Techstrong Gang – January 23, 2025

Hey, everybody. I'm Mike Vior. We got a whole lot of new AI rules, not to mention $500 billion in funding.

And we're gonna talk a little bit about a new report from the Futurum Group about AI adoption. And finally, what's the latest in AI coding tools? You're watching Text on here.

Hey, everybody. We're back and we got a full house today. We're gonna start from the, uh, left hand side of the country and work our way back as usual.

John Schwartz is in Silicon Valley, and of course, tracking all things coming outta DC these days. John, good to see you. Good to see you.

We're gonna have our hands full the next couple years, I assume, and kind of fear, but yeah, it's, it's, uh, we're gonna start. We started off the day one of the administration with a lot of stuff to talk about today. So glad to be with you.

All right. I'm not entirely clear, which is further West Ohio or Alabama, but since Ohio State won the championship, I'm gonna go in our friends in Ohio. First, Stephen Fs.

Good, good to see you again, as always. It is great to be here. I, uh, unfortunately, IW missed a Textron gang, uh, earlier this week, so I'm glad to join the gang today.

See that? You, you just can't live without us. So you, you missed a day and you had to show up for the next one.

Not, not, well wait an entire week. You had to be right there, so we're impressed. Of course, also in Ohio is Lan Neha, who works on Tech Field Day with Steven.

They are longtime colleagues, and Langa is also running tech, drawing AI for us. So Langa, good to see you. Good to be back.

Thank you. And finally, John Willis, who I'm assuming is in Alabama and where it allegedly is snowing. Is this true, John?

Yes. Is true. And, uh, and just, uh, I feel like, uh, I've got my picture on the cover of Rolling Stone.

I've been Tech Field Day, so is that next week? It is. And, uh, How often you there?

Yeah. Yeah. Actually, we got, uh, we got three of us, uh, on this call.

That'll be there. Uh, maybe we should do the gang from there. Oh, that'd be fun.

Yeah. And I think we have to maybe on Wednesday and it'd be Thursday. Um, I was, uh, I, I'm from San Jose, so welcome back.

I think, Steven, you've been there. I'm not sure about you John, but couple times to San Jose. I there a couple of times, Jeff, I think.

I think, I think that's an awesome idea. I'm gonna pencil that in right now. Um, I would love to get into what's going on though in the world, because, you know, there's this new administration and they seem to be, you know, kicking out these executive orders at about a rate of about one every two minutes.

The one that we're keeping an eye on at the moment, though, is what are they saying about ai? They seem to have, first thing was scrapping the Biden rules. And now the second thing is lining up all these folks for this massive investment in ai.

So, and of course, one of the first things they said was, Hey, and this investment will not be subject to as many reviews because John, who needs stinking regulations? Yeah. Who does?

Hey, um, yes. So, uh, first I'm gonna start with a visual presentation. I, I don't usually do this on the show, but I think it's important because in a sense, yes, you know, they say a picture says a thousand words.

This one might tell us about a couple hundred billion. So here we have on this side behind Trump lined up several executives. He might remember them or recognize them.

There's Jeff Bezos, Bezos, there's Mark Zuckerberg, Tim Cook, Elon Musk, Sundar Peka. Evidently, uh, Sergey Bryn was around. I don't see him in the photo, but he was around, and I know that Sam Aldman was around in Washington dc And, uh, basically after the inauguration, uh, Trump signed an executive order, as you said, Mike, to scrap a 2023 executive order by President Biden, our former President Biden, that, uh, that would, uh, kind of held hold companies and developers of AI to some sort of safe safety standards, or look at his potential harm to national security consumers and workers.

So that's out of the way. On Tuesday, later in the day at the White House, Trump hosted Larry Ellison, the soft bank, CEO, Mr. Son and Sam Altman, as part of a announcement of up to $500 billion in private sector AI infrastructure investment in the us.

There's also a, uh, joint venture between open ai, SoftBank, and Oracle called Stargate, which is a little ominous sounding, but nonetheless, so in a sense, yes, unfettered AI development as fast as you possibly can. Uh, these companies, I guess, will be entrusted to self-police themselves, just the way social media. So, uh, self-police itself, going back to Zuckerberg, that didn't work out very well.

But then again, Congress can't pass any legislation invol regulating any type of technology. So now we have this new kind of golden age of America as it's been termed, or even the intelligence age, or the intelligence race. And, um, it's gonna be interesting to see what happens.

I, I anticipate, in addition to what we just talked about, even more actions spilling out, um, it's, it's, it really is a land rush and a land grab. And it will be interesting to see where this all goes. There's also even a report this morning I saw from a Wall Street analyst about this fate of TikTok, which is, um, was extended 75 days on the ban, or from the ban in which the speculation is that Musk and he, perhaps even Ellison, might end up owning the US operations at TikTok.

So never dull moment. Well, that's, that's a whole separate story. But John Willis, how are you feeling about all this?

What's then, well, first off, runs on, right? So that's, you know, there's, there's a little interesting tidbit there, right? TikTok, the US implementation of TikTok runs on Oracle, Oracle's infrastructure, Oracle sort of class.

