Techstrong Gang – January 6, 2025

Happy 2025. We're kicking it off here on Textron Gang. Hello everyone.

It's Alan Shimel for Techstrong Group and welcome to 2025. It's our very first, uh, Techron gang episode of the New Year. We're excited to bring it to you.

We have some of our core gang people with us, and we're gonna be talking about what we think 2025 has in store for yet. Um, got a lot to go over. Let me introduce you to our gang members right off the bat.

First of all, joining us from home in Colorado. He's a future vp, analyst, tech drunk, CTO, my friend Mitchell Ashley. Hey Mitchell.

Happy New Year. Good to be here. Welcome.

Hello from snowy Colorado. We had some snow this weekend, so, you know, it's, we're, we're living the good life. I, enough snow's living the good life, but hey, good for you.

It is here and you're a skiing destination. Yeah, I guess it is. It's good for skiing.

Well, um, it probably is snowing by our next guest house too, 'cause she's also high up in the mountains. She's the deploy hub, CEO open source, uh, open source aficionado extraordinaire, our friend Tracy Reagan. Hey Tracy.

How are you? Hey, I'm doing great and no, we're having beautiful we weather in Santa Fe. We haven't got quite the snow yet, so, you know, I feel I've been working, but I, and I feel guilty for not going out and enjoying the, the weather.

It's been so nice. I'm hoping it continues. I hope so too, bro.

I'm okay with that snow. Me too. Uh, I don't know if it's snowed up in New York yet though, but our chief officer ARDS up in Westchester County.

What do you, what's the weather report there? We, we had a fierce two inches of snow that stuck around for 48 hours. So, you know, it was kind of, you know, new experience in New York.

Apparently it hasn't snowed here in any Significant, no volume in a while. The thing about snow in New York is after 24 hours it gets kind of dirty black snow. Yeah.

I, I don't like that. Well, let me, if anyone's interested, we have not had any snow here down in south Florida. Um, it, it has been around 80 degrees and actually around New Year's.

The weather turned nice for the first time and I don't know how long and I was able to go on my boat and, and enjoy some of the beach in South Florida. What, what we're known for here around Christmas, new Year's time. So it was a, uh, it was a great, it was a great holiday.

We were off for about a week. I hope you all enjoyed your time off as well. I know, Mitchell, you did it.

You were under the weather. Yep, a little bit under the weather. I'm doing better now though.

Much better. Not dead yet. Not dead yet.

Yeah. Did you have a cold? I caught one myself.

It's like a five day job. Oh, I got that strep that was going around, so not fun. Oh, you Were sick too, Mike.

Yeah, I caught a cold. I think it was my own fault. I went to the pinstripe bowl and it was raining and so There you go.

You know, a lot of people I know in, in cold weather had the flu strep colds. Yeah. Yeah, it's tough, tough, tough, tough.

I mean, I'm glad you're all feeling better though. So, Hey, you know, I went, the pinstripe ball was at Yankee Stadium and a team from Boston lost, so it wasn't so bad. Well, as long as we're talking sports, you know, I'm looking for the orange balls coming down here next Thursday.

Uh, Thursday night, and, uh, it's gonna be Penn State. We are Penn State versus the fighting Irish of Notre Dame. Yeah.

Who I spoke would not be Georgia, but they did endedly couple, Classic powerhouses Right in. Yeah, Looking forward, looking forward to it. Speaking of Thursday the ninth, we also, of course, you know, we lost President Carter over the break and, uh, he'd lived to be a hundredth throw at, you know, and had quite a productive life after leaving the White House.

There is life after the White House and, um, Thursday the ninth is going to be his like, I guess national day of morning funeral session, whatever. And we did have predict our annual predict, uh, events scheduled for Thursday the ninth. But in, in respect to President Carter and the National Day of Morning hearing, the us we are moving predict to the following Tuesday, which is January 14th.

We'll be sending out emails to everyone who's registered and updating for promos and social media and so forth. But we are going to, uh, tip our hats to Jimmy Carter. You know, it's funny, not so much for what he did as president, though.

