Techstrong Gang – September 16, 2024

Hello everybody. It's Monday, and we're gonna be talking about all things related to large language models and what's going on with OpenAI and some new competitors that are starting to emerge, and especially this new strawberry update. Then we're gonna jump into, well, are we making progress with software development with ai?

Because it looks like, uh, some survey data says that, you know, we're actually increasing productivity. And then finally, dockers raise prices. And on one level, those price hikes are about the equivalent of a cup of coffee.

On the other hand, there's a lot of money when you have a lot of developers. I'm Mike Ard, and you're watching Techstrong. Hello everybody.

And we're back and we've got an awesome lineup of folks today. Our guests are, once again, David Nicholson, who's a, uh, chief Technology Advisor for Futurum Group. David, welcome to show.

Good to be here. We also have Guy Courier who's CTO for Visible Impact, another arm of the Futurum Group guy. Good to see you again, as always.

Good to be back. And finally, if it's Monday, we must have Tracy Reagan. So CEO of Deploy hub, and, uh, expert in all things DevOps.

Tracy, good to see you. Thank you, Mike. Good to be here.

All right, let's get started with all this stuff that's going on with open ai. They, uh, are previewing a strawberry update that has smarter reasoning capabilities. They are saying they're going to raise more money, and they have a competitor now, and at least another one called Super or Safer Super Intelligence, which says the former scientist for open AI is, uh, raising money for a startup that, uh, you know, a cool billion to start.

So I gotta ask David, um, are these economic models that are starting to be put into these companies sustainable? 'cause sometimes I look at these LLMs and I go, well, it's nice what they're doing, and they have reasoning engines, but part of my soul says, these are interesting party tricks, but how do I kind of like operationalize that? And maybe smaller language models are the way to go, but what's your take on what's going on here?

Well, first I have to say that I was sitting on the edge of my seat waiting to hear the details of, uh, the, uh, the new co code strawberry release. Um, I wasn't sure if they were gonna be announcing something that is better at reasoning, uh, or if they were going to announce something that is, uh, not as good as the existing versions. So it was really surprising to me to hear that they're creating something that's better, cheaper, faster, stronger, uh, in the grand, uh, in, in, in, you know, in the, uh, grant tradition of all things.

It. So, I, so, so first of all, I don't think there's anything surprising, uh, in the sort of leapfrog match we're going to see among the build big LLMs. But, um, is this sustainable?

Uh, it depends on your perspective. Uh, if you are putting capital in, and if you had figured out that you can convince someone to convince someone else that what you invested in is far more valuable than the amount of money you put into it, yes. And that's where these investments are going right now.

This is a, this is a very, very, you know, this is, has nothing to do with ai has everything to do with where capital goes to get a return. Right now, the gold rush is in the hype. The hype is out there.

Uh, can it be sustained long-term? Absolutely not. com bubble burst.

This bubble will burst. And, uh, and I think it's gonna be sort of driven by some of the stuff that we're gonna see next week at, at Dreamforce, frankly, from Salesforce, the realization that that, that a lot of the generative AI stuff, the shiny object, isn't as valuable to businesses. The people who really pay for these things as block and tackle capabilities, essentially revamping the ui, that's gonna be the real value proposition.

So, short answer, no, it's a bubble. It's gonna burst. Lot of money will be made along the way.

You just have to know when to jump off the train before it's going too fast to hurt you. Yeah, guy there, despite all of David's superlatives, there was one thing that OpenAI did say that is not gonna be better. They said that this open AI engine is gonna be slower 'cause well, it's quote unquote smarter.

So therefore, the reasoning engine requires more time to come up with answers. Um, you know, I think most humans, I think when you get smarter, get faster, but apparently, uh, reasoning engines are getting smarter and slower. So what is your take on what is the state of these reasoning engines and how smart is smart?

I, the, the state is good and not good. Good and not good. I mean, it, it, I find it curious, um, how few details were provided have been provided about strawberry, um, AKA 4 0 1 or chat GPT 4 0 1 or whatever it is, um, because, um, how they work is clearly extremely important.

Now, my sense is that they are running, um, the queries, the, the prompts through a couple of cycles, and that's how they're doing this. So-called reasoning, um, I believe it's not really reasoning, it's simulated reasoning. Just like, uh, the answers are not real answers.

They're simulated answers. This is certainly an innovation, no question about it. Uh, I would like to know how they're doing it because, um, they're, they're implying without claiming rightly so, but they're implying, uh, that with this simulate they're not using the word simulated reasoning.

I'm gonna use that. Um, they're not outright saying that it's reasoning. They're saying it's like it's reasoning.

So however you wanna look at it, here's the point. The point is, um, I used to say all the time, if you remember, you still gotta bring your brain to research. You don't just look at data points and say, oh, here's the winner.

That's how we go. That's the strategy. You have to think and reason through it.

Oh, dare I use the word reason for human reasoning. In the same token, we're talking about the absolute necessity of human supervision and human review of what generative AI or, or derivations of it produce. And I know we're gonna talk about that later as well, when we get to using generative AI for coding.

So, what's the state of this right now? We don't know. We don't know.

And so we still need to pay attention to it. I don't think it's an issue at all. In fact, I welcome the sort of slowdown in response.

