Techstrong Gang – November 21, 2024

Hey, everyone. AI agents everywhere as far as the eye can see you're watching Textron Gang. Hi everyone, it's Alan Hummel.

Happy Thursday to you. You're watching Textron Gang. We've got a great lineup on Textron Gang today.

We've got reports from the field. We've got AI agents infiltrating everywhere, and there's always, there's security concerns just about everything Anyhow, and that never changes. Let me introduce you to our cast of, uh, gang members from today, some of them reporting remotely from the road.

Uh, first off, let's go right to Guy Courier who is at the Supercomputing Conference, a C 24. As you can tell from that snazzy background, he's almost in front of, Hey guy, how are, remember to paint? You painted that by hand.

It looks, it looks like you did a good job. Thank you. Oh, you probably had a, a I agent do that for you.

Who are you kidding? This is super compute, Not AI compute. Anyway.

Yes, they're everywhere. They're everywhere. Um, moving on from Super Compute, we're gonna come back and actually get an update from Guy on the, uh, SC 24, but let's move over to Vegas, where our man at OpenText is on the job.

Uh, Mike Ard. Hey, Mike. How are you?

I'm good. One more day in Las Vegas, and then I'm finally heading home. I hope.

Well, it's been, it's been two weeks for you. It Has been. I, I'm ready.

I'm ready to go home. Suddenly New York seems really attractive. Well, you'll be home for a week.

And then back in Vegas, I got maybe two weeks. I'm thinking. Not even.

It's Thanksgiving and then you're back. No, I thought I had a week after Thanksgiving. Well, we'll See.

No, you don't. It's, it's that week now. Rest.

No error. No, not the Reagan. You and Budy single.

Your five country's trip there. Yeah. Yep.

Next I'd like to introduce you. I've missed her. I haven't seen her in a while On here.

Yeah. I Was in Barcelona. Is that where you were?

That's where I was. I was in Barcelona. It was fantastic.

I got there right after, uh, the America's Cup was over. But there was still a lot of excitement about New Zealand being the first Threepeat, uh, in a Very, they wanted for the third time. Well, they, they, they got the down there.

I love bus slope. Yeah. Great team.

So, uh, did the whole America's Cup experience, brought home some swag. Uh, it was a trail. Hope if I didn't Rec Hope Lynch joining us back home in the Raleigh area.

Yes. Our sailor. A a, um, moving on from Hope from our sailor to our royalty.

Our Sor of Phyllis Zai. Don't laugh though. Don't laugh at me.

I'm not laughing. I'm not LA aspiration. We, we live in a world where if you say it enough, it will be true.

Yeah. It's not the truth. Don't get me started.

Let's not go down that road. That was a blistering edition yesterday. That's all I'll say.

Um, it's was blistery be, that was, uh, rollicking. I think he put it that way. You're, yes.

It was a, a rollick. You know what, if it wasn't true, it would be funny. But unfortunately, it's the, the truth that right now.

But anyway, John, John Schwartz joining us, uh, John, of course, one of the core gang members here on Textron Gang. So guys, thank you all. Um, guy, if it's okay, let's go right to you.

You know, I don't know if everyone out there is even familiar with what the Supercomputing Conference is. IC 24. Why don't you, if you wouldn't mind giving us a report from the field there, but also, you know, a little background for people who aren't familiar.

So, super Compute is an old conference. It's usually called at c. Um, it's a pretty old fashioned conference.

It's kind of charming. They only have one keynote, for example, instead of the 12 or more that seems common these days. Um, it's, you could think of it as a sleepy conference.

Uh, 'cause it's so old. It's very scientifically oriented because it's mostly been about supercomputers and HPC for, you know, going on however many years now, 15 years, something like that. I should have looked that up before I joined the podcast.

Um, and, uh, you know, as recently as three years ago, um, maybe it sort of was a harbinger of what might be coming through the data center or what might be coming to cloud native environments, because obviously high performance computing, very exacting requirements. Um, but then this little thing called AI came along, um, which, uh, let's not get confused. A high performance computing infrastructure is very different from an AI supporting infrastructure, whether for training or otherwise.

Uh, no point going into that, into the details here. But they're very different. They have different and particular mathematical requirements.

