Techstrong Gang – September 9, 2024

Happy Monday, everyone. You know, it was only a matter of time till someone sue Broadcom over this VMware subscription pages. We've got that.

We've got AI transforming app modernization for Cloud Native and is, it's 10 o'clock. Do you know where your LLM is? It might be at the edge.

You're watching Techron Bank. Hey everyone, welcome to Textron Gang. I'm Alan Shimmel, and happy Monday to you.

As I said in the opening, we've got some really great topics we're gonna go over here. But first, let me introduce you to our Monday morning Textron gang. We've got some, uh, some regulars and, and some really smart people.

I'm happy to have him on. Let's lead off as I usually do out to the voice of Silicon Valley. He's our roving editor, John Swartz.

Hey John. Welcome. How are you?

I'm good. Good morning, Alan and Panel. It's a big day in the Bay Area today.

We've got the iPhone event this morning in Cupertino. Then Monday Night Football with my beloved Niners hosting the Jets. What more could you ask for For the Niners to kill 'em?

That would be real nice. You Jist fan? Okay.

No, no. I'm a Steelers fan. Oh, okay.

Okay. Anybody but the Jets sometimes don't, don't tell my jet friends, fans. The friends that promise.

But anyway, good luck and can't wait to see what comes out of the iPhone event, though. I think I'm gonna skip this cycle. Yeah.

Um, also joining us from high in the foothills of New Mexico. It's, she's an open source Maven founder, CEO of Deploy hub, Tracy, Tracy, Reagan. Hey, Tracy.

It's great to have you back. Hey, Alan. Thank you so much.

Always glad to be here and start the week on a Monday doing the gang. Absolutely. It's good to see you.

And then joining us from El Dorado Hills, California Sporting the F1 shirt. He's a CTA at a Futurum group with also just a really, really bright guy with his finger on the pulse of the tech scene. Dave Nicholson.

Hey David. Nice to have you on. Thanks for joining us.

Thanks for having me. Always good to be here. Absolutely.

All right. Um, then joining me here in our book of Raton Headquarter Studios. You know, I didn't even watch the Yankees last night.

Football started. I did. They win.

They lose. How'd they do? I didn't check myself.

So it's, it was not there. Yeah. This is what happens when football season starts.

He's our chief content officer, Mike Ard. Hey, Mike Hing, how you doing? Good to see everybody as always.

Yep. Alright, let's jump into it guys. So, Mike at t would may have made a move that may break the dam.

I, I suspect suing Broadcom over the support fees around this VMware subscription hanky panky. Yeah. John has a story about this on, uh, Textron ITSM, but I read the, the filing and I was kinda like, de surprised it, it almost read like, you know, something out of a RICO case and maybe, you know, they're short of accusing Broadcom of extortion.

But John, what's your take? Are you, are you referring to the bullying tactics that they mentioned? The very one?

Yeah, exactly. Yeah. So at and t did what I guess many others would like to do if they had the economic resources.

But in a sense, they're active. They're, uh, suing Broadcom for trying to retroactively rewrite existing VMware agreements. It force at and t to pay for extensive expensive subscription based services that's gonna cost them.

They claim hundreds of millions of additional dollars. Um, the, the two big concerns from at and t is that Broadcom is attempting to retroactively alter existing agreements, which would significantly hinder at and t's support services, uh, would lead to software security vulnerabilities, soft system outages, and even national security threats. Because at and t works with a lot of customers who are police firefighters and government agencies that depend on VMware software for communication during emergencies.

Um, I talked with, uh, the CTO of amenity. I always have a hard time pronouncing that, uh, Eric Helmer. And he said that they are getting a ton of business from people who are extremely upset with what's going on with Broadcom and VMware.

So they're acting kind of as a, as a, as a go between and, and fi filling in. Um, he basically talked about this idea of, uh, vendors locking in customers. And so this is a consequence.

There's been a fallout from Broadcom, VMware. Here's just yet another consequence from that merger. And it's gonna be interesting.