I like the idea that this engine is running through a few cycles and trying to, and that they are improving quality, but we're not at this point of artificial general intelligence where we're interacting with some sort of intelligence in a, um, really, you know, back and forth way as opposed to putting in a prompt. You're still, I mean, you're still doing that. You're still putting in a prompt, getting a result.

You have to look at it. You, you are the truly the one with the full context, not the engine. All right, next time I disagree with somebody, I'm gonna accuse them of simulated reasoning and see how that goes.

I hope you do it to me. I don't really even consider what the way they describe what they're doing. I don't even consider reasoning, because if you think about how human reasons we pull in, we pull in data from multiple points, kind of.

And I, I think, uh, last Monday, I, I, I briefly talked about the interest I have on this, um, multimodal, uh, large language models where data's coming from multiple places to do that reasoning. What they talked about was training, kind of like training a dog, right? You get the answer, right?

You get a treat, you get the answer wrong, you don't get a treat. Um, that to me is not reasoning. That's still training no matter how they wanna spin it.

It feels still like training to me. And the concern I have, if the mo this reasoning model is going to be based on that kind of a logic, then there's, it's much easier to build bias into it because you're now dec decide. You, you, you know, you're, you're, you're, you're giving it a positive response for the answer that you expected.

Um, it kind of sounds like our political bubbles to be quite honest. So that was concern. It's very much Like, yeah.

So that was my concern with it. And I love OpenAI. I love chat ct.

I use it every day. I'm thankful for what they've done, and I know that they have to continue to, um, to innovate and make announcements. And this one felt a little bit, you know, they have a concept of reasoning as opposed to actually building reasoning into the models.

But I'm not in the business. I'm just one looking at it and going, really, you're gonna give it yes or no an or good feedback, negative feedback for the answers it produces. How do we know we're not saying, that's not the answer I wanted.

David, you mentioned Salesforce, and they're part of this move towards what they're calling, uh, agentic reasoning, or the idea that I'm gonna train a small agent to perform a, a task and it's narrowly trained on a set of data and theoretically more accurate. So where does the reasoning need to occur? Is it really gonna occur inside these large language models, or will it be more inside of these agents that we're building that will then orchestrate to create a workflow and complete some sort of task?

I, I think if we start going down the path of defining what reasoning means, it becomes a, a bit of a, a bit of a journey down a down a rabbit, you know, following the rabbit down the hole. Um, uh, first I'd I'd like to say to our AI overlords, I would never compare you to a dog like Tracy, like Tracy did, and I, and, and, and, and, and, and, and, and I wanna say that I'm actually favorably impressed with the level of reasoning that's been achieved. And, and, and, you know, and good boy, good boy.

That's all I can say. But, but no, we, this, This is me patting you on the head right now. Yeah.

This we got, Hey, I, you're all, you're all, you're all new species. ISTs are all gonna be doomed in the future. And I, and I will be at least a loyal court subject.

Um, no, I think that, I think the main thing is kind of take a step back here. I, I, I, I liken our relationship with AI as sort of like, you know, we are, we are all digging with our hands right now. That's how we effectively move dirt around in it.

Uh, AI is presenting us with a set of tools, basically shovels a shovel is an amazing leap forward compared to digging with your hands. When we start talking about reasoning and the kinds of things that are happening on the frontier of LLMs, those are all, that's sort of, that's bulldozer, crazy earth mo move step, move mover stuff. It's really, really exciting.

But I still think the value is gonna be down in those action models, the agentic ai, which has nothing to do with running agents on servers, to be clear, it has to do with an agent executing things like they're your assistant. Those are boring kind of UI tasks. If I can say to my phone, um, Hey, get me my favorite order so I can pick it up on the way to that meeting and let everybody know I'm five minutes late and it can stitch together all of the context and deliver something.

To me, that's a huge step up in productivity. So the reasoning stuff, um, I liked guys quantum answer. You know, it's good and it's bad.

Uh, or, or, or, or, uh, however you started it out. Yeah. The Schrodinger answer.

Yeah, the Inger. Exactly. It's both dead and alive really until you, until you, until you look at it.

Um, I, I like the fact that they're saying, Hey, we're gonna have some more complicated thing. It's gonna take longer to execute however they're doing it. But allegedly, it will be more a, a more thoughtful response, right?

At the other end of the spectrum, we're, we're talking A IPC, uh, you know, edge, where the idea is responsiveness. Responsiveness. Uh, so, um, I guess the one positive I see is sort of the segmentation of, of, of, you know, the way that these tools are going.

Not all ai the same as every other AI thing Can, but can I get back to the funding, Mike, to the funding question? Yeah, absolutely. iQuestion, because, because it, you know, Dave, what you described, I completely, you know, believe, uh, you know, believe in and agree with, which is this kind of idea that, you know, the shoe's gonna drop, the other shoe's gonna drop eventually, right?

And I think we talked about that in the last po uh, the last, uh, uh, Textron gang that I was on. Um, that, uh, that I feel we've, we, we sort of sense this building sort of doubt about AI as people are getting more experience with it and this caution. Um, but I think that, that there are a number of mounting issues, um, in the AI industry as it, as it develops.