Um, nonetheless, AI has been, uh, or HPC has been a critical sort of, uh, birthing bed for AI over the years. Um, and, uh, now is critical to a lot of, uh, HPC. So this show is becoming more and more relevant.

In fact, this year there's 17,000 attendees, uh, which is up from 4,000 last year, which was already up from, you know, 12 or 10,000 the year before. So it's really growing. Uh, and, um, there is a lot of commercial activity.

We were just talking last week about CubeCon and open source and all that other Sure stuff. This is the other side of commercial activity from an open research environment. So, uh, interestingly enough, um, the show this year is less about compute, um, and less about networking, which are long established issues in the industry as much as now it's all about storage and about cooling, which, uh, has I think, a great impact on all the rest of the industry.

As, um, data centers turn into, you know, AI machines, AI factories, um, that's a quote I picked up from one of the, uh, press briefings here. Um, where, uh, energy utilization and cooling of these systems that are not HPC, that are commercially used by the cloud providers or by enterprise, becomes more and more of an issue. So you're gonna see and hear about the technologies here at SC that we are much more quickly going to see out in the rest of the world with a commercial open source, wherever it is, uh, very soon is not already one of the big announcements that here, for example, came from Fon, um, and Soddy, uh, which both have debuted a hundred terabyte plus.

I think they're 20 terabyte roughly, uh, SSDs, wow. This is suitable for cold storage. It's also suitable for basic, uh, uh, uh, data storage in the data center suitable, um, with adaptations that both are targeted to the data center, not so much to ai.

Um, but, uh, nonetheless, uh, uh, soon enough, we will have adaptations for a high speed read and write of, uh, of blogs, low, low level, uh, uh, uh, small sized, uh, data packages and stuff that are used for training or even for inference. Um, so great impact everywhere, but being announced here at Super Compute and first adapted and adopted within HPC environments. The other big thing I mentioned, cooling.

Um, we have lots of direct liquid cooling announcements, uh, from, you know, folks like, uh, cool it or from research, uh, sources. Not direct liquid cooling, but submersion cooling from the Hong Kong Institute of Science Technology. Uh, these are things that are not just related to running these new high wattage systems.

I mean, the latest GPUs from NVIDIA and A MD or something like, uh, 500 watts each. And so now you're looking at a server that's running a thousand watts or something like that, and that needs to be cooled. So, uh, cool plates and, and liquid cooling and all that other sort of stuff is, uh, is gonna help with that.

And then the Hong Kong Institute Science Technology Solutions is immersion one, um, that reduced, uh, uh, um, power consumption by something like 80% in their data center for what's called a low PUE, which is this measurement of how efficient a data center is at supplying power to the IT systems as opposed to having to supply power to cool or to otherwise condition. So, lots of really interesting activity here. Interesting perspectives and need I add, which I will close with quantum computing and the issues, the, the opportunities in quantum and the issues required to bring back to life.

So lots to talk about and lots to look at here. It's good to see this show grow and also stick to its scientific research and sort of old fashioned conference roots of one keynote, one planner recession, and, and, you know, a bunch of simple tracks. Uh, that's, that's, that's an overview Guy.

That's great. You know, hope, hope made a comment on the side channel here that sounds like CES for high high computing, uh, you know, high performance, uh, environments. What I find really interesting is, you know, CubeCon for all the, all the, the buzz and and so forth, they, I think they only had, they had about 9,000 people.

I was thinking about that too. Yeah. This is significantly, this is 150% the size of c**k.

Well, there are two major there, there are two major, uh, super computing conferences every year. There's super compute here. And then there's ICS over in Europe, uh, uh, in, in the early part of each year.

And I think that these have been around for a very long time. And they're, I mean, a lot of the exhibits here in displays, there's this whole academic track here where you can see what the practitioners are doing with these high performance compute computing systems. Oh, I should add another thing.

The fastest computer in the world just debuted yesterday, or, well, no, uh, Sunday. 7 exo flops. 8 exo flops.

It's being run by the Department of Energy. It's an A MD and H-P-H-P-E he system. Yeah.

So there you go. Like, Yeah, no, I, I actually saw a report on that they're using it to simulate nuclear weapons before we headline use. Yeah.

