Digital Disruption

00;00;00;00 - 00;00;14;10

Speaker 1

There's actually a good a good use case toward the end about it was actually being used to like solve a guacamole related problem. Yeah. Lunch distribution. Yeah. Exactly how we can get brunch setups to be more efficient in like, local restaurants. But if we multiply.

00;00;14;10 - 00;00;21;13

Speaker 2

That out across every restaurant in America and then the world, think of the productivity savings, what are we going to do with all our brunch related free time?

00;00;21;14 - 00;00;28;21

Speaker 1

I know, I know, it completely changes the game.

00;00;28;23 - 00;00;33;27

Speaker 3

Coming after you from the Masonic Temple. It's the latest in tech and work and what it means. All.

00;00;33;27 - 00;00;56;20

Speaker 1

Signal, no static and infotech production. This is digital disruption now. And today we've got three stories for you. Number one, anthropic has released its prospectus for its IPO, which we all believe is coming soon. We'll get into that. Apollo says there's no AI payoff beyond the tech sector. And finally, OpenAI says that COBOL may be completely a thing of the past with AI.

00;00;56;22 - 00;00;59;01

Speaker 1

Let's dive into it. What can you tell us about our first story?

00;00;59;02 - 00;01;18;25

Speaker 2

Oh, I love this story. It's our top story, Jeff. And I think it's literally everybody else's top story, too. You cannot throw a stick on the business internet without seeing something about anthropic prospectus. Got people excited. So anthropic has released this prospectus technically form S-1. It's a mandatory regulatory filing that goes out to the SEC in advance of any company going public.

00;01;18;27 - 00;01;39;27

Speaker 2

And it gives us a bit of a peek into anthropic finances and how its executives see the world. In fact, it opens with a letter from the seven co-founders that's worth reading. I think the company, while essentially optimistic about AI, sounded some warning bells. Putting it a little lightly, they're highlighting the threat of rogue agents, state actors, and economic disruption, among other things.

00;01;39;27 - 00;02;02;27

Speaker 2

There's actually hundreds of pages of threats just repeated over and over again. These, it argues, will be, of course, mitigated by cures for disease growth, unparalleled economic growth and improved quality of government services. In fact, they specifically call out the DMV. So, Jeff, if you had a bad time at the DMV or your local equivalent, anthropic is coming to solve that.

00;02;02;29 - 00;02;15;13

Speaker 2

So this is a quarter million words long. And a lot of them, as I mentioned, sort of repeat themselves. So we're going to do our best to give you kind of a high level take on it. But I was wondering, Jeff, do you think they use Claude to write it?

00;02;15;16 - 00;02;27;28

Speaker 1

I don't think they use Claude to write it, but like as you said, there's an awful lot of repetition in here. It's like 355 pages. And I know, like, I know nobody's reading a prospectus thinking they've got the next great American novel, but I just.

00;02;28;00 - 00;02;29;18

Speaker 2

Thought one was pretty novelist.

00;02;29;21 - 00;02;53;09

Speaker 1

It was pretty fun. But yeah, I just couldn't believe, you know, how boring this was, to be honest, for something so exciting. You know, there's a lot of interesting stuff in here, but it's all stuff we've seen before. Right. And also, it's the same stuff again and again. And one of the things that caught my attention you kind of spoke to it is, yes, there's over 100 pages of risks, just risk after risk from, you know, oh, I don't know if we can pay our debts to the world might end.

00;02;53;10 - 00;02;54;01

Speaker 1

It's like pretty.

00;02;54;05 - 00;02;55;25

Speaker 2

It's pretty apocalyptic in a lot of ways.

00;02;55;25 - 00;03;03;03

Speaker 1

It really is. And I was chuckling when you said, like the seven co-founders, like for some reason seven makes it seem more sinister or like more kind of like.

00;03;03;04 - 00;03;04;22

Speaker 2

Sitting around a table that.

00;03;04;25 - 00;03;28;18

Speaker 1

They saw rogues on. But the other thing that's in there is there's 20 pages just of like case studies of all the amazing things that Claude is doing for all these different companies already. And they're exciting, but they're exciting in the same way. We've been talking about AI for like, years now, right? Like I was I was deliberately looking at it like, oh, what new sexy AI use cases are there.

00;03;28;20 - 00;03;41;13

Speaker 1

And it's the same stuff. It's being used for coding, it's being used for productivity, it's being used for cybersecurity. There's actually a good a good use case toward the end about who's actually being used to like, solve a guacamole related problem. Yeah.