I mean, it's still, it's still going gangbusters, right? But, um, the supply of silicon and of power, um, just to, uh, run, I, I, I, I think, you know, we're, we're approaching, we may be approaching, well, I don't know when that would be, six months, 12 months, this point where, um, the global technology is unable to support, uh, maybe, maybe people will stop doing Bitcoin a little bit so that we have some of that power capability to fund more ai. Uh, there may be those trade-offs between that and, um, and, you know, if there's any sort of sar global broad sentiment of doubt and concern and worry, um, about how the, how these ais are performing.

I, I, that I think is what strikes me so much about the strawberry announcement is it is, it feels like it's so six months ago, which this will be six years ago. Like, this is great. It's perfect, it works.

One, don't look, don't look behind the curtain. Um, it's just reasoning now and now it's even better. Yay.

Full steam ahead. And I think that that shows a little bit of tone deafness as to what enterprises are really looking at right now, um, uh, uh, to make good decisions and, uh, not, you know, be disappointed with the investments that they're making. I think Guy is probably as right as an LLM on this, right?

Because, you know, everybody's taken an opinion in, wow, just wow, Smarter than a dog. I simulated that whole response, Thank people, and I, you know, I had the same kind of reaction to the, the dollar amount, um, that's going into one company. It's 'cause because you have to ask the question, um, that kind of money going into one company at the risk of what, right?

What, because there's a, there's a finite amount of money that's gonna go into these companies. So are we, why are we putting so many of our eggs in one basket? And what technology are we not going to see because there wasn't enough money, or there's not enough interest now in ai, everybody got scared of ai, it's gonna take 10 years, it's gonna be a 10 year return.

So let's just put it in open ai. I don't think that's a good strategy in the bigger picture for the broader community when you step back, because I do believe that it restricts investment in newer and, and maybe more interesting innovation in this space because we're early on. These are early days.

And to put everything into one big conglomerate kind of company is, I think, a danger to the technology itself. David, can you explain how we get to these amazing valuations, by the way? Because, you know, if I raise, you know, X amount, a couple of billion dollars, and suddenly I'm worth $150 billion, you know, that sounds a little, you know, Ponzi like to me.

So how did you kind of, how does the, those numbers get arrived at? I was gonna use the word Ponzi, and then, but then I, I held back because I thought, I thought Mike, Mike called me. Did you call me egregious earlier or, uh, or rambunctious or, I, I, yeah, rambunctious, egregiously rambunctious.

I was gonna stay away from the Ponzi scheme. Uh, I don't know. I don't know if, if, if I, if, if we have any, uh, uh, fans of the show survivor, it's a big, my, my family's been watching that show forever.

And, uh, and I would, I would say that it is actually, it is absolutely a societal experiment. It's not just stupid reality tv. Um, they have a thing where they have a fire making contest to decide nearly the outcome of, of, of the, of the entire season.

And in that fire making contest, the goal contest is to build a fire that gets up as quickly as possible to burn a cord that then lets a flag raise, and then you win. Now, if you've ever built a campfire, you know, there's a difference between building a sustainable fire, building a fire that just generates smoke and building a fire that licks at the bottom of the cord as quickly as possible and then, and then goes out. These investors are all in the business of creating as much smoke as possible to convince others to come to the campfire area and jump on board the train so they can jump off with their returns.

That might sound like a cynical approach, but I've, I've done, uh, at last count almost 1500 consulting engagements with private equity and venture capital firms, and that's exactly what they do. So I would like to know when someone says, we're investing a billion dollars, understand no one, no one showed up with a, with an armored truck with a billion dollars in cash and threw it in this guy's lap, does that billion dollars mean a commitment to fund over a period of time? So is it la Yeah, we we're committed to funding you up to a billion dollars.

Here's your first 50 bucks. Go buy some, uh, go buy some sticky notes for your, for your office. I would love to see the way that this is all structured, because frankly, it, it's got all of the hallmarks of, of a, of the pyramid scheme that all of Silicon Valley is based on.

They want a 10 x return on their investment, and they want that within three to five years. So that's, that's where this is going. Watch the money, go in, watch the money go out.

Um, and then my, my final thought on this is that, is that we have too much riding on the success of AI right now. Everyone needs AI to be successful. Everyone needs AI to deliver efficiencies in our economy that are going to absolve us of all of our financial and monetary sins over the last decade.

So there's a lot riding on this, and there really isn't gonna be anybody to call the, you know, to say that the emperor has no clothes. So I think it's a, it's, it's a bubble. Valuations are nonsense.

I think we're gonna, it sounds like a great plane. I think we're gonna, I think we'll leave this here. I think this saves super intelligence.

People are saying that they got a billion in cash, but, um, yeah, Yeah, maybe they, maybe they did. They Got a billion in funding and I don't think it was cash. You know, the, the, how that, how that funding is gated and, and how they realize it and stuff.

I think, you know, it's a really good point. I, I like their mission, um, for, you know, better accountability, safety, security and so forth. Uh, but yeah, um, lot to, like Dave said, they don't just get it all at once and say, Hey, go, go have fun.

We'll get back to you in a year. All right, folks, well, the smoke is here, obviously mirrors are next, and then we'll see where it all goes from there. We'll, a minute.