I think you're gonna see other people, I'll throw this out maybe to Mike 'cause he reaches at the VMware show. But this idea that they're gonna strike back at the Empire, I think Alan has it right in that there'll be more people joining this suit. But David, um, I don't recall seeing these kinds of lawsuits being filed and, and with this level of, uh, transparency into a relationship between a vendor and, and and their customer in the IT space.

Um, is this gonna become more common, do you think? And, you know, what's your impression of the suit? I don't know if it'll become more common.

It's, it's possible. Uh, I think it's, it'll be pretty straightforward if it contracts were attempted to be undone. That's pretty straightforward.

And so what I read, what I go through the complaint, especially when you start tossing in national security, now they're upset that their cost structure is being played with by a company that acquired a vendor of theirs that they feel is too painful to move away from. And, uh, and, and I'm sorry, but at and t not the, not the most sympathetic figure in my mind. This is at Seat Swen orphanage down the street complaining about price increases.

Uh, I would be very surprised if Hock Tan and Broadcom messed up and didn't realize they had an ironclad contract in place with at and t. Uh, I can tell you from firsthand experience that every aspect of pricing and contracting was absolutely scrutinized. And all of this is by design, uh, on, on the part of Broadcom.

So, um, I think what we're gonna find out is if anyone has, uh, committed an FTL error, which is a failure to lobby error, we're gonna see how much political power at t has, how much political power Broadcom may not have. And that's how it will be sorted out, frankly, in this case. So I, I hope this won't affect at and t's stellar customer support.

Uh, Ma Bell, I still remember. I still remember Ma Bell. So yeah, the, see the idea that they are complaining that someone has a little bit of leverage over them in negotiating over tools.

Uh, if I were Broadcom, I would put folks from Nutanix and every other organization telling the world right now that it's super easy to move away from VMware. I put them on the stand. So, So interesting.

Dave. That's what Eric hel. That's exactly what El Eric Helmer said.

He mentioned Nutanix specifically because people are gonna be looking for options. Well, no, because Nutanix, Nutanix is, is advertising, Hey, we make it easy to move our VMware north super easy. And That's, none of that actually works, you know?

Right. I mean, you know, all it does is I can swap the vm, but the entire data set and the databases and everything else that goes above that VM that I gotta move is a painful months long process. Oh, no.

No. Oh, GCP Azure AWS, they would all disagree with you mightily. It's super easy.

It's super easy. In fact, I can get you, Hey, Hey, I do it. I Think it's super easy to swap out the vm, but all the stuff that sits above it, I, I'll put you in touch with people from Accenture who will tell you it's $750 per VM to move your VMs for you.

And they won't even ask you questions about what the I will, I will put you in touch with Nutanix customers who told me how months of weer went into switching over to Nutanix from vm. Put put, Let's not get with spending match over here. Here's what I think though.

I'm being sarcastic. Hold on, hold on, hold on. I'm being sarcastic.

It's incredibly difficult. Oh, what I'm saying, what I'm saying is, what I'm saying is that the world claims that it's not, this is what I'm saying. Right.

And, and that'll be interesting theater in a courtroom. But I'm telling you right now, this lawsuit will never see the light of day of a courtroom. This was a leverage thing to bring some public pressure onto Htan and Broadcom to do something.

Right. And you know, I'm sure there are lawyers representing at t who have visions of class action suits in their head. But that's not what this is going to be.

This will be settled in a they're Negotiate. You're right. They're gonna negotiate some settlement.

What I mean, can I something Just a chi? Yeah. Are you saying that somebody's abusing our court system to settle a score?

No, That's what the court system is there. The force Compromise. It's all about buddy.

Yeah. It's the force compromise. And so that's what you're dealing, that's what you're dealing with.

No big deal. I think we have to remember too that the government is probably one of the biggest consumers of VMs. So how will the government step in and what, what is, uh, Broadcom brought to the government sort The contracts they don't Pay for, they don't care what they pay in subscription fees.

I would, they have, they have budgets. They care. I'm kidding.

I I sar think his Dave Yeah. Know. I'm just curious how the government is being approached by Broadcom.

Are they doing the same thing? Which would surprise me. 'cause they're of course customer of Broadcom.