00;03;41;20 - 00;03;42;19

Speaker 2

Lunch distribution.

00;03;42;20 - 00;03;47;26

Speaker 1

Yeah. Exactly how we can get brunch setups to be more efficient in like local restaurants.

00;03;47;26 - 00;03;55;23

Speaker 2

But if we multiply that out across every restaurant in America and the world, think of the productivity savings. What are we going to do with all our brunch related free time?

00;03;55;24 - 00;03;58;03

Speaker 1

I know, I know, it completely changes the game.

00;03;58;03 - 00;04;14;16

Speaker 2

So we should talk maybe a little bit about some of the details that are actually included in this perspective before we get any further. So the headline number that I've seen a fair amount of is that they lost almost $80 billion or more than $80 billion, actually, if you sort of tally everything up in the first six months of the year.

00;04;14;18 - 00;04;34;06

Speaker 2

So, you know, they love annual run rates. So in the first six months you lose 80 billion. If you keep that up, you're at 160 billion by the end of the year. And that's more than they plan on raising from this IPO. So that's obviously pretty scary. They made $16 billion in revenue, which is up dramatically. I think as late as 2023, they were making like tens of millions of dollars in revenue.

00;04;34;06 - 00;04;56;10

Speaker 2

And now they're at 16 billion. That's like an alarmingly fast growth. But that's of course, you know, one of the things that they anticipate is going to change pretty dramatically now that the loss is mostly related to the value of some convertible notes that they've issued. So it's not like they owe anybody $80 billion directly. They're going to end up diluting some shareholders if the companies that they owe those notes to end up claiming them.

00;04;56;10 - 00;05;04;18

Speaker 2

So it's a little bit more complicated, more of an accounting thing than a true loss. But nonetheless, that's a pretty staggering number. So yeah, Jeff, let's talk about your takeaways beyond branch.

00;05;04;19 - 00;05;19;08

Speaker 1

Well, so I mean let's start with that $80 billion number because you kind of said, well they don't actually owe it. Well actually they kind of do write like that. That's the nature of how this works. It doesn't mean it's cash out the door today, but it means there's existing investors with shares that they owe money to. Right.

00;05;19;09 - 00;05;42;25

Speaker 1

This is organizations like Amazon, like Google. And as the company gets more valuable, they owe more and more money because of the stake in the company. Right. And so what they need to make sure they're doing is having growth that outpaces the increase in value that's owed here. And you mentioned what's going on with compute as well. Like there's just an awful lot of costs here.

00;05;43;01 - 00;06;04;26

Speaker 1

That's a piece that concerns me. And you know they say it right in there. I their IPO prospectus that they have a history of net losses, that they may never be profitable. And the way they've structured this is they've made it like if you actually happen to care about profit, which feels like a weird thing to like make an if in an IPO prospectus.

00;06;05;01 - 00;06;28;00

Speaker 1

But if you actually care about that, it's going to be a long road at best for this company to actually ever be profitable. Now, if you're here for the vibes, if you want a piece of anthropic because it's so cool, then yeah, you know, by all means, by a share it, you can treat it like a Tesla or a space X, but I just don't see it as a company right now that will ever be trading on its fundamentals.

00;06;28;00 - 00;06;50;06

Speaker 2

So okay, let's talk about that because this is where I disagree with you at least loosely. So tech companies obviously have this big vision. And anthropic I think has the biggest vision of any prospectus I've ever seen that is at least vaguely serious. Again, if you go read the the work prospectus for their failed IPO, like 7 or 8 years ago, that might have been more delusional in a lot of ways.

00;06;50;12 - 00;07;16;21

Speaker 2

But like they point out that Claude is useful for literally everything, right from brunch to e-commerce to military applications. And they don't necessarily love that. And they talk about, you know, preventing authoritarian governments from taking over. And they pegged the size of that, like cognitive work market at like $40 trillion a year. And they say that, you know, we can introduce massive productivity gains, 50%.

00;07;16;24 - 00;07;47;26

Speaker 2

And the idea is, well, we're going to capture a significant portion of the $2,025 trillion productivity boost that comes from using our technology and technology that is like ours, right? And when you look at it like that, you say, okay, you know, a $2 trillion valuation on potential total addressable market that big, while obviously like we're making a lot of assumptions here, like a lot of assumptions, this is definitely a gamble.