In a world where every line of code powers the future, every keystroke can introduce new threats As software evolves, so must security, it's time to rethink how we protect our digital world. Join the leaders in DevSecOps and AI at the OpenText DevSecOps Virtual Summit on September 24th. Discover how innovation is transforming software delivery faster, more secure and smarter from AI driven security to the truth behind cloud security.

Get the insights that will keep you ahead of the curve. Don't just watch the future unfold. Be part of it.

Register now and secure your place in tomorrow's world. All right, folks, we're back onto our next segment. There is a report last week put out by OutSystems.

They make a, uh, platform for low code and in partnership with, uh, KPMG. They did this study and it shows that at least it looks like we're making progress with automation in ai. Um, if I remember this correctly, 75% of the respondents said that they think that they have improved the rate at which they're developing software by 50% or more.

Tracy, you and I have chatted in the past about, in many cases, DevOps processes as we know them are broken. And so to David's earlier point in the last segment, you know, is this like we've been digging with our hands and we now have a shovel and we're really excited? Or is this a, a major leap forward?

I think it's a major lean forward. Um, I know from what we're experiencing with, you know, tools like copilot, that we are becoming far more productive. Uh, but it's just in the coding steps right now, and I'm looking forward to it.

Uh, going beyond that, being a DevOps person, I've been looking at what are, what are ways to improve DevOps with ai? And I think that there's a lot of work yet to be done in that area. But what I do think is interesting and, um, encouraging about the survey is the adoption of AI at the, at the coding level.

That happened pretty quickly. Uh, developers got on board pretty darn fast, uh, compared to sometimes other, uh, tooling in the software development world. We can get, you know, software developers are agents of change, but we don't like to change ourself.

So for them to adopt new tooling like this at that speed is definitely a lean forward. However, there are, there are areas of software development that still need so much to so much exploration, like the, the, the life cycle. And they talk about that in, in, in the, um, in the, in the survey, there's been interesting things that are happening in kind, in, in this space that we don't always talk about.

So for example, um, DARPA did a competition, and we go back to talking about the AI challenge. Um, it's called the AI cyber Challenge. And in that challenge, they were able to, and i, I, I believe it was maybe the open SSF was affiliated with this, they were able to do re identify vulnerabilities and do auto remediation of those vulnerabilities.

Now, now we're really talking about what we need to see in terms of code generation. So applying code generation to solving problems is what I would like to see more of. Not just generating code, because remember, the code that we do generate just comes from the, the, the, the data that's out there and the data that's out there isn't necessarily the that.

Great. So while it talked about the possibilities of vulnerabilities going down, I don't believe that to be the case. We're still gonna consume, um, OS packages.

We're still gonna have vulnerabilities. It's not gonna stop the vulnerability, the, the, the CVE march. But what it could do in these kinds of, of applications is to start solving some of our problems.

And boy, DevOps needs a remake. We are way behind the gun here. Everything we do is very static.

It's based on workflows, statically coded workflows, you know, pipelines are brittle. We don't even wanna add an sbo. So there is work to be done in the space.

And, um, I'm glad to see that, that the development community at large has started to embrace these, these code generation tools and using AI to do a better job of what they do. Mm-Hmm. David, what's your take on this?

Because we were talking about the reasoning engine. So can I apply this to the DevOps workflows in a way that will help us deal with all this code that we're now generating across man and machine interfaces? Or is the reasoning engine just not gonna be smart enough to make a fundamental difference here?

And, you know, it'll be a help, but it's not gonna really move the needle? Um, you know, I'll, I'll tell you, I think that my best contribution to this little segment is to maybe ask Tracy Question. So let me, let me answer your question by asking Tracy, and we'll all listen in on her answer.

Um, so Tracy, when we talk about code development in this context, I think you alluded to something, and so give, give me a sanity check here. Give us all the sanity check. Is a lot of this today when we talk about code development, it's not writing stuff.

It's that you're generating stuff from nothing. You are assembling a lot of existing blocks. Was that your point?

And that some of these existing blocks aren't the greatest thing, and until the AI is smart enough to go in and fix all of that, what you're left with is the potential for assembling things that aren't, that are, that are, that are less than optimal. Um, I mean, is that, is that the right way to look at this? Because I I I, I hesitate to talk about a reasoning engine when all it's doing is assembling Lego blocks as opposed to fundamentally starting stuff from the ground up.

Am I, am I, do I understand that, right? Absolutely. Um, uh, my niece, uh, took her first job last summer in, um, coding, and she claimed, she told me that she can generate 6,000 lines of code a day, but she said, but that's those 6,000 lines of code she has to actually work through.

So they are blocks, and it is like, you know, we're, we're, we're putting a Lego system together with this code generation, and most of it does come from existing, you know, code bases that is pulling that information from. So if, if the code that you're using is if, if, if, if what's out in the world that we're pulling in all the, everything that we see in GitHub, um, none of that's perfect. So while we're doing code, we're generating it, it's great.

It's kind of like using chat GPT to write a paragraph, right? You might, you might build, build on what it says, but you do have to build on what it says. So it gets you to it, it, it's a good jumpstart.