He, you know, there was some important things he did in president, including the Camp David Accords, but almost more for just the decent human being He is. He was. And, uh, well, we're not gonna get into politics on the first day of 2025, but he's a, he's a decent human being and outta respect, we're gonna move, predict, um, Mike 2025.

What, what do you, well, speaking of predicts, what do you think? Well, You, you, you said that we weren't gonna get into politics and the first thing in 2025, and since you and I frequently disagree, I must say you're wrong because that's exactly where we're not brand New the mic. Well, I guess Like new year, same as the old year.

Right. But, but I'm gonna start from the far right and I don't know if you've noticed, but, uh, Steve Bannon is outta jail and up to his usual tricks again. And he seems to have targeted Elon Musk and he's gonna bring a boatload of it issues into the political spectrum.

He is already done it with, uh, there's been a fight over these, uh, visas that we give tech workers. There's already been a topic of discussion. I think that's gonna become a much more heated conversation.

He's also starting to tag Elon with, uh, being too close to China, and that's starting to catch some traction out there. So we're gonna see all these issues around semiconductors and everything else become a bigger issue. And then I'll bet two other things.

I bet that he will discover that, uh, AI and EVs are driving up the cost of electricity for the average consumer, and he is gonna start banging on that drum. And then finally he is gonna wait for a big merger to happen and then go after Elon and say, Hey, you're part of this big tech conspiracy and you're basically gonna try to hamstring us all again with the whole conversation about, you know, who's being censored for what. But it seems to me, here's the prediction is that it is gonna be front and center in this political fight that's gonna happen.

And it's not gonna happen on the left. It's gonna happen on the right. I think that's a really good prediction.

That's an interesting prediction, no doubt. Mm-hmm. I think also, you know, we're already seeing, you know, with the Federal Appeals Court striking down the, uh, net neutrality things that the, uh, FCC was trying to do, and you weren't gonna see a lot of things changing on the tech front just because of administration changes and philosophies and supreme courts and courts and all that kind of stuff.

I think it's gonna be kind of a, I dunno if the rough and tumble is the right word, but it's gonna be a, an interesting rollercoaster ride here this next year or plus, I just don't think anything will get done. So I don't know what, it might be a pretty flat ride. I think.

I think we'll have a lot of, we'll have a lot of, uh, um, I don't know, hot air, but I don't know if anything will change. And sometimes hot air is great for disruption, right? And maybe we need a little disruption.

Uh, maybe that's why the country chose the person that they chose. But again, I think it'll be a lot of hot air because incompetence doesn't always get things done. And I think we'll see a lot of that, to quite honest.

That's my prediction. It's true that we may be looking for some more, uh, adult supervision to Mitch's point. If we don't have any of these agencies who are in charge of anything, then you know, who is gonna make the ultimate call?

Because they might just always land on the president's desk, I guess. But, uh, you know, they'll have to have a law. And then to Tracy's point, you have to get somebody in Congress to agree to something.

So, I don't know Alan dysfunction more of It. What do you think? You know, I'm trying not, I'm trying real hard not to be the ma ma mahi or that's gonna take, Oh, come on, come on Allen.

Unleash the kraken, unleash the crack. Unleash It. Haven't we had enough of Steve Godda bandit?

It Apparently Not. This, this, this, this is an anarchist who would be only happy bringing down our way of life. And when are people gonna wake up to that crap?

I mean, I'm no Elon Musk fan. God knows it. If you've listened to me, you know, that's to be true.

But first of all, you can't blame Elon for everything that happens in the world attack, number one. Number two, you know, Steve Bannon praise plays the part of the tri like king, but built his empire by using technology. So he is a fraud.

He's a fraud. And I, I think his 15 minutes are up. And, you know, I think Donald Trump and the administration would do well to just freeze him out as they did the last time.

Right. Freeze him out. He, nothing good comes outta this man for our country or our citizens.

Uh, does big tech need some guardrails? Yeah, probably. Is this the administration with, with the oligarchs, you know, financing it to do it?

Probably not. It's going to be a laissez fair, wild, wild west. We're gonna have ups downs.

We're gonna have the usual chaos that, that Donald Trump likes to wallow in. You know, that. And that, that's gonna be the, the fact, honestly, keep your head low, get through these four years and hopefully people of the United States will, the common sense somehow will return here.