00;07;47;29 - 00;08;03;23

Speaker 2

But like if it works out why that would be absurd. But like, here we are. We've got $5 trillion companies. So why can't anthropic, which is the leader in the frothy space with a product that works, that people actually do, like, why can't they earn out that vision?

00;08;03;24 - 00;08;31;20

Speaker 1

So I agree with every component of that. And I will reiterate, I think they've got a great product. Like I think anthropic does some like, you know, Claude and all the various iterations of Claude are really, really effective and their market leading, you said frothy like it was a good thing, like frothy to me is like a risk of this whole space that there's so much froth here that it's really it's really difficult to separate, like what is actually true, what's actually useful.

00;08;31;22 - 00;08;55;02

Speaker 1

And all of these assumptions are like, yeah, this thing's going to the moon. It's going to keep exploding. We've had, as you said, like really, really exceptional growth in the past, but we've also started zero, right. This was a category creator and it's going to level out at some point. Now the question is where, right. Like do you go from $16 billion to 160 billion to 1.6 trillion?

00;08;55;05 - 00;09;15;19

Speaker 1

I don't think so. I mean, it's very clear in the productivity space, they're sort of trying to like speed run Microsoft. Like how do you just very quickly say, hey, we want this like on everybody's computers, spending, having these massive contracts, the issue with how fast they're doing it, and we've started to see it in practice is the costs for these enterprises are absolutely mounting.

00;09;15;23 - 00;09;35;08

Speaker 1

And they're saying, wow, like suddenly this whole like token maxing thing is backfired. Who'd have thought that saying, you know, spend as much money as possible would lead to as much money being spent as possible? So, so it's really all about where this thing lands. Look, I like anthropic. I like their company. The question to me isn't, are they doing something valuable?

00;09;35;08 - 00;09;56;13

Speaker 1

Are they doing something that's going to grow? It's at what price is this still a good investment versus it in the stratosphere? Like what price to earnings ratio can you accept and say yeah, this can actually get there. And what we've seen with and I keep picking on like the Teslas and SpaceX's of the world is you have these companies where you know Tesla.

00;09;56;13 - 00;10;17;27

Speaker 1

The last I checked it's got a p e of over 400. It's got based on their latest earnings, you know, revenue that's flat profitability, that's, you know, moving a little bit. But it's just like how can you possibly justify that. And so I, I worry a lot about anthropic going down that road. Is there going to be great growth.

00;10;17;27 - 00;10;35;24

Speaker 1

Yes. But like they really very clearly say in the prospectus that like we don't really know where it's going to go. We don't really know how much it's worth. They actually say it's kind of a funny aside, that they've got this principle that you should never, as a company, let your value be more than like $1 more than the company is actually worth.

00;10;35;24 - 00;10;37;28

Speaker 2

They said that was advice that other founders had given them.

00;10;37;28 - 00;10;49;07

Speaker 1

Right. That's advice that other founders gave them that they think is like at the core of their mission, which I think is like an absolutely asinine thing to say because as though, like you have much control over that or you're going to be like, no, bring the price down. Like.

00;10;49;07 - 00;10;51;03

Speaker 2

Well, you can bring the price of your own company down.

00;10;51;05 - 00;10;52;02

Speaker 1

You can you can only.

00;10;52;05 - 00;10;52;24

Speaker 2

Shouldn't do that.

00;10;52;25 - 00;11;05;22

Speaker 1

You can try. But but all of that is to say, there's a lot of oxygen in this, in this prospectus dedicated to the fact of like, trust us, like we're pretty good guys. We're going to be responsible.

00;11;05;23 - 00;11;21;18

Speaker 2

Well, there's a ton of that. And just to the to that example about not your company not being more valuable than it's worth. A few years ago, Elon Musk, I think it was in 2020 tweeted something to the effect of Tesla stock to high back. You know, in his questionable tweeting days I guess those have never ended. And he brought the price down 10%.

00;11;21;18 - 00;11;41;00

Speaker 2

And it launched an SEC investigation for a variety of reasons. Because when you're making materially relevant statements without being vetted, and if you're Elon Musk and you've already engaged in a settlement, that can be problematic. So yeah, there's definitely, I think, a precedent against saying things like that publicly about your own company. So they probably shouldn't come out and say, well, we IPO to 2 trillion.

00;11;41;00 - 00;11;42;10

Speaker 2

That's too high. That would be probably.

00;11;42;16 - 00;11;48;26

Speaker 1

Well. And the prospectus doesn't actually include the words were not Elon Musk. But that is sort of the subtext stressed.

