Transcript
Hi Mike Matchett with Small World Big Data. And we are here talking today about, oh gosh, a confluence of several of my favorite things. We're talking about AI, we're talking about hyper convergence. We're going to talk about storage. We're going to talk about, uh, how you make all that, uh, fungible and software defined and be able to move it around and protect it as if it's, uh, a document. Uh, this is really cool stuff. And who's doing that? It's Verge.io. So just stay tuned with us. We'll be right back. Hey, George, welcome to our show today. Hey, thanks for having me, Mike. I don't even know where to start. Uh, because I'm so excited about talking about the AI parts that people might misunderstand that you've become like, you know, open AI part two here. Uh, but let's let's just try to put that in the background just for a second. Uh, so let's talk about where Verge.io Is, first of all, in this VMware, uh, replacement landscape. How's that? How's that going? Yeah, I think the the big thing for us is we are, uh, we've always been here. Basically what we're using, the term we're using a lot now is infrastructure operating system. Um, clearly the VMware situation with Broadcom has driven massive growth for us. We're onboarding a new customer almost every three days. Um, and so that's been good for us. Uh, but what we see customers doing is I call it the hypervisor trap, right where they're just going to swap one hypervisor for another hypervisor. And, and they don't really kind of fix the rest of the infrastructure. And, you know, 2 or 3 years later, they still got basically the same problems, uh, that they had before. And so what we're suggesting is take advantage of this moment in time and let's see if we can rethink the infrastructure to really make it a much more abstract, uh, and more services oriented instead of just all these rigid, uh, tiers within the infrastructure. So I think I think I hear a hint there of what we're about to talk about a little bit more going into services architecture. So, uh, and just to recap a little bit where, you know, when we talk about, uh, what you guys have done with The Verge OS is you've encapsulated networking and you've encapsulated storage deep in the design of it. So it's it's not, um, a bunch of different components that one has to try to get moving in concert. It's one thing, uh, that they can install and run, which makes it very convenient and very resilient as well. I think this is part of the story. Yeah, we call it our, uh, our single code base architecture. So all of this shares all of these components that we're going to talk about, share a common code base. Uh, and so what that what that gives you is efficiency in when those, uh, different sections have to intercommunicate, you know, if you think about a VM writes a piece of data to storage. Well, that's not exactly what happens, especially in a hyper converged environment. It's going to write that data. It's going to make that write request to the hypervisor. The hypervisor is going to connect to something like vSAN. Vsan is going to go ahead and write to write the data. It's going to acknowledge it. It's going to send it back up to the hypervisor, which is then finally going to send it to the VM that made the original request. In our world, none of that happens, right? You just write it. It goes straight to storage because it's a single code base. Yeah. And I also want to emphasize the fact for people who are maybe just catching up a little bit that this is really software defined infrastructure as well, like the whole thing. And one of the great things I like about, you know, your approach there with verge is that you can encapsulate the entire environment in basically files, documents, artifacts that can be backed up, restored, copied and rehydrated elsewhere or in the same place top to bottom. So it becomes something that's very abstract and encapsulated in an object oriented sense. The whole thing, uh, is not just parts of it or not just the VM. Right. And that's that adds a lot of data protection, a lot of other cool stuff. All right. Uh, we've already covered a lot of that. Go check out our old, uh, bio, uh, past webinars if you want to hear more about that stuff. But today we want to talk about the service orientation aspect. You're coming along. So if you look at the world today and everything is about these aggressive workloads called and a lot of them are AI based, so we'll just call them AI workloads. They're not all it's not everything's about AI. Some it's about bigger data, some it's about big databases and so on. But let's just say we've got people with these corporate initiatives to get something out there and support it. And, um, do it fast. And it goes, well, we would need, uh, bespoke architecture. We're going to need to do this or that. Uh, but now with verge, you're saying, look, we can