For example, you know, one of my str, one of my struggles, and I struggled with it forever as a young coder, was tables, building tables. So if you have something that's gonna give you that code and you know you can work with it, then it's a jumpstart. And I believe that's where we're at in, in this kind of technology.

Absolutely. And where we need to go is to be able to fix problems with it, not necessarily just generate code. That's why I was really excited to see that, that cyber challenge and how they were able to find vulnerabilities and then do self reme remediation, that's where we need to be.

There's so many ways in which AI assists in code development. Um, I'm sure you've done it yourself. I've done it myself, uh, where in the past I would, you know, do searches and, and poke around and look for a, a reasonable example.

Um, it, you know, uh, using, uh, Gemini or some other similar AI tools, uh, to, um, to, to just ask it to produce the code. It's doing all, it's done, all that searching for me, essentially. And it spits out something, you know, immediately that I can, that I can, like you said, work with, I've maintained almost from the beginning here, that the advantages of AI are not really speed.

It's not something that does something for you. It's still a productivity advantage, but it's not that it does it for you, it does two things. It, it improves your consistency of execution.

You don't have to sit there wondering, puzzling, researching, whatever, or in my case, procrastinating. You just go and ask and out it comes, and then you can work from that. That is, that makes me more reliable.

And then the other one is quality. I was really interested to see that the number two benefit, I think it was 56% or something of the respondents of the, of that study, Mike said that they expect AI to improve quality. It improves quality because now you are reacting with maybe, you know, the right side of your brain as much as the left to something that's been presented to you.

And in the same amount of time you can produce something that's higher quality, you can test it more, uh, uh, sandbox it quicker, that sort of thing, right? And there's all these other advantages that AI brings to coding, such as producing test plan, test plans for you, documenting for you, all of those things that, that lead to quality. I think the use of AI in, in ENC coding, um, is a really a landmark adv, uh, uh, improvement to, um, that, that AI is providing, you know, to everybody and to the market indirectly instead of directly.

And these AI augmenta, oh, sorry, I was gonna say these, these AI augmentation tools, especially in the coding world, they're productivity tools. And in that survey, most of them said they would be investing in AI augmentation tools in the next two years. We could have asked that question 10 years ago and put productivity tools in that instead of AI augmentation, and it would've been the same exact answer.

So how are we making, you know, how do we make developers more productive? And that's really the question they're trying to answer. Well, this whole topic about AI and software development is taking place against the backdrop of this conversation we're trying to have about platform engineering.

In my mind, and this is, there's a debate that I hear about everybody kind of nods their head and says, we need to make DevOps work at scale. So we should have a centralized platform engineering team to go do that. But a lot of folks embrace DevOps to get out from underneath the centralized IT team in the first place.

So maybe we should just go back and fix DevOps and the processes that go into that and make that scale. And we don't necessarily need to, you know, put this under the direct rule of some CIO somewhere. I don't know, Tracy, I mean, I hear this debate, it's a cultural issue.

I, you know, uh, I was a programmer in the, in the mainframe world when I initially started, and Endeavor was what we used, which stands for environment for developers and operations. So they were doing DevOps a long time ago, and it was centralized, it was easy. It really was because somebody told me what I had to do.

I didn't have to go and build my own endeavor, you know, lifecycle. I didn't have to manage my own endeavor. Automated configuration management, somebody else was doing it.

Um, I have a lot of stories that justify the, the need for a centralized, uh, you know, a DevOps team or software platform engineer, whatever you wanna call 'em. Now, however, there was a restriction on innovation in that world. And with the number of, of languages that are out there, the changes in the technology, I believe that there has to be some level of flexibility in the, in this type, in the DevOps pipeline.

You might be building now AI applications, which the DevOps pipeline needs to, to, to, to be pivoted for that. So I, I think that the world that we're in requires that there be some level of compliance for the DevOps pipeline, high level compliances, nothing that requires a particular tool or language. The compliance is what's important, not necessarily the process.

And on your, your question about reasoning, I have seen real, and, you know, truth tables are, you know, baby ai, I have seen really interesting truth tables be implemented in DevOps that allows a question process, that allows a fall through process that says, go execute this because this is an emergency release. Take it directly to production, as opposed to, this is a high risk deployment and we have to take it through a longer DevOps lifecycle. So reasoning in DevOps is, is required.

And again, this is another area that AI could help us with, uh, because oftentimes our a our pipelines, our DevOps pipelines are so static that everybody goes through the same one on that team for every kind of release. And that doesn't make sense either. So yes, let's add more reasoning to our DevOps pipelines, but we have to get out of statically coded pipelines.

So there's work to be done in this space. AI could really transform how DevOps is done, but as long as we're really stuck and we're not even doing these, these, uh, building workflows that are event-based, we're kind of stuck where we are. Tracy, don't you think that like, you know, um, just like a lot of these, uh, recommendation engines are helping, uh, it operators to, uh, uncover issues or deal with them proactively or see, see complex, see and be able to implement, you know, somewhat more complex, you know, resource allocations and things like that, that there is this sort of, sort of world of, um, maybe within, you know, specifically within the DevOps problems you're describing, there's this sort of world of application and tooling for, uh, using AI to, to provide the same kinds of recommendations for things that people aren't necessarily seeing or don't have time to look at.