I mean, David VMware will claim, or Broadcom will also claim that they're the last company that had a perpetual license and then everybody else already moved to subscription. So why is everybody beating them up over this thing? They're just kind of following everybody else's pricing structure.

That would be my argument. I think. Uh, I think the Hawk tan argument though, behind closed doors is going to be pound sand.

Uh, he's, he is, he's, he's not intimidated by the public pressure. By what at and t says now the US government. That's an interesting question.

And, uh, I don't think that they have the same pricing issues with the US government. Well, well generally the U US government has no pun intended, most favored nation kind of pricing with vendors like this. I, I used to sell a lot to the US government.

We had to have separate SKUs for the government product. Um, but, You know, given say, climate of the, the, the government and it stands towards a big tech, you think that you, that, that that's the last person or entity customer you wanna rattle at these these days too. Yeah.

But, but don't think that there aren't a ton of companies out here on the sidelines cheering on at and t and saying, yeah, I hope you get something 'cause we're right behind you. Sure, Sure. And I would, and I would say if you are, if you're looking at the friction associated with switching away from VMware, it's even greater.

Going back a little bit in history and looking at some of the titles from Computer associates that are now Broadcom titles in the mainframe space In the mainframe. Mm-Hmm. So, well no, they've maintained that mainframe business.

Yeah, certainly. Yeah. Um, David, how good are, how good are the third party support services that people might look to that John was talking about from Reney Street?

They do support for SAP and Oracle for other folks. And there's a, a lot of alternatives out there for these support services, but it's not clear to me that, um, they're have the lion's share in the market. And so what's your sense?

So why don't people just rely on them more when they run into these issues? So, uh, when you say third party, third party in that they do not have a, an official relationship with Broadcom Or they, they're, they're third parties, Um, kind of stuff, right? They, They have their own experts and their own support services, and they, they have their own squads that you come in and you don't rely on VMware, you rely on them.

Yeah. Now, again, just to be clear. So this is not someone who has a contract with Broadcom to provide support.

So, right, true, true, true. Third, third party. Uh, there's, there's always a benefit to having a, a hotline to the factory and, um, and being able to, and, and by the way, I'm wearing my, my factory, uh, papaya, uh, F1 shirt for McLaren because VMware sponsors McLaren.

So I'm sort of telegraphing my bias here. Uh, full disclosure, uh, I actually worked on the acquisition of VMware on behalf of Broadcom. So, so I, you know, so I I I have a bit of a dog in the fight on this one.

Um, all right. We have a potential witness on the show. Awesome.

You're gonna be used to see nothing else without the advice of counsel. Exactly. Yeah.

So, so good. Yeah. So, hey, You know, mentioned it, I mentioned this guy Helmer, the CTO Remedy Street.

He mentioned they were getting a lot of calls and complaints about this matter, but it wasn't just specifically about VMware. He also mentioned, to a lesser extent, they've gone through this with Oracle and SAP. So I just wanted to throw that out there.

There Look, Oracle's another company that, you know, yes. They're know, they, they don't give away ice in the winter to Eskimos, as they say. And, you know, and there are certain companies that have that kinda reputation in the market.

Oracle's one of them. Broadcom is another, But it was mentioned. Yeah, but you know what?

Things get done in the back room. Maybe at and t extends their VMware licenses and there's a tit for tat. And, uh, you know, hot, hot tag gets free broadband at his house for a couple of years or something.

Who knows? But it'll be interesting. I wanna see if other companies, At some point, the cost of the lawsuit exceeds the cost of the support.

But this is a, this is a fight for principle. Oh, I Principles. Is it The lawyers winning no matter what, where do That, the lawyers always win.

Don't Worry about, where do I put that on the balance sheet? I gotta figure that one out. It's On the principle.

So, And maybe there should be a discussion about moving to containers and off of VMs anyway, you know, maybe this is a story In about two minutes. Oh, Don can take a break here on text. We're gonna come back and we are gonna switch up to maybe we should be trans to containers and Kubernetes.

And, and how does AI go into that? Tracy, we're gonna come back to you. But right now, here's a quick break on Tech 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 Textron Group.