start to bring workload services into the environment and treat them the same way as the infrastructure. So really fuzzing that line. Tell us a little bit more about you just even start philosophically. What was what was the thinking behind that? Well, the thinking is that from a services standpoint, right. Everything should be a service that you turn on and activate, not a new thing that you go by. Right. And so and that's really what we've seen in the I mean, you can use AI as sort of the poster child for this everybody. Last I checked had infrastructure before this thing came out. And all of a sudden now we have to all go buy new infrastructure because of the new thing that's coming out. Right. We got new storage. We don't only just have new storage, we got new different types of storage. We got got a fast storage. You got a deep storage. You got to have archive storage. You you know, it's like, wow. Really? If you're selling storage, it's pretty convenient for you. I don't know how how convenient it is for the customer. Right. And so and you've you obviously got new servers and, and you might, you know, does virtualization work well with AI? That's a I think that's a fair argument. Uh, so you have all these things and we think that instead of doing all that, these should be services or modules, however you want to describe it, that you just turn on. The simplest way that I would describe it is if I needed to share a file with you, I wouldn't go out, buy a MNAs, learn how to use the MNAs, put all the storage in the MNAs, put all my data on the MNAs, and then give you a login to the MNAs. Right? Because we live in a modern world, I would like right click on a folder and say share with Mike. Right. And because it's built into the operating system, that's that's our philosophy at the infrastructure level. So if you want to do when and when and if you're ready to do AI, you click on the AI button and you say turn this on and it just works. Right. So the the deployment time to some extent, you say the deployment time for AI in our world is is zero because it's just there. And so just like, you know, the deployment time was like you installing the OS, if you will. So in this case, let's say you were a VMware customer. You moved over to us, uh, because of the issues there. Now you're running, uh, you got all your workloads running, you're enjoying that. And then all of a sudden you see this little flashing AI button tempting you, and you click on it and you can start doing things. And that's kind of what we'll show you today. And I like how we've sort of tied this together. Suddenly, if someone missed it, that AI as a workload is often thought of as infrastructure dependent. And so now we're bringing it into this world. So it's it's not just a random workload you guys decide to support. It's it's one of the biggest workloads today. And it's very infrastructure dependent in most people's minds. So it makes it a natural progression. Uh, in exactly. That. Way. Yeah. All right. And and you can extend this model. Right. Everything within our environment essentially is a series of services, whether it's uh, um, uh, networking, uh, virtualization is essentially a service that we offer. It's obviously the, the, if you will, the premier service. Uh, and then, uh, storage is essentially a service that you just turn on and start using stuff. All right. So let's, um, let's talk a little then about AI in this environment. You keep talking about turning it on and turning it on. What are we turning on here? I mean, AI, in terms of, uh, an application usually has some, some models, model software, you know, there could be inference, there could be training, there could be the production side of it. Um, there's usually data that might be fed into it. Um, there's, uh, GPUs, of course, and or the heavier compute. Uh, so what are we what are we encapsulating here when we say turn on the service? Well, let's, let's let's just show you. Okay. I think I think that's the easiest way to do it. So a couple of things here. So this is our new release. This is, um, uh, Verge OS 26. So new, uh, new look to it. Uh, we tried to really fine tune the experience so you can move between different things faster. Uh, if you've watched one of the other works we've done with you, uh, there's a new, uh, menu bar at the top. So from from anywhere you can get to anywhere. And then this left menu bar is more, if you will, contextual. It'll change as you drill into certain sections. Right. And so, um, we not important to note, there's probably 2000 different, um, uh, features that we added to this release. A lot of is under the hood stuff for things we're doing next year. Uh, but, uh, this was a major, significant release for us. Uh, the and you can see up here, here's our little friend AI. Right. And I'm going to start there. We're going to leave and then come back, if