It's just through about what Lawrence Lipper's, just what I, that's what I would want. I see Lawrence Lipper once working on, uh, better that than actually blowing one up though, right? So, exactly.

You know, you're making me think of the movie from years ago where, uh, the kid was the game war games. Let's play Thermo Nuclear War War. Yes.

Matthew Roderick, Um, the number one system in the world. There's two other systems that debut on the top 10 because they issue the top 500 list every year at this time. Uh, yeah.

But it was interesting in that El Capitan is, uh, HPE has now made the last eight fastest supercomputers in the world. Now I remember, I don't know, maybe it was 5, 7, 8 years ago, we were worried that we were going to, you know, the US was gonna lose it. The supercomputer race to the Chinese and or Japanese.

And uh, and here we are. It sounds like still on top with that. I don't know if that's made everyone's lives any better.

Maybe it's making our nuclear weapons better. Well, In theory, it should help us with pharmaceutical research. 'cause there's a lot of chemical applications, Folding proteins and so forth.

Yeah. And it should also help with things like climate change. If you believe in such a thing.

It's also a hotbed for the use of AI for research. Uh, it, it, you know, all the focus on generative AI on it passes over these folks who sort of washes over them. Um, they're interested in, um, the research and analytical meets that they have.

So AI can help, like, when you have thousands of possible scenarios to try to figure out which to test AI helps them determine still with human review, the best five or six to test. So you get got Guided recognitions. Yeah.

Are You, are you seeing any folks from uh, three letter government agencies walking around? Because, Because they usually don't wear signs advertising that you know, but Well, the three letter government agency, the funded EL cap, the Department of Energy or DOE, but there's another, uh, one or two of them that are also gonna be using that system, um, for various three letter uses. Uh, I think 100 to Four letter.

I'm sorry. No, no. Okay.

One other very interesting, um, use that I heard about, and this was a while back, is, um, they've worked on creating a digital double, not a digital twin, but a digital double of the planet Earth for weather, forecasting currents, everything. So instead of wondering what happens, they can speed up, slow down, change the earth, and, uh, figure things out. That would not be possible without these large scale systems.

Absolutely. Hey, but we shouldn't also lose sight, what was it, 120 terabyte. SSD and I sweat on my max.

S**t, I go for the five 12 versus the one terabyte. But, um, that's crazy, crazy to think about. It's very remarkable to me, just comparing to last year, how much of the story here at SC is about storage and cooling.

I'm not saying that there's not networking and, and compute going on all over the place, especially that top 500 list, et cetera, et cetera. But a, a, a great deal of the focus this year is on, um, storage, uh, and, and cooling. I I would say that's AI driven, right?

It's cooling down these monster chips and its AI needs increasingly bigger language LLMs to, to play with, to change. NVIDIA's got Blackwell out right now too. So that's, that's the black between Blackwell or Grace Blackwell on the one hand, and Andy Instincts and my three 50 x on the other hand, you're talking about sort of hybrid GPUs now that handle both AI and h BBC workloads in a, in, in, in a pretty elegant way.

Really cool stuff. Alright, let's take a break and we're gonna come back. Actually, I think we've got Mitchell Ashley just, they bailed him out.

He's, uh, he's ready to come on when you're running again. These things happen, you know, we, we've got a great legal plan involved too. So anyway, Mitchell will be joining us and we're gonna talk about AI agents everywhere.

You're watching Text Drunk then. All right, folks, we're back with the second block and well, AI agents, we were talking about it just yesterday, but there's demos of them now from Microsoft and OpenText, and they're gonna be everywhere. Microsoft is turning copilot into essentially, uh, AI Agent Central.

And yes, you might even think of Revenge of Clippy because there are these agents everywhere and they pop up and they're gonna be in your user interface. And OpenText has a similar idea. There's, uh, 15 things that they're calling aviators, and each of those aviators can invoke, uh, right now maybe 75 agents and they're planning on going to a hundred, but they're gonna add agents every 90 days.

So there's gonna be agents trained for all these different tasks. And we're gonna live in a world soon where every productivity application we use is gonna have thousands of agents that are trained to manage all these different tasks. I'm not even sure I can remember what each of the agents does.