All right, folks, we're back. And as promised, we're gonna be talking about well, application modernization, which is sometimes also code for migration. And we have an article this week up on cloud native now talking about how gen AI is gonna make this process a little bit simpler for folks.

So maybe we could accelerate the transition from, as Tracy noted, those monolithic apps running on virtual machines can maybe become containers a lot more easily. And they might even be written in entirely different languages. Tracy, how you doing?

And what's your sense of the state of gen AI as it relates to app modernization? Because I keep hearing more and more stories where people are rewriting entire things and moving from Java to rust, and they are moving from monoliths to microservices, then they're using Gen AI to help 'em out instead of hiring, um, you know, 40 consultants who move in for two years. Yeah, it's a interesting, um, application of ai.

And for those of you who watch this show, you know, I kind of beef on AI a lot, say, come on, it's overused. We never know when we're supposed to use it. But in this case, I think there's two areas that, uh, this gen AI will begin showing, um, kind of its use.

And one is in really understanding how to break down a monolith into, uh, a domain driven design with services. That is an area that companies struggle with. They do it badly.

Uh, they take a long time to do it, and sometimes it doesn't get done. So if we can put that step first, that will be, and gen AI can help. That will, that will be super transformative in terms of moving off of monoliths.

The other areas in the back end, I see the people struggle with, which is really the operational side of it. You know, how to manage containers, how to, uh, to optimize settings. And these things change all the time.

So if we could start employing AI to help in these two areas, I do believe that people will be able to move to a modern architecture much faster than we're doing it today. Because just breaking out the monolith across your entire organization. 'cause you can't just take one application.

You're a bank, you don't just take one application and try to break it down into domains and to services. You have to look across all of your, your applications and do that. And this is where, um, you know, natural language processing can really excel.

And I'm really hoping to see more of that as well as understanding configuration parameters, tracking that, and being able to auto, you know, auto repair. Uh, when, when things are going south. Uh, because right now we have a lot of SREs that sometimes struggle with tracking so many containers.

So those are the two areas I think AI will do the best in. And moving off of, uh, of, of VMs will take some time, but this is the first step. And, and driving into the, the breaking up of those monolithic applications is the first step.

And if you do that well, even the production side becomes easier. App modernization is code ai, assisted app modernization is code. I I think this is a great use case for AI because yes, it, there are complexities and, and you know, again, depending on the organization, enterprise, large enterprise, it'll be more complex, less complex, but, and, and how much they're actually, you know, uh, moving.

But I, I think this screams for an AI type of, does the, the question is, do you, do you create sort of a, a custom LLM for this kind of task? 'cause I think that's, you know, we, we, we talk about AI and is it working? Is it living to its promise, blah, blah, blah.

But I think what we're gonna see is, I don't wanna call it mini ais, what you're gonna have custom l LMS that are very good for specific tasks, and you run that for this kind of thing, and it, and off you go. I, I, I think, and there'll be a lot of business around that. David, do you think that this might reduce the app modernization fear factor?

We've all heard the story. I've got some cold ball application that some guy wrote, and he's probably living on a beach somewhere and isn't taking my phone calls. So I'm, but I'm too terrified to mess with this app.

So can I get in there now with Gen AI and kind of understand how the code was written, how it's structured and, and converted into something more modern? I have to assume that someone is seeing that value proposition. And the reason why I say that is because it's frankly shocking how little pushback I hear against this idea of implementing this new thing.

I, I call it a pluie, a plain language user interface. When we went from CLIs to Gooeys, oh my, there were people who just absolutely couldn't stand them for their lack of flexibility in their lack of accuracy compared to ACL I. The response was, yeah, but look so easy.

Look at this. Look how efficient. And people drag their feet, drag their feet, drag their feet, and now what do we have?

We have a mix of sort of CLI, if you will, writing code and some gooey stuff. People seem to be rushing headlong into this because of the weight of all of that legacy code. Uh, but, but on ab ad modernization, uh, I've become kind of skeptical about, about the subject.