that's okay. Uh, so just to keep people so people understand the system. This this is a system that I'm actually running behind this other door over there. Uh, you can't see it on the camera. Uh, and I've got a a very basic, uh, GPU card, uh, there as well. So I can't run, you know, a hundred different models and things like that. So I'll, I'll swap some things around as we go. But let's jump into the AI section and you'll see that, uh, I've been really busy, uh, experimenting with this, but, uh, essentially in our world, you create assistance and you give those assistance, um, certain commands, you can you can prompt them, you can do different things with them. Uh, and so, uh, you might in a business setting, you might have an assistant for HR, uh, and all the HR data goes in there. But the only people that can get to the HR assistant are people you designate as HR people. Okay. Um, then you might have another assistant for, say, sales. And you might throw a whole bunch of data into the sales section that only the salespeople can get to. Now, I also want to keep expectations in check here. This. Our intent is not to go replace. We're not going to become an AI company. We're an infrastructure company. But just like in the example I use is just like if you're running macOS or windows, there's a core set of applications that come with those operating systems like Mail and Calendar and probably a task manager. Uh, and you can use those. And for a lot of people, they're all you need. But for, you know, a good number of people, they go out and buy something else, right. And so what we're just showing you is we've set the infrastructure to support that. If you can solve it with what we've built in. Awesome. If you can't just know that the infrastructure is there and now you can go deploy whatever more advanced, uh, things you might need. So let's just kind of. Yeah. Before you go, I just have two, two things either point out or question. So one is that top line. When you say AI, I mean that's at the same level as networks and files and tenants and VMs. So you really are treating these AI assistants as, as a, as a core piece of, of, I hate to say infrastructure anymore, but as a core service of what's going on, which is this kind of requires us a little mental change here. And the second thing is you mentioned a couple of times about putting different sets of documents in these different assistants, which seems very closely tied to how the storage might be managed as well. So this actually makes it more of a of a, of an AI per volume of data kind of approach in thinking. So just to help people get their heads around it. Yeah, it really depends on how you how you use it. But so so these are the these are the assistants. And like I said you're well you might and big organization. You might have this many assistants. If you're a guy that has to write labs for people this is probably what you end up with. But so now we're not replacing models. You can select the models that you want based on, you know, your your design criteria. So anything that's in hugging face uh, you can go get you can see here I've got a bunch of models, uh, loaded. And then I can also, uh, install additional models if I want to. Right. And so you can see right now just by the little green dot that I've got the Gemma three four gig model loaded, and you can see I've experimented with the ten gig and a couple other things. Right. So there's a lot of different things you can do from that standpoint. So if I go back to my assistants here, let's uh, let's take a look at the, uh, the financial and an analyst assistant that I, that I built. Um, so once you go into the assistant, uh, there's a couple of things you can do. So if I edit this assistant, pretty basic stuff. Um, I can if I lower this context score, uh, it limits how far it will go outside of the data I'm providing it. So, for example, if I take this down to a, uh, 44 just to pick a number, uh, it's going to be much more. It's going to give higher priority to the, to the data I give it. Okay. Now you might ask yourself, well how do you give it data. That's a great question. So we have a thing called a workspace. And in this workspace you load your data. And so you can see here I've built uh, three, uh, three financial uh, uh, profiles for uh, these different, uh, these three companies. And we're going to pretend we're. Mike, you and I are an investment firm, and we're trying to decide, uh, whether we need to invest in build direct, sonic, clear, or precision plumbing and, uh, their text files. So you can see these are pretty detailed documents, right? A 53 k text file is a lot of data, actually. Right. You're probably looking at 7 to 8000 words worth of information that are in these different things. And Mike, you and I are, are, are lazy financial analysts. And we don't like to read. Right. And so we've built