Not just automating it, but, you know, like I said, making recommendations and allowing people to, to find ways to stay nimble. Absolutely. Uh, we should be able to interrogate all of the Jenkins workflows checked into GitHub and determine which ones are the best ones to execute, right?

The, we should be able to see the flaws in the, in our, in our workflow pipelines. There's millions of them. It's not like we have a lack of data.

There's mil, literally millions. I, uh, CloudBees claims that they manage about 90 million workflows a month. So the data's out there, it's just not being applied.

We're not collecting it. And the other problem in the DevOps space, in terms of AI, is that we do not collect the data of the results of all of the tooling that's running under the pipeline. It's all fragmented.

Some tool tooling is kept in security tools. Some, there's a, there's a deployment log sitting out there. We are forced to generate an SBO m but what do we do with it?

We leave it in a text file and maybe it gets checked into GitHub, or it's sitting under the build directory somewhere. So the data that's critical to DevOps that allows us to do that reasoning and to do that, making recommendations on how to approve, is just sitting under the, it literally is sitting underneath the pipeline doing absolutely nothing. So these are the things that I worry about and think about how to change in the DevOps process, because this has been a passion of, I've done this all my life.

Um, we are just, I, I, I can't explain it anymore. We are stuck because of the technology that we currently use for building out the pipelines. And nobody wants to change it.

There's no money going into changing it. 'cause DevOps is no longer a shiny new object. Well, it is when you put AI in front of it and then away we go, right?

Possibly, possibly Putting AI in front of it. I, I think it, I I, I, I always go back to this principle, garbage in, garbage out. Uh, and, uh, Tracy alluded to, and, and, you know, nothing going in, nothing coming out.

So it's like if the data isn't even accessible, you can't reason over it. You can't do anything good or bad with it if you can't, even if you don't even know where it is. But what we're, what we're sort of presenting to the world in a lot of these cases is, Hey, you know what?

With enough horsepower and enough reasoning, you can go out to your local county dump and mine for diamonds. And it's like, well, but if there aren't any diamonds in there, you're not gonna find any. And so part of this is this idea, and again, and I'm, and I'm, I'm drawing a parallel, Tracy, check me if I'm off, if I'm, you know, if I'm outta line here, but, but I think in terms of the quality of data going in and data hygiene, is there, is there a parallel in terms of code-based hygiene?

At some point somebody has to go in a, a person and, and clean this stuff up? Is is there an issue there? Are we just building on top of a shaky foundation with, with, We at least have a code base.

We don't have a DevOps base. Okay. Right.

Where do we pull that information from? Where is it stored? We don't have it.

It is fragmented everywhere. And even if a single company, you Had one job, Tracy, you had one job, and that was to make sure that all that stuff, you're the DevOps person here. Hey, we're trying to do it at Deploy hub.

This is why it's a passionate topic for me, because when I look at the, the how much data that the DevOps pipeline generates and where, and how fragment, how fragmented it is across the tooling and logs, literally logs that, that information is what's gonna build our, the next generation of DevOps. And when I look at, you know, I, 'cause I watched, uh, all of the CICD tools, all the pipeline tools, and I watched their announcements with what they're doing with ai. It's not transformational what they're doing.

They're adding ai so they can put AI in front of it, but does it really help us? Is it really transforming the way we work? And AI has the potential to do that if we had the DevOps space, like we have the code base, but we don't, All right, well, we're, we could talk about this for an entire virtual event, but you know, David, there is getting a little unreasonable.

So let's get in a different set of prompts and we'll be back in a minute. 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 now we're talking about something that well, maybe doesn't have anything to do with ai, but we'll see. Um, Docker has raised its pricing for its, uh, tools. And at the same time, I was trying to say that there's more value in that.

I think if you add up the price hike, it's roughly in the realm of 40%, but we are talking about, you know, things itself or somewhere between five to $12 per developer per month. So it's not like, uh, shall we say a Broadcom price hike, but nonetheless, developers can be a little bit sensitive to these things. Um, guy, when you look at this, what do you see?

I mean, is there money for value here or is this kinda like dockers just passing on some inflation costs? Right. Well, I definitely see a vendor passing on inflation costs.

Um, I note the, uh, what they claim is a nine figure investment, a hundred million dollar investment in Docker hub, taught to, you know, uh, uh, improve its capabilities. Um, I also see just, you know, capitalism in action. Um, my my general sense, uh, is, you know, almost like, not Docker itself necessarily, but Docker's dominant position in containerization is almost, uh, uh, subsidizing the industry up to this point.

Um, they, they have not lacked for investment. They've not lacked for support. Um, this is not a brand or a company, you know, that has to worry about its future, particularly ha has had to worry most about growth.

So I, I think, you know, possibly that phase has gone by at this point. Um, it's kind of funny, we were talking about, um, how, you know, uh, DevOps, you know, was the, the, you know, the thing of the moment for quite a long time now, it seems, uh, you know, apologies to Tracy, it's like not really getting the attention maybe that it deserves right now because of AI taking over everything. AI's gonna do everything right.