I think that the, that the area that will be truly modernized is narrower than most people represented to be. I think most things will sort of wither away and die over time. And then everything that's new will be developed in a moderate context.

And, and there's that narrow area where you've got this geo VMware based landscape that you're converting. But a lot of that work can be mind-numbingly complicated. So yes, if AI can prove itself, and yes, it will need custom, contextual, very customized models just to understand what it is you're prompting it to do.

Um, it, it will definitely help. And I, and, and, and again, I'm just shocked by the lack of pushback, Tracy, that's, I'd love to get your thoughts on one thing that David just said there. Um, yeah.

Is it easier to build a new app to replace the old app or to modernize the legacy app? 'cause what Dave was kind of pointing at is it's getting simpler to build a, a new app. So why not just get rid of that thing that's been sitting around for 20 years?

No, I would agree. You just build a new app. No doubt.

Absolutely no doubt. In fact, it's, I think you're, you're, you're, you're ahead of the game. Um, when you do so, uh, just simply because some of the problems that you originally built into it, you can, you have an opportunity to move it out.

Uh, we've done that at, uh, open make software and at deploy hub many, many times. Instead of, you know, trying to fix something. We just, you know, tear it all apart, sit, let it sit on the garage floor for a while and pull it all together in a new better way.

So yes, I believe that is, and that's why I believe this, uh, you know, gen AI works for this because you can take a monolith, you can look at, and again, the, the, the LLMs are gonna be a problem here because we don't have a lot of data on the operations side, but on the code side, we already have tools that are generating code. So I believe that we can potentially have, have the ability to look at a, a organization's code base and sort out how you can build services, small independent services, and then generate most of that code. So it's coming.

And that is the way systems will be modernized. And I, I believe that most companies will start retiring old system and rebuilding from scratch. So, you know, I'm, I'm reminded of a conversation I had eight or nine years ago, I think it was at IBM think if they called it Think then they might have called it something else.

It was with the newly minted CIO at Hertz. Hertz had just moved to their brand new worldwide headquarters in es ECE Florida, which Neil Fort Myers. And they were going un uh, going under a huge cloud migration, but it really wasn't a, a migration, it was an extinction event.

What they were doing is they would build from scratch a cloud-based application, and as soon as that application was up and running, they just shut off. The old application, didn't migrate anything, just went shut down out with the old and in with the new, throw the baby out with the bathwater. Shortly thereafter, Hertz went into bankruptcy.

That CIO was fired, right? And, and chaos in suit. And it, it, and it, and if you've ever used, and I, I use them all and they're, none of them are great, but Hertz is it, and app system is far from great, even today, though it is Cloud-based.

Um, I don't know if that's the right answer. I think history teaches us something there. I don't know David, um, technical debt has been with us.

I sometimes refer to Jacob Marley as the patron saint of it, technical debt. You wear the chains of IT decisions that you forged in life forever in the afterlife. It feels like I'm, can we actually with AI maybe start reducing that technical debt to get to down to something that's a little more manageable than it is today?

It's definitely going to help. There's, there's, there's no question. Uh, but I think that, uh, it's still going to be a problem.

And for the majority of that technical debt, sometimes just let it wither away and die is the best solution for that technical debt. So I think AI is going to be more helpful in the creation of net new things. And there's gonna be a gray area between this idea of, of, of replatforming refactoring migrating.

What are you actually doing? You're dealing with technical debt. It's just that you'll be using AI to pull sort of the essence out of that old stuff to create net new stuff.

So the short answer, which I'm completely incapable of getting, is yes, it's gonna help. Right? So coming back to your lawyer discussion and the previous block, if I have the ability to reverse engineer things a little bit more easily, um, does a vendor have the leverage it used to have?

Or can I sit there and look at that vendor and say, you know what, I'm gonna leave you. If reverse engineering was illegal, then it's illegal now IP is still ip. Just because AI can reverse engineer faster than a P person can, doesn't mean that your IP rights are any less than before.

But, uh, I can reverse engineer my stack that sits on top of your stack and move it to a different stack. Theoretically, if it's all your IP and you're not infringing on my ip, there You go. So I just need a lawyer to sit next to my at modernization team and everything will be fine.