this, uh, assistant to, to help us with that reading. So, uh, I'm going to just type in here, um, uh, compare build direct precision. Precision plumbing and sonic. Clear. Uh, for me. Okay. And now it's going to go and look. And now it's pulling the data that's in our workspace. Okay. And so it's giving us a quick summary of that data. Right. And so uh, I can say, um. Uh, uh, you know, again, we're, we're lazy financial analysts, so we just want to know which one we should invest in. Right. So. I like how you make it. You know, summarize it for me. But which one should we invest in. Yes. You can use my money. Well because it's it's yeah it's I we we you write the check. I'm just I'm just. Yeah. So so here you can see, uh, and it's opinion. Anyways. Uh builddirect Pro is the clear, clear winner. Okay. Uh, precision, uh, plumbing is a cautious hold. Uh, and then sonic clear is high risk, high reward. So, uh, you know, and you can take different angles. Uh, you could say, um, we like we like to invest in high tech companies. Which one of these Best fits our profile. And so now we'll go and do the same thing. So based on that it's still like spilled direct. Uh, by the way, I obviously made all these companies up. Um, but, uh, Bill direct was essentially think of the Amazon version of Home Depot and Lowe's. Right. And that was the way I designed that Sonic clear was, uh, you know, you've probably seen, um, Bluetooth, uh, uh, hearing aids. Right. Well, the the problem with Bluetooth hearing aids is they, the, uh, audio quality compared to, say, AirPods Pro is kind of night and day difference, right? So I said, this company has invented, uh, something that bridges that gap. Um, and so those are just some examples. And so you can just go in here and, and do just about anything you need. Uh. Let's see. Right. Mm. Uh. Because I'm the one with the checkbook, I understand. Yes, exactly. Uh. And so now it'll go and, uh. There you go. I like that first part. I've completed a thorough analysis of the retro. But the point is, like, now, clearly you could do all of this with ChatGPT, right? We know that. But if this was a real situation that we weren't lazy analysts or financial analysts, um, you wouldn't put this type of information. This is an internal document that we'd be very sensitive, closely guarded. You wouldn't put that out there. Right. And but these guys have mountains of data that they have to churn through. And so they'll do that sort of stuff. And so I've done similar things with other uh, just examples with companies. So I've got one for healthcare where I've got a bunch of, uh, obviously sanitized and anonymized patient data, uh, and things like that. So you can do different things. The what? Again, I always want to be realistic with this stuff. What what we can process today is, uh, text files, CSV files, PDF files, uh, word docs, uh, Excel documents, uh, so, so those sort of things. Right. And it does a very, very good job with that. And I believe that most companies have some, uh, proprietary data that's important to them that they would use for that sort of a situation. We also support source code, uh, so we can do things like that. The product has a, uh, an open AI, uh, router in it. Uh, so anything any client. So obviously, and I should have mentioned, I'm using our interface to chat with it, but you couldn't because of that open AI router or gateway. You could, um, install anything Lem on all the laptops connected at the organization, and then they could use that to get to your guys, to assume you gave them authorization to get to the private AI. Right. And so they'd have an experience that's very similar to what they're used to say with a ChatGPT or a club. Does that make sense? Yeah, because because now you can wrap your hands and ensure that the data that that IP is not leaking, that they're not taking it and putting it out in the world on some cloud instance that in addition, you know, on the IT side that that that data and even the outputs here are being protected. They're reliable, they're being served, they're scalable. And you've got you've got you've got the whole IT stack now in your hands and you can move it forward as a single kind of concept in your organization, rather than having people do shadow it all over the place and leak corporate IP. Yes. Exactly. Correct. So what I'm going to do is I'm going to start up another model for us, and we'll come back and do some more of this. And then I'm going to show you some of the not AI stuff. Because like I said, there was some important, not AI stuff that we did in this release as well. So I'm going to just. Now, the reason I'm doing this, if you if you were a business and had either a a bigger GPU or probably more likely multiple GPUs at your disposal, you wouldn't have to do this. But considering this is running in my house and actually draws very little power, uh, it's a pretty cool. Thing they're not. Are you doing okay there? 