So anyway, that, um, yeah, I mean, there are pluses and minuses, right? Like, um, you know, the, I, you know, I did some analysis yesterday on that, 'cause I wanted to write my own blog article about what do I think about this? And, and, you know, the, the, the pluses were that, um, you know, regulatory burden, you know, the trade-offs, right?

The, uh, federal outreach, right? Like, you know, that there are things that were in the executive order that there were not everybody was happy with, right? Like they, you know, there was some definite, um, you know, um, and then, you know, and, but then there are things that like, now we have no roadmap.

To your point, John, like, I, I don't think allowing open ai Sam Alman to worry about me. Like I'm, I'm a little worried about that, you know? Um, you know, or can have any concern for what he's doing when it comes to me, right?

Like, so, and then on the Stargate thing, right? Like, I think I'm more interested in like, how is that gonna play out? Because I smell lawsuits all over the place by the ones that weren't invited, you know?

Right. Look what happened with the cloud and Federated Cloud, right? Look how long it took for to, to get that whole thing cleared up, right?

So, um, you know, I didn't hear, hear any mention of Google in there philanthropic or, and so I, I just wonder like, is this just gonna get stall? It's, it's gonna be a lot of noise because it's gonna be stalled for the next year, because there are gonna be tons of lawsuits. Well, I want to, I wanna bring in, um, the, uh, news about the, uh, framework for artificial intelligence diffusion, which was revealed last week, uh, under the Biden administration.

Uh, one of their last, um, efforts to tighten up the restrictions on export of AI chips. Um, the framework is actually pretty impressive. Uh, if you read through it or if you read the analysis, uh, on semi analysis by Dylan Patel, um, it's really interesting what they've done to really kind of fill in the holes in the Swiss cheese of restricting exports.

And on the one hand, and on the other hand, it's really interesting what it does sort of from a forward looking stance in terms of requiring major AI model development to be in tier one countries, and specifically US companies have to develop AI in the United States. In fact, 50%, according to the framework of their, uh, AI chips have to be, develop, have to be situated within the United States. Uh, if you look at this Stargate announcement, uh, and you read between the lines through the filter of the framework on AI diffusion, one of the challenges from that framework was that Oracle invested heavily in offshore AI infrastructure.

And that would be drastically reduced and restricted by the framework for AI diffusion. Stargate actually fixes that problem for Oracle. And another aspect that's interesting here is that, from what I'm hearing, Oracle has already built out too much, uh, AI infrastructure outside the United States, um, previous to this.

And they're gonna have to basically reshore a lot of that infrastructure. So much of this investment might actually not be investment so much as reshoring of hardware that Oracle has already purchased and reshoring of capabilities that Oracle already has, potentially now, if, um, the framework stands, and we have no reason to believe it won't. Um, for what it's worth, the, uh, Trump, uh, rescinding biden's executive order wasn't a specif specifically, uh, directed at this particular ai, uh, initiative.

In fact, um, it was part of the overall, um, executive order called initial rescissions of harmful executive orders and actions that revoked well, dozens of executive orders all at once. And this one was just sort of part of what was swept up in there. I'm wondering if anyone in the Trump administration even knows what they've done in terms of rescinding this executive action.

And furthermore, I'm wondering if anyone knows what they, that, that the, uh, actions of the Biden administration are part of the reason for this big, uh, Stargate announcement. But who knows? I mean, who, who cares?

Frankly, it's better if they don't know, because then they won't kneejerk to kill it, because the, uh, framework for AI diffusion makes a lot of sense. Yeah, that's an interesting, you know, perceptive. It was a very perceptive point by you, John, earlier about, uh, the, the executive order and how it was written, because I did talk to a couple of people who had a hand in crafting it, and they acknowledged there were flaws in it.

Their only concern was that now there's no real oversight or a roadmap, as you pointed out. The other thing, um, Steven, that, and I, I was thinking about this, and I'm gonna ask Mike as well, you know, I always have a bit of a skeptical eye around these major investments or these plans that Trump announces, like this goes back throughout his career, Especially with SoftBank. Yes, yes, exactly.

That's where I was going. And, and, and I'm thinking, I'm not thinking Foxconn, I'm thinking SoftBank. I'm even thinking back to Trump's earlier, you know, massive projects that he was gonna build in New York City that he talked about for years, and they just never came to fruition, even with the, the latest announcement in the Middle Eastern investment, which where there were projects that that didn't, didn't develop into anything.

So I have a, a bit of a skeptical eye. Maybe this is a bit of a repurposing around Oracle and, and SoftBank and open ai, but I, I just, I'll believe it when I start seeing, uh, real action or, or, or tangible evidence. Mm-hmm.

And there's a lot of speculation in the real estate side. We talked about this on a show earlier, but, um, a lot of the folks involved in this seem to be more interested in buying the land and hoping to sell it to somebody and maybe, you know, mark that up, then they aren't necessarily in solving the AI compute challenge. But Lanman, what's your take on all this?

Yeah, Trump himself has said that the, this target venture is going to help build more data centers and create more jobs and, uh, in the United States, but it really remains to be seen how it all pans out. Um, there are a lot of ifs and buts, really. Uh, so I think it just, it's something that we'll have to wait out and see.