You never get one lawyer whenever there's one lawyer, there's at least another, that's the way it works. Alright, so I guess the question is out there to you guys that modernize your apps for cloud native using ai and it's coming and, but I, I don't think we're gonna run into the IP issue. I I, you know, I don't think most people are lucky necessarily to do that unless you wanna get out of paying subscription license fees or something.

Yeah. Or I, that's when that comes. Alright.

I just wanna migrate faster from one virtual machine to another virtual machine that runs a little bit better on Kubernetes or something of that ilk. I can get there. I think we'll see.

There'll be ways of doing it. All right, let's take a break here on TechOne Gang. We're gonna come back for our third block today and last of the week.

And it's, uh, LLM at the edge first of the, oh, the first of the week. Excuse me. Yeah, it's only Monday.

Yeah. Try that again. If You want football season's here and I'm, I'm the All right, I will outta sorts.

Okay. You're watching Text Drum gang. All right, folks, we're back.

And yes, we're talking about some more futures in the land of it. And this time we're talking about, well, our large language model's gonna find their way out to the network edge. And there's an argument that says that, um, if I have code running at the edge, I kinda want it to be processed and analyzed, or the data at least at the point where it's being created and consumed.

And so making API calls to the cloud for every little thing is not sustainable. Um, we are seeing smaller, uh, inference engines starting to emerge, and language models themselves now come in t-shirt sizes, small, medium and large. And some of the smaller ones can actually make it out to the edge.

And so David, what's your sense of, um, a what's going on here at the, the Edge when we see this happen? And how long, and are we gonna have this kind of diversity of language models that are out there that are some gonna be managed by your traditional IT operations team? Yes.

So I think that, uh, absolutely the edge is the frontier. Uh, is that a cliche to say the edge is the frontier, uh, for where LLMs will live. Uh, they'll get smaller, they'll get more customized, um, in the very near future, having an LLM running locally on a device at the edge, whether it's a mobile device or, or, or a PC, will be akin to having a calculator app on your phone.

Um, to do those calculations on a mobile device. You don't need to make calls out to the internet. You don't need wifi access.

Uh, and then the question becomes, you know, what does that balance of, of, uh, data storage and processing power that's necessary at the edge, what makes sense to run at the edge? But the key thing you, you referenced is this idea of customization. Salesforce is, um, is, you know, making a lot of noise around its large action model and ag agentic AI stuff that they're building, um, with, uh, with open source code.

But they're also telegraphing to the world that it's extremely complicated to do this kind of customization yourself. So they're saying, Hey, come use our platform and our tools and we'll make it easy for you. But so we talk about l lms, I think the really exciting stuff is gonna be in the form of large action model stuff.

So small large action models, uh, so that what you are prompting on a mobile device or on a device at the edge, the processing power necessary to actually go out and do something as opposed to simply generating something, uh, will be a game changer from a productivity perspective. So yes, look at LLMs at the edge. Look at what we'll call large action models.

This idea of a agentic ai, it's all getting pushed out much to the delight of the AMDs and Intels and, uh, and snapdragons of the world. Absolutely. You know, I I but LLMs at the Edge are not gonna make extinct LLMs at the, at the core or, you know, at the central.

Nor will it, you know, I I envision the future. You have LLMs on device, you have LLMs at the edge, and you have LLMs back at the core. And they're, and they all serve different functionalities.

We are, we're launching a new TV series here on Techstrong TV called the Connectivity Cloud, and it's sponsored by our friends at CloudFlare. And you know, they buy into this in a big way, right? They, they think LLMs at the edge, computing at the edge is, is the future.

But they also recognize that's not to say that the core cloud data center goes away, but that we all live in this world. We, you know, we never seem to have enough computing, power storage space, you know, places to do computing. So we're gonna need it on the device, we're gonna need it at the edge, and we're gonna need it in the core, and we'll probably find other places to put it too.

So stay tuned for that. It's called the Connectivity Cloud. You'll see it, I, I think it'll debut next week on text, on tv.