Um, in a lot of ways, containerization had its moment in the infrastructure. Um, also now somewhat being taken away by ai. AI is just such a dominant, you know, topic of conversation.

So is Docker trying to recoup, like they say, this huge investment that they've made? That's kind of a weird thing, especially when they all also are saying at the same po same time, it's only gonna affect 3% of, uh, their qualified, uh, user base. I don't know how they're gonna make a hundred million dollars back, even with a near doubling.

Um, 'cause that's how I read it. It was like a doubling of the subscription price at the end of the day. Um, the question we want to ask ourselves is, is this going to affect how infrastructure is run?

Um, developers are using Docker, enterprises are using Docker. There is an, an ecosystem surrounding every enterprise that has its own developers and a network of, you know, uh, open source freelance developers, a whole, you know, sort of cloud of, uh, uh, pardon the word, but a cloud of contribution to what they're trying to do. And a lot of it anchors on standards like the Docker standards.

So is this going to cause any problems there? I think you're right. It's a low enough price point to begin with, and we do want Docker itself to be, you know, healthy, um, Docker Inc.

To be healthy because of the contributions it's made globally to how applications are transforming and being run on a scalable basis. Um, we're gonna find out, Docker's gonna find out if this, uh, causes any glitches, and then everybody's gonna feel it, and then they'll lower the price. And capitalism will once again reassert itself.

David, speaking of capitalism, you know, we saw Broadcom raise prices and now Docker, and I can't help but wonder if I have a sneaking suspicion that every vendor out there is trying to figure out a way to maybe raise prices. Yeah, I, you know, look, I, my educational background's not in computer science. Uh, it's actually in economics.

And, um, the, the, the, there's a challenge in any field where people are really smart and accomplished and they've got domain expertise, uh, sometimes they fall prey to having that level of expertise in their minds spill over into other areas. But I will tell you that it is staggering to me the level of lack of understanding of how the economy works. In, in what I, in what I hear, you said the word, the only word that matters in this conversation, if you're talking about pricing inflation, there are $21 trillion that exist in US currency.

40% of those dollars were created out of nothing in the last five years. So this is across administrations. This is not a political comment.

This is monetary policy. It's called monetary inflation. When you increase the amount of something, each unit of that thing becomes less valuable.

So a $5 to $9 increase is actually pretty much in line with what $5 bought you 5, 6, 7 years ago. It takes $9 to buy that today. No one wants to touch that because that's sort of a political hot button.

Uh, but the fact is monetary inflation affects everything. So yeah, they're trying to recoup it. It's a, it's a tax that they're trying to pass on to consumers.

And the thing that determines al as guy says, you know what, how will the market react? It's elasticity of demand. How much do you need it?

If I tell you, you know, I'm ri I'm raising the cost of air that you breathe, how much are you willing to pay for that air? Well, it's like, yeah, in an infinite amount of money, you know, until I, until I pass out from lack of air, we're gonna see how elastic the demand is for what this organization is offering. But everyone at some level over time has to bake in the effects of monetary inflation.

Uh, the, the thing though, Dave, is Docker made this decision somehow or other, and I don't think it was by thinking about, you know, the money supply. I think it was by looking at two things. One is their p and l bottom line, and the other is the expectations of their funding sources, their investors, their stakeholders.

If, I mean, let's think about, you know, the Amazon model for, uh, uh, uh, approaching any given market. They've done it successively at various markets, starting with books, which is their model. If you're familiar with their, their pinwheel or flywheel or whatever they call it.

It starts with offering extremely low prices, if not at a, you know, at literally at a loss, which is how AWS worked for quite some time. So they build their market that way. Then they work on whether it's quality or, or pricing or whatever it is, then they work on profitability afterwards.

I'm not saying this is what Docker did, I don't know, but it, it looks a lot like that. It looks a lot like Docker was a a, a market darling for a long time had no problem worrying about its future. The executives there had didn't have to worry about their bonuses, their compensation, or anything like that.

And the signal here to me is, uh, if they are starting to become passe, not really passe, but less hot, then they have to think that, well, we, we need to find a way to be hot and that just might be in profitability. That's why we raise our prices. They, the, but the point, again, I gotta push back a little, Please do, because I'm really way out over the edge.

This Increase the price, increase the price increase, all it can do is maintain profitability. If they lose no customers, that's all it can do. So, so, so if the folks at Docker aren't factoring in monetary inflation, then they should all be fired because that is, that is like I, I guarantee you the CFOO there understands this, that do that, that a dollar today is worth far less.

All dollars are worth far less than they were five years ago. It's just, it's, it's physics. It's, again, I know it's hard in, in like the computer science realm.

It's like if I, if I said if I drop this ball, what's going to happen on earth? Everyone would agree it's going to fall to the ground. But you can find really, really smart folks in computer science who are confused about what happens when you inflate currency.

The only answer is each unit is worthless. And so it's gotta be reflected so that the best they can do with this price increase is maintain their level of profitability. That's, that's the challenge.

Now I'm giving, there's a fudge factor in there. Maybe their price increase is accounting for slightly more than just monetary inflation, set aside energy inflation. Um, but, uh, but yeah, so again, I don't mean to be cynical on it, but it does come back to sort of first principles of, uh, of, of economics here.