So they're, I guess they're gonna pop up and tell me at some point when I'm doing something, and maybe it might feel even a little creepy, dare I say, but Mitch, you're here with me. What's your take on all this? Well, it's interesting, uh, OpenText is very much taking an approach of, uh, multi-cloud.

I mean, they, one of the big themes is how do we tie all these clouds of content of information, whether it's an SAP or Salesforce, or, you know, you're storing it in your local documents that are part of OpenText. And so they're trying to really kind of blur the lines to make it easier. And they're using AI to do that, to be able to search across that.

And they showed a number of demos, um, and they are, they are delivering kind of predefined agents. And there, there's a lot of look ahead here too. So they're talking about kind of what's coming, uh, underlying all this is what they call their titanium X platform.

And the message was, titanium X is real, it's here. And, uh, it, it's very much of let's start to make things more seamless. And they gave several examples, several demos where it's, you know, we want to do weather analysis on, you know, weather patterns for hurricanes and some historical analysis of information.

So we're gonna pull down information from Noah and different sources, incorporate that with content that we already have. Uh, there is also interesting that in their, in their AI strategy there, there's very much a, you put a lot of trust in us and we will not let you down. We are being very cautious about making sure that all the, all of your content is not going into other people's LLMs training and anyone else's data.

And maintaining a real clear separation of that. Now, as you talk about going across those clouds, things get a little bit more complex, right? With what do I have in SAP, what do I have in, in Salesforce, et cetera.

But they, they gave an example also of kind of walking through a sales opportunity and how they internally are using their own AI agents in their own RFP responses, responses to other people's RFPs, uh, very much using natural language interface. Here's the question that comes into an RFP, feed it into the gen ai. Here's a response that comes back.

Validate it, put it into the, into the response and cutting down, you know, days of work into a few hours or so. A lot of copy and paste right now. But, you know, that'll, that'll get smoothed out.

So it is the age of agents. You've gone from chatbots agents, you know, you're talking about Microsoft and their announcements. Um, you know, AWS has announcements that are coming out too about their own technology.

It, it's everyday things have changed. So, you know, I would like pick some partners to work with. I wouldn't pick any one technology 'cause it's just changing too fast.

I'm not sure that we're even gonna see all these agents, because if you think about it, let's say Microsoft's your primary application and they're gonna have an agent, well, you can invoke an agent that's gonna invoke an SAP agent and the ad agent will go do something and bring it back to you through the Microsoft agent. In fact, OpenText is supporting copilot. So we may live in a world soon where the only agents you see are the ones that, uh, Microsoft presents to you and everything else becomes a headless service because, you know, you're having an AI agent that's gonna go invoke that application, then you're never really gonna see it.

So, um, I think we're playing for dominance over the, uh, front end of the user experience, but I, they can't, they simply Can't all win. So they, uh, OpenText even redesigned their interface. They call it JO deceptive takeoff, um, to be able to better incorporate AI into the experience.

So it doesn't look like it's just kind of hanging on. It was bolted on at the end. Um, to your point, Mike, AWS inter introduced a multi-agent orchestrator.

Um, so, so this is agents of, I mean, agents are really becoming the code name for AI code, right? You can call it an agent, call it whatever, uh, but doing tasks and doing work and doing things behind the scenes. But you're right, it, it is agents managing or processes managing other agents.

It's getting very kind of layered here quickly. So I, I think what we're seeing, right, you know, pendulum swing, right? This AI agent stuff is, is new and the pendulum is swinging way to its side.

And so as we're seeing here, everybody and their mother is not just making an agent, they're making dozens of agents. And, and, and Mike, you're right, the, we're not even gonna remember what all these agents do. And the other thing is, you know, you know the old saying in software, right?

Today's product is tomorrow's feature. Well, I say today's agent is tomorrow's subagent. We're going to see someone get smart about this and say, Hey, you're gonna be a master agent, secret agent.

We probably write songs about that like secret Agent man, and, um, anyway, but this master agent will, will have multiple facets of subagent routines that it can run. It just makes a hell of a lot more sense to do that than to try to manage literally dozens of agents per person. That's not going to Work.