39? Yeah. It's, uh, you know, again, goes to the efficiency of the code. Uh, so you can see I stopped it. You see, it went offline, uh, because of the efficiency of the code. This is running on, um, two very small mini servers. Uh, that, like, I don't have my, uh, back from my Storage Switzerland days. I don't have my specialized air conditioner running or anything like that. It's just it's just running. I mean, they they basically kind of look like little Mac minis, uh, to some extent, uh, now. So that's why I did that. So I'm going to start up the Fi model. And so we'll let that get going. And while that's doing that I'm going to jump back to my dashboard and show you a couple things. Uh, not necessarily related to AI, but certainly important. So the the first is, um, as you know, and we did a webinar on it. Um, we have a very good story around, uh, ransomware, uh, not only protection, but probably more importantly, recovery. Uh, there was a vendor who I won't mention who put out an article recently that said, uh, immutability is ransomware's Kryptonite. Now, let me be clear. I think ransomware immutability is really, really important. But to call it Kryptonite is probably a bit of a stretch because you, even if you have it stored immutable, you still need to be able to recover, right? And so that's, uh, that's kind of. And so we've got the recovery part down. And where we've really closed the gap now is on uh. Recovery. I mean, it's on immutability. And so now if I go to so I went to system, I'm going to go to snapshot Profiles. Uh, and I'm going to create uh, so for us the snapshot profile is, is a holder of uh, schedules. So uh, so we'll call this uh mix. Okay. And then we can hit submit and then this sort of now I can go in and the first thing I'm going to do is define some snapshots. So the first thing we're going to want to do is, uh, we'll call this, uh, um. And I can also specify what tier of storage I want to store those snapshots on. I can specify if I want to quiesce the snapshots so I can get, uh, something that's application, uh, consistent. Uh, on this one, though, this is the new feature I'm going to turn on immutability. Now, immutability means for us, the only thing that can delete this snapshot is this schedule. Okay. So you got to be a little careful, like you don't want to have when we get into the retention section. Uh, so let's say I'm going to have this thing, uh, happen every hour. Our, uh. And I'm going to. But because it's immutable, I don't want to retain an hourly snapshot for 30 days because I can't again, I can't get rid of it. So I'm just going to have it retain it for, uh, I actually I'm going to have it retain for five hours. Okay. And we'll set, uh, minimum retention to five. Right. So, so basically what I'm saying is I'm going to have five snapshots that are immutable all the time. And then after five hours that snapshot becomes available for deletion. Uh, so I hit submit, uh, and you can see here it shows up there. And the other important thing is because I do want probably some long term retention. So I'll come in here and say, okay, uh, I'm going to take these. They're going to be application consistent. I'll do them daily and I'll, I will retain those. That's, uh, just for grins. Let's retain those for 45 days. Right. Oops. I didn't give it a name. And these are not immutable. Uh, but they're exactly right. So that way, if I need if I did screw up and, you know, like, 45 days is a pretty healthy, uh, retention period. And I start to the big issue is do you run out of storage space, right? Um, the system will protect itself to make sure it you're, you know, mistake on snapshot scheduling doesn't crash us. Uh, but, uh, you, you know, you can kind of paint yourself into quarter immutable snapshots you want to be careful with. Now, there there is a debate. Uh, there are companies that will do sort of like a dual key. Two people have to log in at the same time. And, and there's others where you can call support and get a super secret password to delete an immutable snapshot. We think that's wrong, right? Immutable means something, and it means that nothing can delete it. Because once you set up a back door, back doors were made to be entered, right? And so we we set it up that way Now, by default, our normal snapshots are read only, right? And I and I also will say that if you have to restore an immutable snapshot, there's probably been four other security mistakes made before you got to that point. Right. So it but it's there in case of a last resort. And of course these guys are always getting better at spoofing us. All right. So, uh, I get one the latest ones, I'm getting stuff from DocuSign, which I have to sign DocuSign all the time, and I'm like, oh, nope, nope, that's not DocuSign. So you got to be careful. Yeah. Anyway, so