Heads up. I'm co-hosting that with Mitchell Ashley, and it should be a great show. Tracy, I'm not necessarily buying the argument from Salesforce that says that AG agentic is too hard, and you shall leave that to professionals like them.

0 and it's just adding the reasoning engine to a, um, rag process and creating an agent that's trained for a very specific task using the data that's at hand. Um, I don't know, what's your take on it if you looked at this just yet, but I know you've kind of looked at RAG and it seems like it's just the next logical follow up. Well, you, you, you hit on several points that are my, my my kind of my complaint points.

First of all, why are we using sometimes, are we using ai? And we have to question that. And I haven't looked at this.

I, I've, I've, I've, I've read some articles on it, but I don't consider myself an expert, even though I did stay at a holiday in last night. But the other thing is, is the agents, I'm, so I, we we're seeing too much reliance on this, the agent technology, and I don't understand why in particular Salesforce needs to do that. That was my question for, for Salesforce, why do we have to have an an agent for this?

Because that's another type That's, we always have agents, I, we have to stop. We, we, we just have to stop doing that. And the other thing I wanna say is, uh, Alan touched on something that we haven't talked about on this show before, and that's a multimodal l LLMs, you know, we will have LLMs talking to each other, uh, and that I think will be far more interesting, and that's where it'll be pushed out at the edge.

And in terms of, um, who's gonna do this, I really believe logistics companies have a lot to gain by having these small edge LLMs that are taught, are able to talk to each other with different, uh, topic information. Um, having been worked in warehousing and logistics for some time, when AI first came out, I thought about how warehousing could be completely transformed, uh, with, with, with LLMs and ai. But of course, what do we have to have?

We have to have agents. No, I really wanna know how we could do it without having agents on every single edge and every single, you know, compute node that has to, um, interact with the data. That, that's the question I feel like we have to answer because we're just expanding our, our, um, attack surface every time we put an agent into our environment.

But I've seen this, you know, I was in security for 25 years, right? And we went through this, do we need another agent? Why do we need to have another agent?

And then people start saying, but wait, it's a light agent, it's a lightweight agent, very small footprint. It's still an agent. Then people say, okay, that's not good enough for you.

Agent list. We're going agent list. It's just a piece of code we're putting on your computer, but it's really not an agent.

We're agent list. It's still an agent. An agent by any other name would still smell as rock.

Well, We gotta think of a better way to do things. There's gotta be a better way to do it. All right, David, let's paint the future a little bit for a second here on that point, right?

So we're creating little agents that are trained for different things and we're using RAG and LLMs to create these what may, you might consider the next generation of agents, but they're smarter and they do better things. Um, the issue is how do I orchestrate all these agents? And if all these agents are sitting around on different platforms, am I gonna create a workflow around that?

And then finally, um, I could have five different agents doing the same thing and they're all trained slightly differently. So how do I know which agent might be better at something than the other agent? Kind of like a real estate agent, right?

Oh, I'm sorry To ball. I'm sorry. I'm sorry.

I thought that, I thought you were pitching me and you thought I was a vc. Is this, that's, it sounded Like, so How do I do all that stuff? Agent To roll the wall?

You, no, you, you have Mike, Mike, you, you, you've got my attention. What, what, what is your solution for it? I've got, I, I wanna invest in this.

I, I think that that's what's gonna have to ultimately happen. It's gotta be some sort of orchestration engine, but I haven't heard anybody really stand up and say, these are how we're gonna orchestrate this stuff. So we Are a, there are a bunch of very, very small outfits that are getting investments, that are looking at exactly that.

And, uh, and to the, to the broader point, the co the, the, the question of what constitutes a Gentech AI versus running agents in different places, people will make arguments that, well, no, no, no, don't worry about it. We're not gonna be putting agents everywhere. Uh, but, but yes, you're gonna have all, you're gonna have varying, um, you, you, if you have, uh, if, if you have versions one through five of something, um, orchestrating them, testing them, optimizing them.

Uh, think about the classic AB testing, uh, when testing consumer preferences out on the web. Imagine having five agents, if you will. And again, if I, if I, if we're not using the term in the same way, uh, you know, point that out to me folks.