Well, and isn't it kind of inconsistent? Like I read that like a hundred thousand jobs. Well wait a minute, because the Facebook has a model now.

You can build data centers in months. They have models where they can be operated with less than 15 or 20 people, right? And AI is supposed to reduce head count.

So like, why, why is this AI initiative now? They're magically gonna, you know, 10 x jobs, um, for what is traditionally being done now in, in data center. So, and well, there's, there's, there's gonna be two guys managing the robot, building the data center.

Yeah. Not there yet, but like, I, you know, oh, I, yeah, for now, at least until we have a new super intelligent AI agent that runs everything, it, the one last thing I I'll say is if, if anybody's read the Fifth Risk by Michael Lewis, that was, that was baby steps compared to what we're going into right now in every sort of level. The fact that AI is so pervasive, the technology of ai, we know of what, you know, how pervasive and how impactful this is.

We've all been doing this for years. This, there's nothing been this impactful and to just completely fifth risk. It just seems awfully scary to me.

Hey, Steven, have you heard the one about the future of the data center? It's, it's, it's gonna be managed by one guy and a dog, and the dog is there to keep him from touching anything. That's funny.

Alright, what do John, so what do you expect to see John Schwartz in the future here? It seems like this is just the beginning of, of series of things, but what's your speculation? Oh, I was gonna, I think we already are starting to see like this, like a chain event or domino principle.

I mean, I don't know how it's all gonna work out, but we, we we're seeing a, for instance, Google just invested a, a billion dollars in anthropic in which Amazon's already invested $8 billion. We're gonna see all sorts of vaccinations, people moving in crazy speed, you know, the interest. One interesting thing is, um, you know, it's, it's, it is open, open game, open track.

What interesting thing was there was no presence of Nvidia that I know of at the inauguration, or that we didn't hear anything from them. Um, or for that matter, even Microsoft, which I found was kind of interesting. I'm not sure if anyone has any thoughts on that, but, uh, perhaps those are two companies that really don't really need the help of the administration or have to cozy up to it.

But I found that kind of interesting, the absence of those two companies. Yeah, I found that interesting too. And I, I heard the same thing, John, that, um, you know, Jensen Lang and, and, and, and Nvidia were nowhere to be seen, uh, with the inauguration funding at the inauguration, et cetera.

I believe Microsoft did contribute some money, but, uh, again, I didn't hear about, uh, you know, the presence of the, uh, you know, Microsoft execs or anything like that. Um, you know, it, it, it does seem, you know, you could, I guess say that vm you know, Nvidia doesn't really need, um, you know, the government's help, but frankly, Nvidia is another one of the companies that's gonna be really, um, impacted by the framework for AI diffusion. If it con continues through, if I was, and they push back on, they push back on that immediately.

Oh, Yeah. They push back strongly. And one reason that we heard about it was because Nvidia was pushing back so, and, and got so much publicity.

Um, I'm surprised that they're not, um, well in their, um, uh, let's say, uh, to be nice, uh, pressing the flesh, uh, in order to try to change, uh, the, uh, administration's goals. Maybe they don't trust anybody. John Schwartz, though, I would love to get your opinion on this.

Where does all this money come from? Because, you know, in the state of New York right now, they're trying to scrape quarters off the floor. Yeah, that's a good question.

In somewhere. So, I mean, going back, going back to, uh, I don't need to make this personal, but going back to Trump and just based on, on his projects, and I've read a, a number of Lucky Losers, a very interesting book about him. And it, it just kinda looks at his career as kind of a microcosm of where the political side goes.

And they're, they would just create numbers on projects. And when I see 500 billion, it's a nice round number. I, you know, I will believe it when that money actually comes to fruition.

I, I just think it's, it's a pumped up almost imaginary number they just threw out there just to get our attention. It has to be, he had no inference with OMB. Like, how does he even start Yes.

Throwing numbers out on day one when there's been no discuss discussion with OMB. I mean, it's totally from the, from the get go, it's, it, yeah, it's the total, you're right, John. It's the total mo.

It's, it is just, this is the method of operation. You throw out an outlandish number. This hap this has happened for decades with him.

And, and it goes back to this first administration. And again, I mean, this is all, a lot of, a lot of bluster, a lot of, a lot of, uh, you, these, these plans are written in, in two paragraphs. There's no due diligence as far as I know.

It's just pie in the sky for now. You know, I'll give them credit if something does come to fruition, but for now, I'm very skeptical about it. That was my point about the lawsuit.

It hasn't even gone through the office management budget yet, right? Like, it hasn't gone through any economics reviews. It hasn't.

And then we're not even talking about the lawsuit. So it's all just fantasy. Now, I, I will say, if they create a hundred thousand jobs in Texas and they build data centers and they insource infrastructure, um, yeah, great job buddy.

But, but like, but just, you know, throwing your finger in the air and saying it will happen. Um, you know, And This is like, this goes, yeah, this goes back to all the manufacturing jobs that were promised in the first administration. And how did that work out?