And uh, and, and of course it's a lightning rod. 'cause people think it's a political discussion. It's not, we've all decided that we're okay with this.

This, you know, this, the currency inflation, a lot of it happened on the back of the pandemic, but it's a reality that everyone is having to face. It's not just groceries, it's not just luxury goods, but it's, It's always been the case. We're not, it's, this is not new.

Inflation is not new, it's the rate of inflation and potentially the, the pandemic created that, but this is not new. Now, when I was curious, when I read that I, you know, if I, now I'm not docker, so I, but if I was Docker, I would be looking at some of the trends that we're seeing, um, they charge per developer. I've never understood that model, to be quite honest.

I a, you know, a per developer model is, has always been odd to me because what if we do, what, if we're looking in the future and we're looking at AI and we're looking at fewer developers, they're going to be losing money out of that model because they're gonna have fewer developers are gonna have to charge for. Uh, and we have to remember the docker is open source. So companies have an, have an option they could bring on their own folks to do maintenance for Docker and potentially, uh, be a lot less expensive for them.

So I felt like Docker, while I do believe that they deserve to bring their prices up, I think that they should have looked at their entire pricing model. And this u this idea of a per user basis just doesn't work for me anymore. It doesn't, it, I I don't feel like it's a, a good model for pricing any software.

In my opinion, Inflation, inflation is nothing new. Inflation is nothing new. 40% monetary inflation over five years is unprecedented.

And what it, and what it has driven in terms of the layer, you know, uh, uh, debt servicing, uh, you know, is getting to the point where the interest on our 35 trillion in debt in the US is the equivalent of the entire defense budget at this point. Or it's, or it's, or it's close to it. So it is, it is, yes, it's, yes, it's more of the same, but we are in unprecedented times where we don't, we don't quite know what's gonna get us out of this.

That's why I go back to this, this idea that the, the thing that is supposed to deliver us from this is increases in efficiency in the, the economy delivered by ai at this point, technological advancement is the only thing that can push that thing we call the production possibilities curve in economics outward and account for this monetary inflation. So it is, it is. I I hear you.

It's, I I'm, I I don't mean to be chicken little Davy downer. Um, but it is unprecedented. No, I, I agree.

And that, and I'm saying that the efficiency, potentially the efficiency will hurt them more than anything else if they're charging on a per user basis. Yeah. So Really, yeah, I'm Gonna, when they went to change the pricing, they should have thought about that and maybe decided that enterprise support and, you know, different levels of support you could have, have bought into.

I think customers are gonna ask following, and at the risk of delving into some simple supply side economics, they're gonna be, is the cost of building software getting more expensive? If so, how? And software.

It's like, you know, if I need another copy of it, I just make more bits. It's not like I'm gonna have to go manufacture it all over again when I want to give it to more people. So how does one really assess the cost of building software and then put a value on Energy?

Mm-Hmm. Ener, I think e ener energy's energy's a big cost. And we, we alluded to that earlier, the, the lack thereof.

Um, and, uh, so that is, I mean, one of the big co obviously there's labor going into it, and, but yeah, I I get it free generated next, you know, uh, copy next is sort of free, um, except to the extent that it's being that, that there's processing involved that that's drawing electricity. So, um, the, the, there are certainly incremental costs that have to be accounted for. My, my my main point is that there are sort of fundamental things going into this that they have to account for.

And, um, and if they're clever about it, great. If they're not clever about it, it can, it can crush them. Uh, Tracy, they say AI is gonna drive the cost of building software down to zero.

It's gonna be free. Is that true? No, sorry.

No, it's not. No, it's not. That's not gonna, that's not gonna happen anytime.

Not in, not in our, uh, lifetimes anyway. Not any software. This is a really Interesting, this is a really interesting, uh, uh, uh, point you're making though, Tracy, which is, um, I, I su I don't know, you know, what you have against, uh, seat based, you know, pricing for SaaS or for software, um, exactly.

But I can add, until you mentioned that I really hadn't thought about it, and I could see a lot, lot of problems with it, you know, um, when you're running, uh, the, you're providing the same, you know, price with variable costs because the cost per seat, you know, doesn't indicate, I mean, the price per seat is not reflective upon how much each seat, how much of your resources each seat is using. Maybe that's what you're talking about, maybe not. But um, if the number of seats you have, if you're expecting them to go down because of something like ai, then uh, you'll want to raise your prices.

And so maybe that's their motivation, um, prior to some longer term strategy of changing how they charge. Um, I think it is probably two things, right? Because it's not just per seat, it's per seat per tier.

Like you, like you alluded to, and you may need to multiply the number of tiers or provide a menu of options, you know, each would, each one of which is a fraction of the seek cost or something along those lines. Well, It's not so straightforward. Yeah, It's, we, we ran long on a couple of the earlier blocks, so I'm gonna have to end this here.

But I suspect that seats and software are a lot like seats and airplanes. Nobody's paying the same price for any of you. All right guys.

Hey, thanks for spending some time with us, and thanks for everybody sharing their insights. I wanna thank you all for watching the latest episode of the Textron Gang. Please stay tuned for all the episodes on techron tv and till then, we'll see you tomorrow.