But, um, but imagine having these sort of virtual assistants that are programmed to do things in slightly different ways that you can rank against one another. Uh, the classic example, and again, it's, you know, because Salesforce is in the news news very recently on this subject. Um, they're talk, they're talking about being able to do something like say to this virtual assistant, uh, send an email out to everyone that fits this category and have a conversation with them about this thing.

And they go off and autonomously carry on conversations on your behalf. Now having five versions of that happening, probably a good idea at first to see who is, uh, who's getting the best results. So I, I, I think we're talking about two separate things though.

And here, here's my point. I think on one hand there's absolutely going to be an opportunity for a Kubernetes of, of ag agentic ai, right? And maybe Google has something in open source right now, or open AI does, or anthropic or it's somewhat, the other issue though, David, that you hit on and you touched on is in a war of AI agents, how do I pick what's the best one for my given job?

Right? And that's no different than how do I pick what cloud provider do I want to use, right? Microsoft's good at, at this stuff.

Uh, Google seems to be pretty good at Kubernetes and, and glad cloud native and AWS is good at this, and Oracle is good at that. The, that the market will ferret through and you'll say, Hey, if I'm doing this test, this, this is the best ai. And, and that's just market dynamics at play.

Uh, Tracy, any of that making you feel better? No, in fact, you know, it doesn't, but in terms of Salesforce's, uh, you know, their, their rag, um, I, when I don't trust it yet, maybe that's, that's just me. But I'm, I'm, I'm getting to the point where I feel like, do I really want to put my customers under that kind of a process or prospects?

So I, I think maybe for me, I'm having a trust issue and I understand some of the agent problems that we're having and the security issues around them. So for me, I think it, uh, is I feel like slow down, let's make sure this is working correctly and let's not be hearse and put something out and, and, you know, try to upgrade to a new technology and find out it's not the best way to treat our, in our, our customers or end users. In any case, This is like autonomous driving, right?

I mean, I've had autonomous parking and somewhere driving on my cars for a couple years. I've never used it because it just gets me a little freaky. And, um, I, I don't, I don't trust it.

I guess at the end of the day, you're right, Chay, it's a trust issue. Yeah. But I think we'll have to overcome that or be left behind this dinosaurs.

Yeah. And when you think that one outta 10 people driving on the road is impaired, how much worse could it be? This, this is true, this is also true.

I hope those impaired people are using autonomous driving Crazy. Alright, um, well that was a good way to end it, I think. I think.

We'll we're gonna call a wrap on this version of Text Gang. It's been a great Monday. Um, Mike, you're gonna be with me out this week in Austin?

We are. We will be a j Rog Swamp up. There you go.

And we are going to, I think we're going to do a, a, uh, Textron gang from there. Probably be Wednesday's Textron Gang. Right?

Well now that you said, so we are, yeah. Okay. We'll be there.

Um, David, you traveling anywhere, Uh, headed up to Portland to go visit Intel? So, uh, looking forward to that. John, I think you're off somewhere this coming week.

I will be going. Well, let's see the time. I, yes, I will be going to Dreamforce in San Francisco and I some interesting interviews that are being lined up, including one with Will, I am of all people who, who's very big participant in tech shows.

And I actually might go to Boston for something, but I'll talk to you about that later. Very cool Chi, enjoy. Tracy, how about you?

I am staying home. I am grounded probably until, you know, there was a couple of shows I thought about going to and I just decided to stay at Stay Home this fall. Haven't done that in a long time.

Good for you. Cube Con though. No, no Cube Con even.

It's too big, Man. All right, well, we'll, we'll this is too big. We'll keep wraps on it.

Hey, at all of these places, you, we will have text drug coverage, you can check it out there. And, and for those who don't know, David does a fair amount of, of writing reporting videoing over at six five Media and Fu and group sister company. So do follow him and check him out there.

But until tomorrow, hey, we've got a full day of text Drunk TV following the gang today. So stay right here and see what else we have for you. But that's gonna end it for the gang.

We will talk to you tomorrow. Everyone have a great day.