This is the point where I, I'm, I'm expecting Alan to show up, you know, out nowhere and talk about Foxcon and talk about the con of foxcon. And I, and I, I, I suspect that we're seeing history repeating itself again to, to a bigger extent in a more pumped up extent because it's ai and AI has to be bigger and more hype than anything else. Alright, Steven, I'm gonna give you a last word on this, but I kind of feel like we're just watching an episode of AI Game of Thrones here.

Yeah. I'm, I, with all of this stuff coming outta the administration, I'm really trying to figure out what the reality is. Like, like you all are saying, $500 billion, it's absolute fantasy.

Uh, remember SoftBank, MACI Shisang was right there with Trump in the first administration announcing a hundred billion dollars investment in the United States. That never really happened. That a hundred billion dollars was basically the size of the vision fund, which famously became a massive loser and a lost at least a third of that money.

Uh, a lot of those startups closed down. They didn't amount to anything. And it wasn't like that money actually ever came in 500 billion.

Yeah, it's a nice big round number. It's five times the size of the vision fund. Um, a hundred thousand jobs.

Sure. Nobody's even saying what this thing is gonna be. What we have to do, just like everything that comes out of the Trump administration is we have to read between the lines and figure out what this is gonna do and how it's gonna nudge the industry.

Um, you know, just like, you know, the, the, the repeal of the executive order on AI safety, just like the AI diffusion framework, these things can have real impacts. They can cause the industry to turn, they can change the trajectory. We just have to try to not focus so much on the announced, um, bluster and focus really on the reality of what's gonna happen when these things hit, hit the ground.

When, when, when, with, with the news of these a, you know, AI data center build outs and this investment and so on. No, it's not gonna be $500 billion. It's not gonna be a hundred thousand jobs, but there is gonna be some investment.

There is gonna be some reshoring, especially if the AI diffusion framework stays in place. And that actually will have an impact. And we just all have to try to figure out, kind of push away the noise and try to figure out where's the signal in this.

And, and I'm gonna add one last thing, Mike, which is, I talked about this before the call. I don't think what's gone over the radar is what they've already done with CSA is now you have a perfect storm. You are literally pulling out all the, the, you're basically saying self-regulation and you're basically dismantling csa, right?

Um, it, you know, I mean, I I follow critical infrastructure. It's not like my space, like Josh Corman is got me scared to death if anybody knows Josh Corman about the cyber critical infrastructure, right? Water supply, telecom, and like this is doing an amazing job there.

And that stuff's all getting gutted. Alright? We will be coming back to that topic in a future show, no doubt.

But, um, in the meantime, what I learned today so far is if you want the entire AI oligarchy to bend the knee, you just have to point to some imaginary cache and they show up. So there we have, all right, we'll be back in a minute. Hey, everybody.

We're back and we're talking about some research from the Tuum group, the Parent of Techstrong, and it points to, well, what is the level of adoption among larger enterprises of ai? And it seems to suggest that this whole notion of a fear of missing out is maybe going away. People are taking their time a little bit and maybe being a little more thoughtful about how they're planning to use AI so long.

And this is on text drawing ar the site you run. Now, give us the high points. Um, so yeah, it appears that the speed of AI innovation and success is a much slowing down, at least in the early stages for some companies.

Um, this is owing to a number of factors such as, uh, disconnect from the realities of implementation. Uh, adopters expectations are not well managed. Uh, poor ROI focus expectations for quick gains integration issues, and most importantly, lack of hands on strategy from CEOs.

Uh, the future on groups, uh, state of market survey, which, uh, recently released, uh, shines slide on this. Uh, some of the stats say that, um, 62% of CEOs view AI as a powerful force of, uh, change. But 38% just see it only as a hype.

And, uh, only 46% of the companies surveyed, uh, said that they were very prepared for AI adoption. And 31% were only moderately, uh, prepared. And, uh, only 48% prioritize rigorous ROI measurement.

So, and John also did an article on this. So I feel like that, uh, there are some takeaways from this. Uh, it feels like, uh, to re all the benefits of AI companies need to have some things in place, such as, uh, for, to start with, uh, they should be preparedness, which is crucial.

And, uh, sort of internal training, the ability to accommodate and manage the growing purpose of data, making sure that people are trained to handle the technology at a professional level, um, putting in place best practices and guard rails and slowly and steadily incorporate the change, sort of shoving it down imp implieds, What, what, what you had me at CEOs and disconnect from reality. So I've been laughing ever since. But Yeah, that was another, that was another interesting thing to read between or read in this study, um, was that incredible disconnect.

Um, 78% of CEOs strongly believe in their own ability, their personal ability to guide AI for their companies, but only 28% of mid-level managements agreed with that. Oh s**t. I mean, the people actually in the, in the trenches, right?

Yeah. What, there's, the one thing I, I was there was, I think there was a takeaway, uh, I'll get outta the way because I talked too much in the last segment. There's a takeaway that, um, from the CEO sentiment study that the analysis reveals, as you said, Stephen, a striking disconnect.

The most successful companies are those where top leadership deliberately steps back from hands-on AI strategy. That kind of stuck with me. I know, right?

It's incredible. I mean, if, if, and, and the disconnect with customers as well. You know, the, this, this whole idea that, um, CEOs are all jumping in there, but according to this study, only 24% say that customers are specifically asking for AI based solutions.