It'll be interesting to see what kind of an audience, you know, he continues to draw. Whether that'll become narrower people that, you know, strongly view what he views or, you know, is it kind of the larger MAGA audience, if you will. Uh, I, um, I remember I love this quote by, um, you remember the comedian Bobcat Goldway Sure.

In popular long guy. He had this quote, he had this thing in his, one of his sets a long time ago. He said, you know, when I was in high school, Done anything recently, Mitch No, No, I don't think he has either.

He said, you know, I found out in high school that I was fifth from the bottom of my class, so I went and found the other four guys and I became their leader. So maybe that's what we have, we have here with Bannon. We'll see.

Yep. I, uh, I agree. But, you know, and, and quite frankly, I, I don't think the US will be the center of government action in regard to tech.

I think it'll stay at the eu. Oh, certainly in terms of regulation. If anything, you'll get less in the us right?

Yeah, I, Yeah, we will not, I don't think we'll drive any of the, uh, new policies that's gonna come out of the eu. And if we wanna do business there, we're gonna have to follow it. So, But, but what's interesting also though, not from even a tech point of view, but from a political point of view, is, is this his classic, you know, political theory, right?

So now you have the right ascendant with all three branches of government and the judiciary for the most part, right? With the pact Supreme Court. And here they are self-destructing, right?

You'll have the tech heads at war with the diehard tri light mags. And so, you know, will they be able to govern with, given that divide? And probably, you know, it's gonna create some chaos.

It's gonna create some paralysis and so be it. So be it. Exactly.

So what are you, something you said, Alan, you know, you used the, the term laissez-faire. I predict, well, let's, let me back up. Okay.

We have to talk about American Airlines. You have to talk about American Airlines sending out a update on Christmas Eve. We, I had a friend coming in from Chicago and it took it, his flight was delayed for 24 hours, and he spent the night Christmas Eve and Christmas morning at the Dallas-Fort Worth airport with 800 other people who were stranded because of that release.

Now, I wanna point out that we have this, John Schwartz talked about this once, and I think about it all the time. Authoritative bias. I feel like we have in, in, in 2025.

We are gonna be looking at this authoritative bias across many industries. Politics. Somehow Elon Musk is the person that we should turn to, right?

Um, when it comes to fixing our government, he has no background in it. But let's go ahead and put him there. I don't know why.

And we have, we haven't tricked that kind of philosophy in that Annette Twitter. I know. It's like, can you think that would be the exact opposite?

But why do we still have after Delta, you know, had their experience with CrowdStrike? Why do we still have third parties updating software? And I believe, and I predict this, that we're gonna see more of that kind of problem in 2025 because we have forgotten how to do DevOps.

We have thrown the DevOps process away because we were relying on authoritative bias. And it's becoming part of our culture in a very odd way. I mean, there was no reason for anybody to ever do a release, uh, Christmas morning, Christmas Eve morning, that was just stupid.

Or whenever they did it, maybe on the night of the 23rd, whatever it did, it stranded a lot of people. But why would they have done that? Why didn't they, why didn't the testing process go through what happened to DevOps and where has it gone?

So I, I really think that we're gonna see a lot of these, these massive dis disruptions because I believe that we have that authoritative bias now in our culture, and it has taken over some of our processes. And in DevOps, that is so true. 'cause now we've seen two airlines have massive shutdowns because of an update done by a third party, which I find to be absolutely crazy, crazy crazy.

How on earth are we doing that At this point in time? We've been doing, talking about DevOps and tightening up releases and having a process for how long? 30 years.

As long as I've been in business. So where, you know, that just baffles me. And I'm afraid that 2025 we're gonna see more of it and we'll probably see a lot more vulnerabilities come through and hit the hit, hit the market.

You know, having been in security as long as I have been, I have a little different view, Tracy. I don't blame the third parties, right? I, Mitchell will tell you, we, we had the brilliant idea when we were selling IDs that instead of, instead of just selling it as software, we were gonna sell, well first, excuse me, we were selling appliances.