Yeah, you're starting to, you're start, I'm starting to hear that out here in Silicon Valley and in some of the conferences that have, there you go. The concept of the super agents. And I also believe, and I, I don't know Mitch and, and Mike, it, did Satya Nadella refer to the, uh, this new era of scaling laws and things like test time compute, because that's also starting to get a lot of traction among not just people at Microsoft, but Mark Andrees has been talking about it.

Some of those, some of those types of concepts. Um, that was specifically what I mentioned, kind of that, you know, that sophisticated of a conversation. Not, not that we aren't having good conversations, but, you know, going down that particular path, I think this is more, you know, remember the OpenText community, this is all about data access management doc, you know, start starting from the document management days, but there's a whole of course, security layer to this.

'cause they have security offerings with it too. So it, it's, from a general user standpoint, I think it's how, how agents, if you will, in whatever form, if you know they're an agent or not fit into someone's workflow and how you can take content and then use that to create other content or maybe feed into processes. Um, they were, they were using some examples of, you know, how many of these kinds of attacks have I had?

What help me trace down the, the, the kill chain that were happening on a certain kind of attack. And what it reminded me of is we're gonna be writing agents to go find out what agents are running to see what are using too much resources, or maybe they're doing stuff we don't want 'em to do, right? So there's gonna be a lot of na natural language queries just to find out what's happening out there.

'cause everybody's craving agents. I, I think I am seeing and hearing, um, the advent of a, of a new role, a new role in organizations, right? Because I tell you, when I log into applications, I get popups, oh, we have this new agent and here's what it can do.

And I click away from it because at that moment I don't wanna hear about it. I just wanna get my tasks done and reading through all of the instructions on how to use that agent, uh, I'm not here for it at that particular time. But if you think about a large organization rolling these agents out, buying applications, there has to be someone who's gonna take charge of chain management, educate people, teach them how to use it, because they make it seem like it's super easy.

But there's gonna be a little bit of a curve. So someone that's, that's interesting because like, you're gonna almost have to have like a chief agent officer, Someone In a weird sense gets overwhelming, right? The more you hear, I mean, it's, it's gonna be a lot.

And these agents will interact. What are the interactions between the agents? Do you have, you know it Yeah.

So I'm here for agent ai, but does anybody remember when we hated agents, there was this period in, in op systems operations management, and my calibrations management, every project had to come out and say it's agentless, it's agent free, no agent, you know? Yeah. Yeah.

So sure it's easier to manage agents now, I suppose, especially if you have AI agents managing your agents on your AI enabled platform and so on and so forth. Further disaggregate or moving us away from what, what the, I mean, we're all gonna become George Sheen. We're gonna walk into our office and we're gonna have For Agent Smith to show up and say Mr.

Anderson, But why do we hate agents? Mitch, can you do that again, Ash? I would Wait.

I had a frank conversation with the CEO Open. Let's keep Frank. No, no.

And he said, you know, the funny thing about all this is that he doesn't think they're gonna be able to charge extra for all this. He says it's all table stakes now and every application has to have it. And um, you know, it's just gonna be par for the course.

They may not be what Wall Street wants to hear, but you know, AI agents are essentially Free. So the reason we paid a we bid agents was because they impacted system performance. So now, you know, if there's a new type of, I don't, I don't know that there would be, but if there's a new type of a system overload caused by agent spawning and patients going outta it or whatever it, I don't know, but, uh, I just, Yeah, I scratch my head then Gonna be a process manage dream for somebody.

I'm waiting for Hugo Weavings to show up as, uh, agent Smith And Mr. Anderson, The Patriots. Well, but you we'll say the Jetsons.

I don't think it's a Jetsons. I think it's more of a walleye wally, where we're all tubby little guys sitting on chairs getting no exercise or worse yet a true Mr. Anderson where we're just plugged in, in the back of our head into, uh, a matrix or something.

We are batteries out, bro. And lets keep that. I say though, if there is an agent for, for, uh, instant translation, no matter what country I go to, I might sign up for that one.

I don't have to get it on my phone. Uh, give me some little, I don't know, put a chip in my head. A little ear thing.

Like Lieutenant Horror hat. Yeah, Yeah. Something unobtrusive.