this is all set up now. I have immutable snapshots. When I apply that to a VM, uh, it'll follow the schedule, uh, for me. And I can have infinite number of periods here. I could just keep going in and defining, uh, different ones. So that's snapshot profiles. The other thing I wanted to show you. Now we have a breadcrumb system here. If I just want to get back to a certain section, uh, and then again, I can switch to anywhere in the product or from this top menu. So I'm going to switch back to the system level. I'm going to I'm going to snooze that alert. So we have a new alerting system that's a little bit more, uh, in your face, so you can't ignore it. Um, so we'll put that there. So we also now have tagging and we can tag at any level, uh, so across all the different resources. So, uh, my screen is not wide enough, but you can see basically all the different tags. You see, I created one, uh, let's call this, uh, so there's two levels with our tagging. So a hierarchy of two, you have a category, and then you have the actual, uh, tag. And again, I'm taking this because I want to use this, this grouping for. For. Other system management tasks. Yes. Yeah. So we have a I also have a new task engine, uh, that you could use these tags in, uh, to, you know, you can say, okay, anything tagged with this, do this snapshot or replicate it here, you know, different things like that. So I'm going to tag I'm going to create, uh, this is going to be a virtual machine tag. We'll hit submit. So you can see here I've got my analyst tags. And now I can go in here, view those tags. And I can say uh. Small world big data tag uh data VMs. And then. I could do storage Swiss VMs. It used to be competition, so you probably wouldn't let me get away with that. So. So now I have two tags right created. And so when I create a VM, I can I can tag those VMs with whether you created it or I created it. But I could also look at it from the category perspective. I could say, okay, show me anything that's a VM tag, and I could see it that way as well or anything. That's an analyst tag. So it gives you a level of hierarchy to to do things. And then that, uh, real quickly, that task engine is over here, and it's really sophisticated. There's a lot of interesting stuff. We'll we're going to start building, if you will. Pre-done. Um, I shouldn't use the word recipes, but scripts, if you will, for these different types of things. But you can go in here and you can see all the different options, right? We can, uh, trigger this on email, uh, system snapshots, virtual machines. Uh, I can tell it to use a tag. This is where that comes in. So I could do analyst tags, for example. And then I can select an action, take a snapshot, power it off, do whatever. Right. So you can imagine the capabilities that we have here with this. And uh it's a little bit up to the creativity of the system administrator today. Uh, but now we're going to go like I said, we're going to go through and define some things that people can do with this, uh, that they could just download and apply. So anyways, I can imagine as your user base grows and the size of the organizations may also spread out within that user base, some getting very large. Um, that this kind of facility is really going to help someone manage something that starts to starts to cover more than one region, more than one, more than one department, more than one organization within their organization. This really is a way to manage things at scale better at that policy oriented thinking and tagging. Yeah. Exactly. Yeah. And so, for example, one of our bigger customers is Topgolf. We are the if you go to a Topgolf, you're actually using il S you probably don't even know it. Uh, it actually it's really good that you don't know it because that means everything's working. Uh, now that I think about it. But but they're going to tag, um, uh, different areas by so that they're, they have virtual data centers, uh, or take on multi-tenancy by time zone. So in the US anyways, it's East central Mountain Pacific and then they're going to tag by state. So show me all the you know, show me all the, uh, sites because you can I didn't show you that, but it was there. If I do a new category real quick, one of the things you can tag, you can create a tag for a site, right. So they're going to tag all their sites in New York with a New York tag as an example. Right. So that that's where that comes in. Um, so let's do one more thing on AI. Let's see if, uh, so uh, let's go over here and go to so you can see this agent is now online. Because I turned I turned that model on before we left. Okay. And let's see let's do a new history. And so then I just go in and ask it now. Oh, okay. That's pretty quick. I was going to make an excuse for it. Um, so you can see this is a big model. This is a 11 gig model. I've got a 24 gig GPU. Uh, and that's