Now, there's nothing fancy about the appliances. They were X 86. But what was great about it is we knew what hard drive, what, what network card, what, you know, what software, what, what, what was the foot physical footprint of our appliance, even if it was just a Dell box under the hood.

And then we weren't selling enough appliances and the, and there wasn't a lot of margin selling hardware. And we said, you know what? Heck with it.

We'll just send, it runs on net 86 Linux. We'll just sell software. And we started selling software.

What a frigging mistake. What a mistake. Because all of a sudden we had to make sure our software ran on every network card, every theta hard drive, every configuration, every motherboard, every, you know, back in the bus, in the IRQ conflict era and all of that.

And it drove our support and, and development. And Mitch ran support and development. He could tell you this firsthand.

We couldn't support all the infinite variations, uh, you know, let alone the Dell compact. Hp IBMs what about all the white box stuff, right? Coming outta, Well, in that case, we owned the whole stack operating system up.

So it was a software appliance if you'll Yeah. That's why that caused so many issues. So it, it's The same thing for third party software vendors who are, you know, putting their apps out on platforms.

It's very hard to anticipate every configuration, every nuance, every variable that can be on any particular platform. But to me it's more, Mitch, you remember our friend Peter Fisher at Citigroup back then? Citigroup, Yeah.

CIO at Citigroup. Yeah. One of the three global CIOs at Citigroup.

They used to take 90 days to do a patch Tuesday. So Patch Tuesday came out every Tuesday, right? Or excuse me, once a month or whatever it was the third Tuesday of the month or whatever.

It's, they were always three months behind because they wouldn't roll out that patch until it had been thoroughly tested on their particular systems. They didn't care. There's a middle ground though.

Equifax is a perfect example. They didn't update that, uh, vulnerability for way too long. 'cause they didn't wanna push it out.

But we have gone beyond that. DevOps has actually allowed us to do much more in a much shorter timeframe. And there is no reason why a, you know, somebody like American Airlines doesn't have a model office system where they deploy that those patches before go out to production.

Yeah. But to your point, it's really to your point, it, it, it's, it's not an all or nothing, right? We don't have to live in an all or nothing world anymore.

And that was one of the lessons with CrowdStrike is don't push the update whether you do it on Christmas Eve or morning or you know, the tu third Tuesday of, of January. Don't push it all out at once, you know, incrementally and don't push it in your main hub. Right?

Right. In in Dallas, that, that was sort of the idiocy of that, right? Put it out there on the edge somewhere that if, you know, Fargo goes down, okay, we're sorry Fargo, but it's not the main hub or whatever it might be, sorry Fargo, but, you know, be intelligent about how you run it.

Be incremental. You don't have to push it all out everywhere. And that's part of, if you've, if you're taking advantage of the cloud of the network, that's some of the capabilities that you have now.

You have to be able to architect design things, rework things so that you can do things incrementally. Not everything is easy that way, but that's, that's the world we live in. And you don't build res resilient systems without doing that kind of thing.

But ultimate, I don't think It's authoritative bias though. I just think it's, uh, inertia. I think it's probably the, you know, the last Monday of the month there's a regular update cycle and nobody thought through the fact that that was probably gonna impact Christmas on that Wednesday or the Tuesday when they normally do it.

And so somebody just hit a button because they were supposed to, 'cause it's that day of the week. And I don't think they had a plan. I think they were just kinda operating on autopilot and didn't realize what was gonna happen.

And I think that's all too common. I think there's just a lot of these processes that are, uh, mindless because nobody actually correlates them to, you know, airline traffic. So.

Absolutely. And I, I agree a Alan, it's not that it wasn't the third party's fault. Mm-hmm.

I've, it just like I said about Delta and CrowdStrike, they are responsible for their own systems and making decisions on who can push things out and the process in which it is pushed out. But I do believe that we have forgotten some really basic DevOps principles and makes me really sad. Oh no, It Really Does thing here.

There's a cultural thing. It's Cultural. Yes.

Lemme go back to, I was still secure days I was talking about earlier with Mitch and I, right? So we were one of the first to market in what we called, or what the industry called IPS Intrusion Prevention systems. Up to that point, state of the art was IDS intrusion detection systems.

