette that I hope doesn't come across too much like a, a product pitch for my company. But I, I very much believe that uh the way to approach problems like you just I identified is to is for um innovators, private industry, government developers, whomever to deeply understand the use cases uh in in parentheses, kill chains, kill webs, whatever whatever sensing and effects operations are desired by the by the echelon. Um deeply understand those and then build flexible software and automation technologies to facilitate real-time uh analysis of a context to uh almost preempt the question from the human operator. And here's the vignette, imagine that the four of us are uh outside of Kiev in uh somewhere, somewhere somewhere in the hinterlands of, of Ukraine. Um We have a some sort of robot platform above us uh scanning the area uh almost doing things that are transparent to, to us, but uh it understands what our needs would be on the ground. For instance, uh A a big need is casualty evacuation planning and coordination. It's incredibly time intensive and labor intensive and cognitively intensive for humans to go out and survey for helicopter landing zones as an example. Uh both in premising and during mission. Um So imagine that uh I get shot. Uh Nick's providing security, Scott's moderating a panel and how, and, and Howard needs to figure out where to take me to get lifted out. You know, we're the, the whole golden hour thing you gotta move, you gotta get this, this underway. Uh In olden days, you might just like accept all the risk and let the air the aviation platform decide where to land, that's close enough to you and then you have to move to that location or perhaps, you know, one of us needs to go while I'm bleeding out on the ground. How has got to go and like survey a couple of areas to see if those are viable helicopter landing zones. What if instead our robotic teammate in the sky already has the answer and, and says, and says to us, you know, here, here's or understands from us that there's a casualty. We need CS A and immediately provides here are three candidate helicopter landing zones all within a kilometer south of you, their enemy north. And if and if instead of like living in this abstract world, we build uh automation technologies that are edge based, perhaps even totally humans, human hands free and uh and are deployed at the edge in, in other words, on board the sensor platform and can preempt our questions or our problems with solutions like that. Uh We, we're gonna save lives, we're gonna make our kill chains tighter. Our decision dominance is going to be greater and everything's gonna be happening much faster. Um But it, it, it's about building for specific use cases and it starts with really good product development um and deploying things to the edge. Thanks here, I know we're out of time. I just want to close this, this group with, I think for the audience, you know, I think the lesson you know, that we should all take from this is helping our clients figure out how to convert and translate our needs and mission use cases to come. Companies like these guys are, are part of and, and really helping match, you know, sort of the government needs and the government data with innovative technology solutions. And I, I think that's something we can all sort of take from this as, as sort of our, you know, marching orders. Uh So I want to thank my, my three panelists. Uh I could have listened to you guys for about two or 34 more days. Um I really appreciate the time and the prep time that we had and, and great to meet you all and, and have you all here with us uh with in this event. Thanks for having us. Thanks, Scott, Garrett, Howard, Nick. Thank you very much. That, that was actually a really quick session.