pretty quick, right? Um, uh, I can do how does. And, you know, for, for us, the ability to do this. And by the way, the answers are kind of accurate. Uh, that which is also good. Um, and if I go into this workspace, you can see, as you know, probably, but maybe the listeners don't. I write a lot of the white papers, uh, for, uh, verge. And so, uh, those are all, uh, in there. Okay. And so I can, uh, query them. Uh, honestly, this is going to be helpful for us internally because I get sales people calling me every day and saying, hey, you know, do we have anybody that's, uh, you know, I need a summary of how we're working in telco or something like that, right? So I could say, for example, I could say, um. Just sort of a broader AI comment. This could be something that your AI assistant, your human assistant, who knows how to use AI, would do for salespeople. Eventually, we hope that salespeople could do this for themselves and actually ask the question, uh, and and be assured the answer is coming out for them. Right. And so the the so, you know, it's basically a knowledge base on steroids, right? You can load all this stuff in. You could give it to your telemarketers or your outbound salespeople, your inside salespeople, and uh, do that. And it's essentially, uh, and I'm sure you get this a lot, too, but I get calls every day from sales guys saying, hey, you know, do you have anything that compares us to so and so? Right. Uh, so like for example. No comparison. And so it'll go through. And now again this information is spread out across multiple different documents. And it's doing like you would expect AI to do. And pulling that information in, uh, you know, more or less in real time. Right. So, uh, very fast, uh, and on this one, if I look at the assistant, uh oh, I have the context. I thought I loaded the context on that one, but if I wanted to make sure it's only pulling my data, I could lower that. You could also dial in the temperature. Like how many? Um, uh, how creative, if you will. It gets, um, you could also limit. So let's say you're a service provider and you're providing AI as a service to the people you're hosting. You can also limit the number of tokens that they're allowed to use. Uh. Uh, as part of that. Right. So you could say, you know, if you said 100 tokens wouldn't be very much, but you can, you know, let's say you're doing a lite version. You could say, okay, a thousand tokens. And then when that person hits 1000, send them a that could be part of that task engine that we were just showing you when a token limit is reached. And so and so an email that kind of thing. Right. Okay. So so it's very powerful. Again for us it's sort of like I don't want to say it's no big deal because obviously it was a pretty big thing for us. But for us it's just another service that we activate. Right. And so it just it's just there. And you don't have to use our vector database. In fact, there may be situations where you don't or you shouldn't, uh, but you can see the plumbing is in place to support anything that you need to do, whether we provide it or you bring it in externally. No, I think it's worth. I think it's worth considering. If you are an organization that has some initiatives, you know, business level that are filtered to the IT level or departments to say, hey, we need to roll out some AI out that's useful, that's productive, that's protected, that we control, that we can get out there without hiring a whole bunch of assets we can't even hire on the market today. Uh, and somebody could say like, hey, you know, we're doing verge for this, you know, this infrastructure as a service kind of approach to replacing our VMware. And we could just add AI, AI to that. And just and with a few clicks, it doesn't look like it's that hard to set up some models and say, set that model up over this number of documents that we already have in the storage, uh, and uh, create some services that people are really looking for today. And just like quick, quick wins. I mean, in your in your mind, if somebody started with already say somebody someone already has verge installed for some other reason. How long would it take them to actually roll out some AI assistance? Oh, uh, less than an hour. Less than an hour. Right. Just the kind of clicking you were doing there, right? Highlight some documents. Turn on the model. Uh, yeah. Because you're not training. You're not training new models. You're not getting into. You're not running the vector database yourself. That's all baked in. Uh, so this is just a way to really go from A to Z really fast on, on some use cases that you may have. Um, yeah. The only thing you need to do is the only thing I'll tell customers when they do it is you need to give the, um, the vector database because because it is so fast that the, um, the knee jerk reaction. I don't know if that's the right word, but