Now, back then, I was a young whipper snapper. And, um, I thought, ooh, who would be satisfied just detecting an intrusion or a mal, you know, an attack, potential attack. Of course you'd want to prevent it.

So if I had the ability to prevent that attack instead of just detecting it, no brainer. No brainer. Well, no bueno.

It didn't work like that. People were like, oh no, I can't just block traffic worm. If that's important to my CEO, I'd rather suffer the attack than have the CEO get mad at me 'cause I blocked his porn, or something like that.

Right? And Mitch, you know this, it took, Mitch and I were both gone from still secure before IPS became the re gore, you know, the, the Andrew versus IDS. Same thing in vulnerability management.

Is it enough to find vulnerabilities? Heck, why can't we just patch it while we find it? No one perfect.

Yes, hundred percent My stuff. Just tell me, gimme the telephone book of vulnerabilities. I'll patch 'em on my own.

Right? No, I put the patch just be done automatically. At least a poll request.

This says a patches there. Tell you what, the problem that caused this hot down foil here is I reach into my pocket. This thing, this thing, this has ruined software updates.

This has ruined security for updating because we've gotten so used to this thing updating on its own in the background that we think that's how it goes. We no longer, none of you can tell me which apps were updated on your phone when, 'cause it just happens. We have said, oh no, it's just an update goes on all the time.

We don't worry about that crap no more. We don't do our QA testing, we don't do our DevOps. We don't do a digital twin, we don't do a test environment.

My phone updates twenty four seven and it's been the ruin. And it, it's responsible for CrowdStrike, it's responsible for America and airlines. It's responsible for all of it.

Same can be said for all the SaaS applications, which some of those are, Well, you're starting off 2025. A little growly there, Alan. So, no, I'm sorry.

I'm sorry. So lemme take, somebody has said AI yet, so I'll bring it up to your point about author of bias. A lot of, a lot of what is coming out is agents that will do work for you, like updating software, applying patches, maybe doing upgrades.

That's, that's, that's the path that least we're headed on, headed to, you know, are we running into a, a brick wall there of the authoritative bias as well? AI do it, it figured out it was supposed to do it, and it's supposed to know how to do it. So we let it, you know, we, I blanked up.

I trusted it, you know, just to quote a famous movie, Grander, what, what am I gonna tell Fred? Oh, my car. So not his car, but it seems, well, we can, that's the next thing we're gonna run into, right?

If we continue down this good, Mike, go by. So let, let's, let me put a theory out there that you think we'll see in the coming year, an effort to take back more control from these SaaS apps. Because people are also gonna wanna take more control of their data destiny and the age of ai.

They're gonna want to own that data, be able to apply it and train in and use it where they want. And too much of what they're realizing today may be that, you know, in addition to being dependent on third parties too much, that they don't own their data and they're gonna kind of say, Hey, we need to take back control of it. In general.

I think it's all relative to what it is, right? If it's the CrowdStrike agent that runs at the OS level and can block anything if it fails or cause it to crash, yeah, I might want to be more diligent about it. If it's the fitness app that my people are using on their phones, going over my wifi network, I don't care about that stuff.

Or that's sort of an extreme case, but maybe there's a lesser, you know, if, if, if my, so whatever system goes down internally is not gonna stop the manufacturing lines. It's important. I, I think the thing to take back it, it isn't necessarily just a pendulum swing, Mike.

I think what we should take back is how do we make our systems resilient, right? Just taking it back just means we now own it when it fs up, right? We're the ones gonna be on the hot seat, not the vendor.

Um, and we already are with the vendor, but in this case, we're gonna be the cause of it If we didn't properly, you know, do whatever the steps are, do be diligent about it there. My question is, well, what we want to have is, hey, things happen, right? Stuff happens whether we do it or the vendors happen do it, or the, the SaaS providers, whatever it is.

How do we recover? How do we quickly recover, come back up, go back to the old version, et cetera. We're just on kinda this automatic upgrade path.

Then I think in many cases, we don't know what we would do to get back to the last version of a And that was the case with American Airlines. It took 'em an hour. Mm-hmm.