That's wild, because that doesn't say 28 4% of customers, it says 24% of CEOs say their customers, which means that that could be much less in terms of customer demand, because that's only the ones that are going up to the CEOI. It, it, it is incredible. Um, but it kind of matches.

I mean, do you know of anybody who's like out there, like going to their, I don't know, their main, you know, software as a service supplier or whatever, and saying, you know, show me your AI strategy. I need more ai. You need to put some more AI in here.

No, customers don't want ai, customers want solutions. AI is a way to give to, to make solutions. It's not the solution, Right?

It's investors who are screaming about, show me your AI strategy, right? At the end of the day, John Willis, I know we have talked about this subject in the past with you, and it it rankles you because you see so much the eye progress, I think wall. Yeah.

I hate the Wall Street view on this, right? Because, um, what I'm seeing in the companies that I'm visiting, well, first off, if the CCEO says it's so, it will be so all right. Like, that's just the reality of the, like, so saying they're disconnected and you know, like, um, but there, there are incredible projects going on.

A lot of 'em, I can't talk about that. And, and what I do know is there's a lot of reluctancy to actually promote what they're actually doing because of the competitive advantage. I mean, they're created, they're creating so much innovation internally now how it all winds up.

And I, you know, we'll go back to the, what's the famous quote that if, you know, Ford, if I asked the customers what I want, we'd still be driving, you know, horse and carriage, right? Like, um, you know, I, I don't, I don't think that's a, a really, to me, that that doesn't stick right? The organization, like I know companies that have incredible amount of data.

One company has credible amount, you probably figured it out, but they have incredible amount of data on every restaurant on the planet. Like they're figuring out how to do something with that data, right? If they know every transaction of like 70% of all the humans who eat out in America or maybe even in Europe, right?

Um, like there's gold there. So it's the data stupid. Um, it is that organizations are not actively telling you what they're doing.

I've, I, on a previous Textron, we talked about, I had one client, Lily made me sign a second, NDA, not only that was I under NDA with the client I was working for, but I literally had to create, do another NDA to not discuss what they were doing specifically. Um, you know, they're very, you know, like, rightfully so, like if you're in a competitive space and you believe you have, um, an innovative idea. So I don't, I don't know what the right answer is.

You know, wall Street are usually better at this stuff than I am, but, but I do know, you know, my ground level knowledge of what's going on with some of the large corporations that I've been talking to, I, you know, one I can talk about publicly is what John Deere's doing. John Deere is across the board, um, figuring out how to use this stuff strategically. You know, like a transformational, you know, and that's the thing, wall Street's not looking for.

They're looking for like, where's the money? Where's the budget? What they, there are organizations that are completely transforming, there're talent and innovation strategies based on this.

You know, that's interesting that you said that, John, because I was in, uh, Pittsburgh last week at Carnegie Mellon, and I went through the robotic robotics lab, and I talked to some folks in agriculture, especially ai, agriculture and John Deere's, that name kept cropping up and they, so to speak, no pun intended, but that, that was one of the, the key case studies actually something we should look into. I'm just, but the fact that you brought that up is, is really interesting to me. They are gonna transform and they're somewhat a little bit more transparent than others.

That's always the, the key, as you said, you, a lot of these companies are reluctant to say what they're doing. So where's the hype and where's the reality is somewhere in between. Steven, I'd love to get your opinion on this though.

When I talk to people, they're in a quandary. They cannot fund every AI project, and about half the AI projects are rapidly becoming table stakes, right? They're not competitive advantages.

They're things I need to do to stay competitive. And then they're trying to figure out, well, what are the things that are gonna really make a difference for us to be competitive? And it turns out everybody's kind of working on very similar things.

And so what is that window of opportunity around ai? Well, that's an interesting, uh, aspect of this study as well, is that it does show a real differentiator between sort of these, uh, younger digital native startup companies and the more established companies, to the extent that at least I, I'm still working through my way through this. I'm still trying to really digest it because this is an incredible data source.

Lemme tell you that I, I've never seen it this much information direct from these CEOs. Um, it seems like established companies are smartly focusing on how they can use AI to transform their existing businesses, how they can use it to, um, yeah, as, as we were just hearing, to ingest existing data to come up with novel uses and novel applications for the existing market they have. Whereas, um, the more, uh, digital native, smaller, you know, scrappier companies, they're looking for transformative business models and they're looking for ways that they can use data in a way that nobody has thought of yet.

And that I think is really what we're gonna be looking out for here. We've already seen a divergence. I mean, you know, John, uh, Willis and I, you know, we were at the early stages of cloud.

There was that incredible divergence between sort of the cloud native companies and the cloud followers. We're seeing that 10 x with ai, the, the AI native companies, the data native companies, they really understand the capabilities of this, of this technology. They're looking for novel ways to use it.

They're looking ways that, that can serve as accelerators and differentiators. What has to happen for the rest of the market is what happened with cloud technology, which is where they look for the most useful tools, the juiciest, marcels, the really, the things that can really transform what they're already doing and adopt those things, embrace those things. That's what we're seeing with cloud technology.