the immediate reaction is, hey, I loaded up some documents up. Let me start asking a questions, and depending on the documents it'll take, it takes a little bit for the vector database to chunk all that stuff up. Also depends on what kind of GPUs you have right. So it's a so. There's a there's some caveats. When you say an hour or 15 minutes, right? No, no, no. That's actually why I said an hour. Right. It's not five seconds after you upload it. But if you if you upload it, if their text files like was in that, um, if they're this almost instantly right, I think I started asking questions like two minutes after I did it. If it's if it's structured like, say, a CSV file or an Excel document, that takes a little longer because structured data, ironically, is harder for a vector database to process. Um, so that that's the only thing I would say there. But but again, let's say you did upload a CSV file. Uh, 15 minutes later, you're asking quality questions. All right. Right, right. And I can see I can just see right out of the gate, you know, in addition to some of the things you're doing here in terms of research or marketing content. Uh, but someone in finance can load up, you know, several years worth of, of past performance and help get immediate help and insight into the reports they probably have to write going forward. Somebody in security who's probably got a number of other tools, a number of them probably already have AI whatever in them, but wants to keep something, you know, sort of constrained and IP protected, if that's the right word here in security, uh, could start something like this and say, here's our security postures for something, and let me just do this in a segmented, uh, what's what's the Air Force word? Compartmented. Yeah. Compartmented way. If we're talking military. Uh, so. Well. I'll give you an example of what not to do. Like, I have a colleague who I think you probably know at a company. I know, you know, uh, who downloaded the company's entire Salesforce.com database as a CSV file and uploaded to ChatGPT so he could start asking questions about it. Like, I don't know if I would have done that. Yeah. That's that's you don't want to train. Train the train the. Yeah. Wide open AI in your corporate IP data. But but you could do that on this and get very, very good results right now. And you don't have to do the extensive like everybody talks about training. You might have to say hey this field means such and such, right. But it's not like you have to teach it all kinds of things. You just have to kind of teach it what the structure is about, uh, and let it go from there. But that's no different than, you know, putting a good together, a good prompt in ChatGPT or Claude or whomever. Right? So it's sort of the same thing there. But if for what I will say is, if customers want to see this, I'll what I like to do with customers is if they'll give it to me is let's do a demo, send me a few text files that are comfortable for you to you send me that. You know, I couldn't do anything bad with like a great one is an employee handbook. Nobody ever reads it. Why don't we ask questions of it? Right. And I can load the employee handbook in, give it a minute and we can start asking questions. Right. All right, all right. Uh, I think we're we're kind of running to the end of our time slot here, George. A little bit. Uh, is there is there anything you'd like to sort of last words or final advice for someone if they're sitting out there going, I don't know, verge. Is that is that something I should kick the tires on or take a risk with? Yeah, I would say, look, the our bread and butter remains and will remain for 5 or 6 years, I bet. Uh, VMware replacement. And then frankly, now we're starting to see other hypervisor replacement do that. Right. Do the thing that's you got a broken leg. It's called a hypervisor. Fix that. But take an infrastructure wide approach to it. And one and the value of doing that is we solve a lot of different problems simultaneously and set you up nicely for an AI future. Right? And there's definitely those AI initiatives there. That's great. If someone wants to then get started on this or take a look at it. Uh. Yep. Uh, if you want to talk to me directly, hit the contact me form. Tell em you want to talk to George, and I'll talk to you personally. All right. Uh, this looks really. I want to use some exaggerated words, but I'll just say this looks really, really interesting, George. As an interesting as my best word for the day on this. But, no, this is really, really cool. Uh, and and I want to I want to get this in on some of my servers here, uh, so I can create some assistance, because otherwise it just takes too long to set everything up as a as a home baker for AI. Yeah. For sure. I love this, all right. Uh, thank you so much for being here, George. All right, take care. All right.