It took them an hour to do a rollback. You know, we, I think that, you know, maybe in 2025 we should revisit chaos engineering and understand that the ability to quickly fix a problem is so much more important than the root cause analysis because we don't know what it, you know, it could be, you know, it could be somebody sneeze and hit a key wrong, right? That could be the chaos that that caused the outage.

Probably not, but it could be, right? It could be just a a a f figure typo and a configuration that takes forever to find. But chaos engineering taught us that being able to quickly respond to outages and prepare your team to do so is the way forward.

And I believe that, you know, I don't think that most companies have really, I doubt if they have game days in trying to be able to teach their staff to, to be able to, to quickly get back online. Certainly American Airlines has make, because an hour is a long time, an hour stoppage means that anybody who has a connection's gonna get screwed. It's about the system.

And that's what happens. Rebound not just, you know, not just points. Right?

How many assists, how many rebounds do we get? Yes. Yes.

Well, maybe, so that's my prediction for 2025. Maybe that connection shouldn't involve me trying to get from one end of, uh, Charlotte, North Carolina airport to the other in 10 minutes though. Right?

That's true. I read a, I read a blog by a, a flight attendant, um, and basically she said, if it's less than several seven hours drive, Well this, there's something to that with all the prep time and everything. I think connection times have gotten shorter too.

What we think are acceptable. Well, no, what, so here's a little pet peeve of mine that I'd like to see change in 2025. If you speak to the airlines, they're actually gonna tell you they're on time more than they ever were.

Because what they've done is they've built, built-in buffers into their schedules. So like for instance, you know, the, here in Florida we have sort of an underground railroad up to New York, Mike, Mike notes that from an airline point of view, every airline here runs like buses from Miami, Lauderdale, PBI into Kennedy, LaGuardia or Newark. And it's a two and a half hour, two hour, 20 minute flight.

But now the airlines are scheduling, they used to schedule 'em for three hours. Now they're three and a half hour in flight because they're building in time off the gates and stuff like that. But what they're really doing is building in extra time so that they could show their early more than they're late.

And so, even though you got a 55 minute layover right, connection point, you probably have closer to an an hour and a half because there's that extra buffer time built in to these flights, assuming everything goes well. And, you know, you could fool some of the people some of the time, but quite frankly, Delta and America reap what they sell because they're, they're not, they don't play nice in, in all honesty, right? And it's not just the two of them.

JetBlue does it. Southwest does it. They all do do it.

They, they've buffed up the, the flight times to make it seem like, you know, they're, they're on time or even early, you know, and, and that pilot when he lands and he thanks you for flying their, for the airline today. And we remember this the next time you're late, that we got you in early today. Oh, you got me in on time.

It's just your schedule. Yeah. Look like a hero.

So, yeah. Right. And we worry about bias in our large language models, right?

We've had bias in data for a very long time. Yeah. I, you know, again, I'll go all Steve Bannon on him.

This is a, this is, this is like the old sales rep who, uh, you know, when his spouse was complaining about the fact that he was traveling too much though, and then he started telling his spouse that, you know, the show that he was going to was four days long, when it was actually only three days long. And then he comes home, Hey honey, I missed you, right? Yep.

So if I can, Yep. The surgery was a success, but the patient died. Yes.

If I can turn the, turn the lens and, and focus an area one, one of the areas that I'm really most interested in because of the analyst coverage that we do, is around code generation with ai. And I think 2025 is the year of the code generation specific LLMs including code quality, secure code, things like that. We've, we already have some of those available today.

I mean chatt, ptt, excuse me, chatt P four, um, you know, things like code at five and five plus Code Lama, et cetera. Open AI has their own code X for code generation, but also the other providers like the Microsofts, the GitHubs, the AWSs of the world, uh, having not just codes generation specialized LLM capabilities but specialized to their environments. I mean, who knows better than than AWS what their environment is like and how best to build applications within it.

So I think we're gonna see a lot of bifurcation, not just, I use Microsoft or I use uh, GitHub's copilot and that's my, that's my pilot. No, that's your window into whatever LLMs or services that you're gonna use. So I think that's gonna be the big focus around at least the code generation part of ai.