Make me seem like a local, have all the accents and everything That Barcelona hangover going on though. I do, I Do. That's a bad.

So all those people who teach second languages are in trouble, is what you're saying. I see. Alright, let's take a break here on agents everywhere and come back and talk.

You know, security concerns, lions, tigers, and bears. Oh my. You are watching Textron Gang, Discover Textron Group, the epicenter of tech innovation.

We are your go-to for reaching IT leaders and practitioners worldwide. Our secret impactful content that sparks awareness, engagement, and top quality leads with us. You'll access editorial websites, streaming videos, virtual events, custom content analyst research, and more.

Join our satisfied clients. Let's revolutionize your tech journey. Contact us today and tell your story to the world in the most powerful way with Techron Group.

Alright, folks, we're back in. com talking about how there's increased adoption of gen ai, but it also says that people are getting a little bit concerned about the implications gen AI has on DevSecOps because, well, we're not sure if the models themselves are secure or maybe they have dependencies and issues that go with that. Hope, hope.

I'd love to get your thoughts on this, but you know, frankly, I was kinda, my positive spin on this is, hey, we're talking About this Much earlier than I expected. So I think that is the positive spin to me. It means that the, the message is getting across.

It's not just people standing on the corner, uh, you know, screaming the end of the world. Everybody is walking by, right? Uh, people are listening and understanding.

There, there is a potential threat there that they need to address. So there is the benefit, uh, for code generation testing and deployment. But yes, understanding how they can maintain the integrity and security, uh, of their software supply chains while using these tools.

Because we know, um, it can introduce vulnerabilities inadvertently, um, everyone knows about issues where you have, uh, a SQL injection vulnerability or maybe there's some insecure authentication. They could, the developers unknowingly incorporate these into applications, um, using generative AI because they are not hands on keys with every piece of the code necessarily, depending on how much they're using these tools. And, um, risks related to dependency on external models.

Because a lot of these solutions are using third party models and even issues related to data exposure risks. Because you may be sharing code snippets or your data with external services, because how do these services learn how to write the code? They're bringing it in from other places, and maybe they're also sharing some of your code out if you don't have things configured properly.

So I think there is, um, there's upside, but there's a lot of risk. But again, yes, it's great that people are starting to pay attention. So I, I have some thoughts on this, you know, and I, and, and, and take this from a security person where it's coming from.

I call bull crap on this, right? This is right outta central casting, oh, gen ai starting in the software supply chain. Hey, roll with solutions with question 26 B security concerns.

We hear this, we hear this about every, that, that's the, the knee jerk reaction every time something comes out, oh, this is raising security concerns to me. This is more, yeah, what's, what's more secure, closed, soft, closed source or open source? Well, we used to think it was closed source.

Now we think it's open source, then we'll think it's closed source. Gen AI isn't going to introduce any more security concerns than we already have in there. We already have AI and agents looking to make sure and test in real time that we're not interjecting sql, uh, or buffer overflows or SQL interjection type of vulnerabilities into code.

There's no reason we can't use what humans use for the AI generated code, right? I think all of the things that AI introduces and is no more than what humans can introduce in as well and no less. This is just typical FUD that we see come in every time in some security companies rubbing their hands right now.

'cause they're about to release their AI security scanner, like NAMIC scanning, static scanning. SCAI now have AI scanning. I think, I think, I think there is some fud yes, people want to, you know, they want to make a sale.

But I think there still is, there is a tinge of reality to it because I know, uh, through personal experience, you can get a little lazy using ai, right? You can get a little lazier. Maybe you're not, But here's the beauty of ai.

AI never gets lazy. And I could use my AI to check my ai, you can, But the human has to do something to ensure that that happens, right? And that is the crux.

You, I, I think as a, as a part of my CICD pipeline factory, whatever, or it's my I and to put some code into my repo or whatever, I automatically run a different AI security scan on AI powered security scan looking buffer, all that flow That, that would happen. Well, here's, here's how I look at it. Compared to what gen gen AI code generated code is introducing security vulnerabilities anymore, or any less, say it's the same.

It doesn't automatically gonna mean that it's gonna be less secure. It's just a different programmer writing it. Different development, right?