I think that's what we're gonna see with AI technology in the coming years. John, what do you think, John? Oh, I mean, I was just gonna pipe in 'cause I do talk a lot, but I, you know, I talked about this Mike on a previous call, you know, the Friday y if you will.

Um, but the, um, you know, I talked to CIO, um, uh, near the end the last year, and, you know, he said something really insightful about, like, John, I'm not in this for the ROI, I mean it, what I have to do in my organization is create a sort of seamless, horizontal, um, you know, sort of talent and innovation transformation, right? And, and like, like you listen to that and say, oh, it's just CIO speak. But he knew like this whole question of like, what's gonna happen?

This comes up a lot, right? If, if we don't need junior developers, how do we create apprentices, right? This comes up all the time with ai.

Well, the, the, the savvy business, or CIO or the, the executive team who understands your business, they're realizing that they're going to have to create a talent like from the 30 year Java developer to the two months out of Carnegie, me and AI expert student, they're gonna have to create a seamless innovation strategy. And again, I, if I knew that I'd be making, you know, $7 million a year for a corporation, but like, you know, but that, that I think the organizations that are thinking alike that way, you know, again, I wanna be careful John Deere, I, I'm talking about the stuff that's public out there, but, but like, like those are the type of things I'm looking for when I'm interviewing, when I'm hanging out with Jean Kim's crowd. And, you know, I, I want to hear how you think.

Are you just like letting people go off and create pilots and co-pilots and chat bots and like, because you can, or are you setting sort of an across the board, or at least attempting and, and back to your Steven, like that was the, the haves and have nots and cloud, right? The haves understood this was a technology transformation and they, they, they step back and some of them, you know, capital One, capital One is a great example of a company embraced it strategically in the financial sector, probably one of the earliest ones to truly take advantage of cloud. And they did it, you know, head to toe.

And so that's what I'm looking for is the organization. CI talk like this. So, so let me ask you this question.

What do you think the odds are the next great thing in AI that a company does is gonna be driven from the top down? Or is it really gonna be driven from the bottom up by the people who are closer to where the AI is gonna make a difference? Um, bottom up is how ideally it should be, but oftentimes, uh, like you mentioned, there is, uh, this strange disconnect between, uh, the Csuite executives and the people that are working at the ground level doing the real work.

So, uh, as long as that, uh, there's a bridge between the two of them, I see it's very difficult for companies to really implement that kind of a systemic change where, uh, it is done in a discipline sort of way. Um, uh, but, uh, surely these smaller companies, and it's actually really following the cloud trajectory where companies are at first, like they're overwhelmed with, suddenly there's this AI rush and then with all of the aggressive strategies and, uh, stuff like that. And there's that myopia that, you know, if I can embed AI into my, uh, organization, I'm, it is my ticket to success.

So I think we are we'll slowly get out of that phase, uh, to a point where companies start to see more clearly and see how it can be actually implemented step by step in a way that works. Um, so I hope to see that kind of change happening in the future for Sure. Here's my bet, my bet is that the middle managers are gonna wind up being the heroes that implement this stuff working from the folks at the bottom up, and they're gonna become the next people taking the CEO survey.

'cause they're gonna replace all the guys that couldn't figure out how to make it work because they're too far removed from the actual thing that the company does. Steven, am I crazy? I think you're not crazy.

And I think that that's what we've seen in the past as well. Um, and, and, and one more thing from the study. Uh, it, uh, once again shows companies are desperate for people with this kind of knowledge and skill.

And I think that, uh, if those of you listening are, uh, interested in, uh, what to get involved in, uh, well, uh, I know, uh, this is not gonna be a news flash to you, but, uh, hey, maybe learn a little bit about Ai. Ai. There you go.

Hey, I think that goes for everybody. If you learn about ai, you're gonna write your own ticket because as they say, it's not AI that's gonna make you lose your job. It's the person who knows how to use ai, who's gonna take your job, right?

All right, we'll be back in a minute, folks. 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. Hey everybody, we're back and we're returning to one of our favorite subjects, which is just how smart are these AI coding tools?

And we talked about in a previous episode about a survey from folks that suggested that the amount of bad code being generated by these tools was high because they didn't understand enough context about the deployment environment. So developers were spending too much time debugging code written by the AI tool. Since then, we have seen some advances, uh, Devon's talking about a new version of their tool that's smarter and has more reasoning capabilities.

And then there's even a startup that's, uh, I'm gonna butcher this, uh, RA or something of that effect, which design their tool to keep the developer engaged. Their argument is, is that developers, if they're engaged, well understand the code and it won't be so hard to debug it. 'cause part of the problems with some of these other tools is they're just black boxes and they generate code and nobody knows how they did it.

John Willis, you cover this space more closely than any of us. What's your take on, where are we? What's going on?

Yeah, I mean, it's a mess just like everything else, right? Uh, I think, you know, I, my personal opinion to Devin, it was dead on arrival. We've gone through that.

You know, the, um, they don't even show up on the benchmarks. They don't even try to show up on the benchmarks anymore. Uh, the sweet benchmark, which is good or bad, it is the litmus test for a coding agent.