00;11;48;28 - 00;12;14;06

Speaker 2

But they they do a ton of business. They do over $100 billion in commitments with space AI. Right. That's one of the things that they've highlighted. And that's actually maybe a good segue into another takeaway for me, which is like they are heavily intertwined with the economy, like to an alarming degree. I was reading the New York Times piece on this, and they tallied up all of the obligations, and they came out to something like $807 billion in stated obligations over the next decade or so.

00;12;14;07 - 00;12;38;19

Speaker 2

And every, every term is different with the different companies that they're working for. But these are leases. They're, you know, purchase commitments. That's an insane amount of money. So they're going to raise $100 billion in their IPO, and they owe $807 billion over the next few years. That could be potentially problematic if these assumptions don't come true. And so I like to say, because I've heard a million other people say it, that, you know, if you owe the bank $100, you've got a problem.

00;12;38;19 - 00;13;03;26

Speaker 2

If you owe the bank $100 million, they've got a problem. They owe the bank $800 billion. The bank is the broader economy. And so just imagine if those promissory notes are never paid. Like, imagine if Amazon doesn't get their $100 billion in compute. You know what that would do to Amazon's price, right. What that would do to the value of the convertible notes that Amazon owns in anthropic, and the spiraling effect that that could have on the economy.

00;13;03;26 - 00;13;20;28

Speaker 2

Imagine if the same thing happens to Broadcom, who they have a significant relationship with, or space AI 2 trillion almost dollar company that they've intertwined pretty heavily with. And so as I was reading this, I was thinking like, oh, they kind of have like a gun to the head of the market. They're like, continue to believe in us.

00;13;21;02 - 00;13;29;03

Speaker 2

Because if you don't, we could fail. And if we fail, that's actually like a big problem for everyone. That would be the bubble popping, right? Well.

00;13;29;06 - 00;13;50;12

Speaker 1

I mean, they may have their gun to the head of the market, but I think there's an awful lot of guns pointed at their heads to. Right? Because there's so much investment here, there's so much owed that, like, I think it's probably too strong to call it a collective delusion. But the issue is that right now everybody's got money in their under the belief of, yes, we're going to make a lot more money.

00;13;50;18 - 00;14;02;06

Speaker 2

But that's every investment though, ever, right. And people are going to make a lot of money on this. Amazon is in this to like $190 billion in value. Like people have made a lot. And they'll continue to.

00;14;02;08 - 00;14;26;21

Speaker 1

Right. But because of how intertwined it is and because of the astronomical sums and the astronomical belief of of future revenue and future growth, we're in a situation where the numbers are so big that if somebody starts getting cold feet and saying, I'm taking my money out, you know, you can basically have like sort of a bank style run and see a lot of this come crashing down.

00;14;26;22 - 00;14;40;24

Speaker 2

I mean, we call it a sell off, right? Again, I should say to to everyone, I'm going to look to camera, Jeff, to everyone watching. This isn't financial advice. You shouldn't get your financial advice from any podcast. Certainly not this one. That said, you know, when the market sells off, we call, you know, it's a sell off, right?

00;14;40;24 - 00;15;08;09

Speaker 2

It's bad. There's a bit of a collapse here. Public companies are especially susceptible to this. Right, because everything is public, obviously. I mean, I think you're you're notionally correct that anthropic is kind of a bellwether, right? It is the gold standard in AI. Companies like this company is only five years old, founded in 2021. At its core, I think they realize that they are the linchpin that is holding together this economy.

00;15;08;09 - 00;15;25;21

Speaker 2

And when we talk about our second story in a few minutes, you'll see that, like the rest of the economy isn't actually doing that well. So, like, there's more and more being piled on Atlas's shoulders here. And I will say, I do sort of buy it. Like I wrote in my notes here, I kind of get it.

00;15;25;23 - 00;15;27;06

Speaker 2

Do you kind of get it?

00;15;27;09 - 00;15;46;06

Speaker 1

I kind of get it. I just think it comes with colossal risks. As you've said, the rest of the economy is not doing all that well. We'll, we'll talk about all that soon. But you know, if you pin all of your hopes and dreams into one company or just a few companies, the the pressure to perform gets very, very high.

00;15;46;09 - 00;16;05;06

Speaker 1

And it just again, to use your word, it gets very, very frothy. And I see a lot more based on the numbers they're talking about. I see more downside potential than upside potential. Now, you know, could we have had this conversation about Google 25 years ago? You know, probably we could have said the same thing in some degree.