Writing code. So, you know, it, it's interesting when I talk with test people about, like, I was just talking with someone from the online grocery shopping company yesterday here at OpenText. I said, so, um, if, if your developers start using generated code coming out of co-pilot or any other thing, how are you gonna change your testing strategy?

He said, I'm not changing it at all. We're gonna treat 'em just like another developer. Um, you know, like hiring a new person on staff.

We don't know how good they are at the security or quality or any of those things. Um, so, you know, the test people are taking it with a, you know, in good stride, why don't the security people, You know, Hope do, do you think that that the approach here is gonna be the same as at least my understanding of a lot of security approaches over the last 20 years, which is a little bit of like, uh, lip service, but basically not a whole lot of investment, um, in, in control. And then when the disaster happens, a lot of scrambling to Tell this, this is guy, this is an old story in, in security.

You know, when people, when people give a crap enough to vote with their pocket books, we'll see better security for AI generated code, not before I, It tends to disaster. It's, it tends to be responsive and reactive too. That's always the best type to sell security.

Mm-Hmm. Right? When Mitchell and I did still secure, I remember coming in one day, I said, Mitch, I got a great idea.

We should go hack people and then try to sell. There are a lot of people out there that think security companies do do that. So More than one afternoon sitting in the waiting room to visit with a customer that they got hacked that day and we're like, okay, we'll be back tomorrow.

And they had a conversation. But, uh, but to your point though, I think, I think some of it is just, um, for security companies, they need to change their software maybe in some minor ways to account for AI tooling, generative AI tooling, how those, uh, On the market demands it not one day before. Yeah.

And the market, and a lot of them are going with you. And this two words, Zero Trust, zero trust of code that gets created by whatever source. I don't care who it is, what it is, I, you know, it's, let's, let's make sure that it's actually you were gonna go all Mrs.

Robinson on us there. I was gonna drew my nicks in for a moment there, but I, I pulled back and just did a, I will, I will tell you the problem from the human side, right? So these AI agents show up and they create this code for you, and they tell, you know, they make, it feels like, you know, this is the best code ever written in the world and it doesn't like say hi, you know, we created this for you and you should maybe, you know, verify this and check this out and run a test on it.

And so people, you know, suddenly trust the AI too much. And then the next thing you know, there's these vulnerabilities in there. So, I mean, it would be great if there was a process and it was automated and everybody did the right thing.

But I wish that some of these AI agents were less, uh, you know, confident and just kind of just said, here's our suggestion, rather than, you know, here's the best Thing in the world. Mm-Hmm Mm-Hmm. Well, the, um, uh, AWS's, um, not copilot, they're their, their coding environment that just went at went GA App Studio, that's the name of it.

They very much have a process of, you know, you bring up your code from the, from the repository. I wanna make this change to it. You give it a natural language prompt, it comes back with this recommended change for you to review it.

It's not necessarily doing a qualitative review. It's expecting you to say, you know, to your point, Mike, is this good bad or do I wanna implement it or not? So there, there's changing code and just kind of sticking it in into the pipeline and just like another developer would and assuming it's all good.

And there's also sort of a guided process to this. And, and until we, we kind of trust what the gen AI is doing, I think we're gonna review it in most cases. Now, whether that reviews any good human in the loop, that's our job to do a good job, absolutely.

Be interesting. Anyway, hey guys, I gotta tell you, this was a very efficient, very efficient day, strong gang today. Kudos to our gang members.

Well, I guess we're getting good at it. Um, we've got a full day of text on TV following this. So as usual do that, Mike Mitch, you guys are, uh, are, uh, broadcasting live from that OpenText event.

So, and that'll be up here on Text Trunk TV as well. So good stuff there. And we'll be back tomorrow with an awesome Friday show of Textron Gang guy.

You headed home today? You got another bio here? That's it.

Yes you go. I mean, I was home for the weekend, but yes, I'm heading home and heading into Thanksgiving week. Yay.

Well, enjoy, and to all of our viewers out there, if we don't catch you tomorrow or, and maybe you're not working many days next week, as many people don't have a happy Thanksgiving everyone. Thank you for watching. Text Drunk Yank.

We're out.