Um, you know, they, they, they, they hover in on their last peak was about 14 to 20%. The benchmark now is a 55% resolution to 60%. Um, and there's just a lot of players since, uh, you know, even the Open Devon team is now called, uh, all Hands.

They renamed themselves, right? They were basically crew. Literally about a week after, uh, Devon announced you, you know, their sort of fictitious resolution capabilities, um, which were just chopped up by the industry, and Open Devon was a response to, um, like, Hey, like, this is not that hard.

The, the AI engine does most of the work, right? Um, and so, you know, uh, I find that two simple open source tools, ADA and Klein literally do everything I need to do. I mean, you've heard all about Cursor.

There's just a lot of players. Guy Pur Purdy that sneak founder. It's got something called te.

I mean, it, it, you know, Amazon queue developer shows up really high on benchmarks, uh, you know, um, co-pilot workspace, which still I don't think has been released by Microsoft. So it's a very competitive space. But the key thing that you said there, which is when these things first started coming out, the context windows weren't as large as they are now.

And what, what a lot of people are realizing, and I think eight is the first one I've seen do this, is they allow you to take all the non functionals, like your, your Jira tickets, your um, your documentation. And, and this has been sort of the biggest problem with these tools is if you just ask it to generate code, um, it really doesn't have any strong context of all the other resources that are related to the code's generated. And so now these context windows are so big, like you can literally, I mean, some cases you just load a repo into these, uh, engines and just say, start explaining this to me.

So there's, instead of two worlds, there's org, there's tools that are staying far ahead of everybody. So like, pull in all what I'm gonna code now. I want to take all the documentation, I'm gonna take the terraforms, I'm gonna take, I'm gonna take everything I possibly can and throw it in the context window when I ask to create a service.

Patrick Debar talks about this elegantly in some of his latest presentations about, like, the world has to be this massive context of not just code, code is such a small part of the operational cost of an application. We all know that. Um, and, uh, so I, I think, you know, again, I, and, and the, the balance then is the technologies.

I mean like all of the major, you know, sort of LMS or models now have their own agentic capabilities. A lot of the early Devon Open Hands, all those things, uh, were that they were implementing Agentic capabilities on platforms that didn't have it right. Built in.

Um, and so the, you know, the, I think the, the message is, you know, learn, learn how to create agent based solutions. Learn how to let a AI understand the full context of your service and then, you know, don't get locked in right now to anyone solution because it's moving so fast. I mean, I, you know, I, I mean just what's happened in the last quarter, or not even the last four months between g PT four oh going all the way into oh three.

I mean, it's just mind boggling what's happened in the last four months. Other reasons. Hey, John?

Yeah, check. Can I ask you a quick question? I, I don't, this might be a non sequitur, but what you, I mean, you know this so well, what about operator from OpenAI?

This might be coming up, some version of it, or early version might be coming up soon. I've read to the point where it has ability to autonomously code or even Jew, Jew functions without any human interaction. Is that, is that kind of play into what you're mentioning?

Yeah, no, I, again, they, they've already sort of created the ent, you know, capabilities and now, yeah, I mean like, again, the point being, well that goes back to a bigger picture. And I always felt this, I've been surprised at, and I'm gonna sort of wiggle around this and get to the point to your question, which is, I thought when, um, when Microsoft co-pilot workspace, which I think is still in beta, um, they were gonna win it all. 'cause all, if the big providers provide a high percentage of resolution or ability to sort of create code autonomously, then I don't see any of the little players surviving, right?

Um, and so, yeah, I mean, you know, if, if open AI can take, you know, what they're doing and, and create, you know, sort of an operator level efficiency, I'm still waiting to see what Google's publicly gonna come out with in this space. They have it, you know, it's, it's internally I've talked to people, they use it internally. Um, so there's one element of that question, which is I think the big players are probably gonna dominate.

If you've gone all in on, on Google and Vertex and they have a reasonable solution that that benchmarks pretty well, then you're probably not gonna jump over to the operator or, or, um, philanthropics or, or certainly not like Tesla or, and then again, I think the, the, the interesting then becomes the open source edge players, and I, I'm probably the only one that calls it, but Eighter and Klein and these tools that are really effective is instead of big picture, $500 a month or a thousand, $2,000, all you can eat Devon Solutions. So John, lemme ask you a question here about all that context, because one of the things I'm starting to hear from folks is do I need an AI coding tool that is tightly integrated with the rest of my repositories in my DevOps platform to get that context? Or can the AI coding tools be, are they more disposable in that sense where I can just swap them in and out as I feel like No, I, I think in, in fact, I, you know, I I, the first time I heard it, it's Pius soer, so I'll, I'll, I'll replicate your, uh, like destroying the name.

Uh, but like, I think that's interesting 'cause that, you know, one of my arguments that I get a lot of pushback from, you know, the sort of all the sort of AI kids and is that the, the, the, the main repositories are GitHub, like all the engines that are coding system, all been trained by open source GitHub libraries, basically, right? Or gi you know, gi, right? And, and there's a big difference between, for the last 10 years, capital One with 18 to 20,000 Java developers and how you build services in an organization like Capital One on, you know, probably, you know, certainly hundreds of millions of lines of code, maybe billions of lines of code and code that's been generated in open source projects.