00;16;05;10 - 00;16;05;14

Speaker 1

Well.

00;16;05;14 - 00;16;28;05

Speaker 2

Google is always been like especially except for their very, very early days, like the Google business was was pretty clear, like Google wasn't posting an $80 billion loss in advance of IPO. If I had to summarize this, Jeff, I would say it's a bunch of guys saying, wow, this thing's really powerful and it's really dangerous and everybody else can't be trusted to use it.

00;16;28;05 - 00;16;51;13

Speaker 2

But us seven founders, you can trust us because we recognize how dangerous it is. And if you do trust us, we'll all make some money along the way. You know, that's sort of how I read this. It's it's very doomy, but also very optimistic. It's very distressing in some ways. They say some of our decisions may look irresponsible, you know, especially as it relates to compute spend.

00;16;51;14 - 00;17;10;25

Speaker 2

But it's also like kind of optimistic in other ways, like compute as a portion of our revenue is going down. So like I read that and I thought, you know what, if you buy their story and there are elements to it that are very purchasable, I think it could make sense. Maybe now's a good time for you to try and summarize it in one sentence or a few sentences for our for our viewers.

00;17;10;25 - 00;17;11;01

Speaker 2

Yeah.

00;17;11;03 - 00;17;12;02

Speaker 4

As I as I said at the.

00;17;12;02 - 00;17;30;04

Speaker 1

Beginning here, I don't think there's much material that's new here. I think we all knew this was coming. I think that this is going to, you know, whether it ends up being $2 trillion or $1.5 trillion, it's going to be massive. It already is so critical to what's going on in the economy. I'm a bit worried about how intertwined that is.

00;17;30;04 - 00;17;48;16

Speaker 1

But, you know, this is the company, as you said in many ways, at the center of it, right? They're sort of building the future that everybody is talking about it. And that's exciting for people. And does it make sense to buy a few shares of that because you're excited? Sure. Are we going to see, you know, as it's added to indexes, just having more and more exposure to it?

00;17;48;17 - 00;18;07;18

Speaker 1

Yeah, that worries me a little bit. But you know, it is, I think frankly, much more than space, which you know, it's the record holder currently. It is, you know, the biggest, most exciting company right now in this space. So that's that's worth something. But at what price or is it add read time already.

00;18;07;18 - 00;18;08;29

Speaker 2

It's added time. Let's get that out.

00;18;08;29 - 00;18;31;25

Speaker 1

Of let's do it. Everybody knows that running a successful IT shop is extremely hard and AI has only made it worse. True, you know the drill. Immature processes, a fragmented data ecosystem and unlimited demand for it. It all makes for an impossible job. And yet nobody has invented a unified operating system for it. We're literally the last function without one until now.

00;18;31;27 - 00;18;50;25

Speaker 1

That's why Infotech created CIO analytics, a single platform to run your IT department. We call it API driven advisory. It's powered by your data, our analyst insight, and it works with your existing tools right out of the box. If you're interested, check out the link in the description to request access and see CIO Analytics in action for yourself.

00;18;50;29 - 00;19;01;03

Speaker 1

An Infotech team lead can get you started. Same day. Better context, better insight, better IT performance only with API driven advisory. Only with infotech.

00;19;01;06 - 00;19;16;10

Speaker 2

Thank you Jeff, I should say that we are employed by Infotech and that is actually the infotech logo behind us there that's sort of etched into our wall here. So definitely check us out at Infotech. I know I'm going to head there after the show. Maybe I'll look right now while you're talking. Let's get into story number two.

00;19;16;12 - 00;19;46;16

Speaker 2

This one actually really entertained me. I enjoyed it a lot. And so here's my summary in today's piece of what I'm calling Numa Chao, covering all of an AI Dumas nutritional needs in a single portion. Apollo economist Thorsten Stork says that only The Magnificent Seven and that is alphabet, Amazon, Apple, Microsoft, meta, Nvidia and Tesla traditionally is receiving a payoff from AI for the rest of the S&P 500.

00;19;46;19 - 00;20;01;17

Speaker 2

Nothing. The other S&P 493, as he calls it in the post. And it's actually a very short post, and you should check it out if you haven't read the thing. It's just it's a couple of paragraphs and some charts. It's pretty insightful. Of the S&P 500, only The Magnificent Seven are growing their revenue per employee, which is a key metric.