I think they, the chip manufacturers are gonna get more involved in software too. I think like Nave is gonna get much more involved in AI building AI systems. It just chips.

Here's My prediction from left field since I started with right field earlier. But on the left field, I think you're gonna see org charts in IT organizations and enterprises be ripped up this year. And I think a lot of that's gonna come back to this platform engineering conversation we started in 2024, where a lot of those people are gonna decide that, hey, you know, all these little specialist teams that we have created over the years are counterproductive and they're gonna start pulling all that into more centralized functions and require IT people who can do multiple things rather than, you know, standing around saying, I'm the infrastructure specialist for this, or I'm the cloud architect for that.

I think we're gonna see a much more, uh, homogenous team of IT folks who can manage multiple things, working hopefully more hand and glove to do things at scale. Ocon trip. I hope so.

Oh, contra. Alright. Point Point.

Johnny Carson talking. Ed McMahon, the amazing Carac, Oman Shrimp, Mr. Sure.

So I think you got that platform engineering thing all wrong. Do you, I think platform engineering is about building silos, not knocking them down. I think platform engineering says I want my little guys who do that little do that thing and I wanna keep 'em as a separate team doing their due dates.

I want them to communicate, right? I think platform engineering at its heart is about cross silo communication, not busting down silos. And so those teams are not gonna go away.

They hopefully will just communicate with, with the collective and be assimilated. I believe they'll be assimilated, that's for sure. Yes.

Well, the thing that would drive, drive collapsing it to your point Mike, is cost, right? If you can demonstrate true cost savings that don't have a major impact on delivery speed or things like that, maybe it's improving it, that would certainly push you to consolidate. I think The issue too is like even today, if you look at the cost of it, the single biggest factor is still labor.

And I think people are gonna start to look at that harder and say, how do we reduce that particular cost? Yeah, Well maybe we'll have Code Bakeoffs and we'll have, have, uh, platform bakeoffs, we'll take my platform design or platform design from one LLM and com use another one to compare it it against two different L lms. So things like that.

I think we think of the ability we have to compare allies information. I finished some papers in 2024 around using AI to really di dive into the depths of what an architecture and what functionality and applications performs before you modernize it. That's half the battle is just knowing what things do and who the people who wrote it on around anymore.

And if you're gonna, you know, move it to the cloud or start updating parts of it or maybe rebuild parts of it, you know, what is it you gonna, what are you gonna leave on the cutting room floor? You don't realize. And AI is very good for that.

There's a lot of really good analysis tools for that now. So I I I gotta jump in on this whole LLM issue though. So here's my prediction for 2025.

Something's gotta give with LLMs and there's two, there's two issues of it and two points to it in my mind. Number one is, are we serious about organizations near and far being able to create their own LLMs and maintain them? I don't know.

Not unless we somehow make it a hell of a lot easier to integrate and stuff like that. And so the logical flow of that is okay if they can't create and maintain their own LLMs, except for the big guys, they're gonna have to use prepackaged lls. Who owns the market for prepackaged LLMs?

Well, at the very general level, I think we've already see this market being defined. You've got open AI chat, GPT, you've got Claude Ro, you've got Gemini, probably from Google, you know, apple surprisingly is borrowing or partnering with all of the above. But let's face it, eventually they're gonna have their own As, and you know, Microsoft's open AI thing.

I will, who knows where that goes? Musk invested how many billion in Right? Musk has X AI or whatever X but Mitch, here's a, you know, something Brad Feld always told us Top three, If you're not in the top three, get the hell out.

So my, my projection is by the end of 2025, you're gonna have three major, let's call 'em general purpose LLMs, mega hyperscale LLMs, and they will be owned by hyperscale tech companies. Then I think you're gonna see a market underneath that for the specialists, for the specialty LLMs, for law, for medicine, for finance, for, you know, I'm, I'm, I'm hitting the usual suspects for deck, right? You'll, there will be a market for specialized LLMs that are kind of off the shelf 80, 85% and can be modified 10, 15% for your individual company.

And that will be, instead of developing your own LLM, that'll be the LLM you use when not using one of the hyperscaler ones. So I think that's how the LLM market shakes out for 2025. And you heard it here first, I think we're, I think you're, you're right, and we're living the first half of your prediction already.