So they, you know, my argument, there's gotta be a difference. So organizations that are helping, and there's a few of these happening, helping go in and understand your libraries, your code, uh, I don't have any data on it. I think they're gonna be far more efficient.

You know, you think about what Capital One or a large bank or Goldman Sachs has to do for their code base to work it, it's just the, the, the decision and the training and all of the sort of code generation is just gonna be significantly different. Yeah. I want to jump in on that, John, Because I think that you're really onto something there.

What you've been saying a lot here makes a lot of sense to me. And that's that essentially if, if the code generator is just generating code, that's not all that interesting. If it's generating your code, then it's much more interesting if it's helping you to generate what you want.

One of the interesting things about this Pyra product is that it doesn't just spit out code. It's a conversation with the developer. Um, I'm not sure that that's the right way to do it, but it's a good way to do it or better way to do it.

To your point on Microsoft copilot, I had the same opinion that copilot was gonna be radically successful, mainly because it's right there in your IDE, it has access to your code. I mean, Microsoft has GitHub, the same with Vertex. I mean, if you're a Google developer then you know, Google can do a lot with what's already there with, with the rest of your, your corpus of of, of code, a lot of these open source, what you're calling edge, uh, co-pilots or edge products.

I mean, that's kind of what they're doing too, is they're watching you, they're helping you. It, it's kind of like, you know, sono when you're, when you're working with, uh, an AI writing tool, uh, there's a big difference between one that actually knows your writing style and one that's just spitting out words, right? Yeah, absolutely.

Um, I personally do not put a lot of, try to not put a lot of faith on, uh, tools like that that do coding. 'cause you know, I keep hearing from a lot of people that, you know, sometimes the coder earns there are bugs and it takes away a lot of time to really fix them and get them in shape so that they can run it. So what I personally think is it, it's wiser to actually take it with a grain of salt, really.

And uh, see that the, it is these tools are basically meant to help you than to just, you know, do your job for you. So you can maybe use that code as a template and do your changes. And even if they're tying in all the context and all that information and personalizing it to your needs, it's still important that, uh, there is a human in the loop who's like actually looking at it and making sure that it's on well and good to run.

'cause it's code after all the smallest mistake and make the biggest, uh, difference. This last question on this, um, I also wonder if we're overly obsessed with coding and if I look at a developer's job, coding is maybe 15 to 20% of what they actually do. And so do we need AI to do all the other stuff?

Because, you know, we're not really gonna drive productivity just 'cause we can. Well, that's the point, right? Like coding is a lot of copy and paste.

It's going, finding routines. It's like most coders if you write, you write 10 lines of code, it basically turns into a thousand, 5,000 lines of code anyway because all the libraries, right? So like, so, so I I I don't really adhere to this sort of idea that AI tools are gonna make worse code when you've got 20,000 developers in your organization.

There's a lot of those. I ask those people like, oh, I can't use, you know, copilot or I can't use this. I'm like, how many of your, you know, that that base of Java developers you have, you think actually writes Eric free code?

And, and the truth of the matter is, these tools are gonna do better than they average coder. They already approving it on benchmarks. So the argument that AI generates better, every, every implementation of coding will have bugs and errors.

The question is, will these things get better? Are they, uh, like at a point where they are better? I do.

They are. And then to your point, Mike, like coding an application is a lot more than just sitting down and writing code. It's design, it's requirements.

The, it's like the same thing. I use a lot of tools for, uh, writing. I don't let the AI write my, my articles, but what I will do is I will feed every concept and idea and all the things I want into it, right?

So I think, you know, when you talk about a service that's being built, the, the human in the middle is the design. You don't just say, Hey, I, I am a business that sells cars. Go ahead and figure out how to write software that will, you know, that will do a better job.

You know, like that's not how it's gonna work anytime soon. It's gonna be, we have this business campaign, we have this idea, and we're gonna go through design and we're gonna requirements, it's gonna be humans all part of it. But the point of where somebody just sits down and literally 80 to 90, uh, 90% of what they're actually doing is using reusable libraries.

You know, I, I've said this before, Topo Pal talked about a Capital One where he was asked one time to find how much of their, um, code base was open source and it was it like, it was, uh, not open source. How much of their code had they actually written? It was 1% 99% of their code.

He's publicly said this, 99% of their code was libraries. So like, so now we have AI tools that literally are doing the same thing on that 1% or 5%. It's definitely never more than 10% in any modern organization.

So like, like what are the decisions that are going wrong by that, that AI is doing worse than human picking, the wrong libraries, putting in the wrong configurations. Those are things that you just do ter pretty terribly. So I don't know.

So folks, I guess, you know, we're not gonna miss copying pasting code ultimately, which may account for why a lot of these repositories are seeing a, a decline in overall traffic 'cause people are just taking it from the open AI tool. Hey folks, thank you all for being on the show as usual. You guys were awesome, shared great insights.

And with that, I'm gonna point everybody to the rest of the lineup coming up right behind this show, Textron TV is coming up. Stay tuned and we'll see you guys tomorrow.