We're already see who the major players are and then there's some things coming up along the side like, um, deep seek, which is funded by the Chinese, you know, we'll see if that becomes a, you know, a, a hands off kind of thing. But there are things that are gonna come up like that that will be new advancements there, there's a lot of innovation around how to create LLMs much faster, lower cost, and more efficiently. So there's a lot of, we're in that, how do we make it better, faster, kind of phase two to your point about having specialized I agree with that.

I think the other thing is you'll see more things like their neural magic acquisition by Red Hat, which is about deployment of LLMs deployment of AI software or, and access to multiple LLMs. If you look like even at with at Elastic, you know, they have an environment where you point where you want it, you know, in your, uh, vector, vector database if you will, front end to your, your rag, where, what model you want to use to be able to do that part of it. So a lot of this is gonna be choice, right?

And it's gonna be built into the software that we use. You only have to download the LLM necessarily to set it up and use it. You'll use it as part of the service that's online.

I think it's gonna be more about, well, I guess what people are calling, uh, small language models. And I think some of those will be based on foundational LLMs that the big guys are providing, but over time, those LLMs or the small LLMs will become foundations for other LLMs and they'll start to multiply. And each of those things will become easier for an organization to create.

I don't have to create a massive LLM for every task. I can use a smaller language model and then I can, uh, theoretically at least create my own AI agent off of that SLM and I might go from there. So I think that there'll be a move towards, uh, small, beautiful in the space of ai.

You can see the big guys, I think the ai AI agents, Yeah, I think AI agents are gonna be a problem as we move forward. I agree with, I always talk about it, but we're all, we're gonna, there's gonna be so many of them and to not, if you're not able to kind of version 'em and, and track 'em, we're gonna get serious drift in that, which is gonna be a issue and probably bring down airlines. Come on, you're gonna need agent hygiene, right?

You'll need Ag Agent hygiene, otherwise you're gonna have rogue agents that are floating out there. Like you have subscriptions for SaaS products now, right? And I mean, and this is not new in tech.

I mean, think about cloud. How many, how many unused sessions, instances are there on an AWS that were open and never closed, right? And they're getting charged every month for it, but it's pennies and so they don't recognize it.

Software on your computer, how much stuff Do we Have installed on our laptops? It isn't use. Hey, Google my refrigerator, I spent earlier this morning clearing out all the, if it, if we had it during Christmas and it's January 6th, it's time.

I'm not gonna eat it. F**k that out. I think we'll see a lot of cleanup on aisle six agents, you know, go out there Like clean.

Well that I think is what you're gonna have. You're gonna have Agent Smith, right? Who cleans up Rogue Agents?

That one's gonna be called oh oh seven, right? Yeah, I know it'll be a hero, but, you know, red pill, blue pill, Steve Bannon don't care. Or the above.

Anyway, hey guys, we're, we're 45 minutes in. Um, I'll, does anyone have any other prediction they wanna throw on the pile today or no? We To talk about regulation.

This is the year of seeing some enforcement in there is like software security starting of it. So we should spend some time talking about that. Another Time.

I think regulation about, about, about Steve Manon. I'll share some wisdom from a former publisher of mine who said, you know, the one thing in the world you should always remember is it's a lot easier to throw grenades than it is to catch a Absolutely, at least as it applies to the us. This is not a year for regulation.

No, not a year for regular, sorry, it's la as I said before, it's laissez-faire, whatever goes batten down the hatches. Keep your head slow, guys. I'm sorry.

Um, if we don't have anything else, let's call it a wrap on our first show of Textron Gang. Well, just a quick reminder again, predict is moving to Tuesday, January 14th. Textron Gang will be on five days a week.

We have, we're gonna be recruiting some new gang members for this year too. com and we have our producers contact you. But, um, I'm looking forward to it and we have a lot of, a lot going on all around.

So, Mitchell, Mike, Tracy, thanks for joining us today. Thank you for watching. As usual, we've got Textron tv all day following our gang this morning.

Uh, until tomorrow though. This is Alan Shimo for Textron. Have a great day everyone.

Happy New Year.