00;20;01;17 - 00;20;20;16

Speaker 2

Anthropic reported that over the previous trailing 12 months and it was way up. Margins are only up in tech. They're pretty flat everywhere else. Productivity is flat, which is kind of concerning. Productivity growth is flat, and AI spending is concentrated very heavily in VC backed firms. And this is actually something that anthropic called out in their perspectives as well.

00;20;20;16 - 00;20;42;10

Speaker 2

They said that only 0.1% of expenses of the S&P 500 were actually like AI and were anthropic token related. So there's obviously a lot of room to grow there. I would argue this is bad news for anthropic, but it might be a needed reality check for the rest of the market. Again, it is a short blog post made a lot of waves, so you should definitely check it out.

00;20;42;10 - 00;20;46;04

Speaker 2

So Jeff, what did you think of Thorsten Stock from Apollo Global Management's piece?

00;20;46;05 - 00;21;00;10

Speaker 1

I thought it was interesting. As you said, it's short. It's mostly just a handful of charts that to me like don't actually paint. Maybe the picture that they think they're painting. So first of all, this is a story we've heard before many, many times over the past handful.

00;21;00;12 - 00;21;02;04

Speaker 4

But but this but this is the productivity.

00;21;02;04 - 00;21;15;01

Speaker 1

Paradox, right? This is the same conversation we were having in like the 70s, the 80s, the 90s, about like, oh, there's all this IT investment. How come it's not increasing our productivity? Right. Like Robert Solo talked about this solo paradox.

00;21;15;01 - 00;21;15;28

Speaker 2

It's famously. No.

00;21;16;00 - 00;21;41;15

Speaker 1

Exactly. You know, was talking about this. And you know, I expect to see the same thing again. I just don't think you can, like, put this tool on everybody's desk, say, have Adour with, you know, Claude or ChatGPT and expect to see it, like radically increase, like the ratio between number of employees and amount of revenue. Now, like, let's be really clear, that's exactly what they're looking at when they talk about productivity.

00;21;41;17 - 00;22;00;11

Speaker 1

They're looking at tech and they're looking at non-tech. And they're saying, how much revenue is there per employee? And what you're seeing is non-tech. It's basically been flat for the last handful of years. If you look at tech, it's pretty interesting. You can see actually since about pandemic times it was much higher. It dipped and now it's kind of going back up.

00;22;00;17 - 00;22;28;28

Speaker 1

And to me, the story first of all, like let's just acknowledge that AI is at best implicit in that. Like there's no direct line here to AI. But the other thing that we all know happen in tech post-pandemic is there was massive hiring in tech, right? They hired all these new positions. And then as interest rates kind of crept up again, as there were sort of concerns broadly in terms of market performance, they've said, oh, you know, maybe we don't need all these developers, we don't need all this.

00;22;29;00 - 00;22;34;13

Speaker 2

Well, they they initially juice the denominator right and then right. And then the result of the equation is a smaller number.

00;22;34;14 - 00;22;35;05

Speaker 4

Well, so that's.

00;22;35;05 - 00;22;54;19

Speaker 1

Exactly what I'm saying is going on is there's a story here where it's actually driven by AI. There's a story here where they over hired, they corrected for it. And now you're just seeing that number go up because they were able to correct for it without it directly impacting their their revenue numbers. Right. And if you actually look at like if you look at the past year, you're like, wow, it's actually gone up a decent amount.

00;22;54;19 - 00;23;15;16

Speaker 1

I think it's like 20 or 25%. If you look at where it was, where the chart starts around the pandemic time, it's a much smaller increase in productivity. So to me, I mean, two things going on. One of them is if you actually believe, AI is going to completely change productivity in the economy, you know, I feel a little bit sorry for you because I've been to this.

00;23;15;16 - 00;23;16;28

Speaker 2

Good to know that you feel sorry for me.

00;23;16;29 - 00;23;22;29

Speaker 1

I've been to the circus many times before, and as I said, looking at these numbers, they don't even tell the story that I think the article.

00;23;22;29 - 00;23;43;07

Speaker 2

Was trying to. So okay to push back though, in defense of our friend, probably doctor stock. Honestly, my thought here is I don't think the point of that blog post is the impact on the tech companies that might be engaged in something cycle. I think it's the rest of the S&P that we should be looking at who didn't necessarily engage in that over hiring to that same degree.

00;23;43;08 - 00;24;03;27

Speaker 2

I mean, an observable implication of AI should be that we are seeing people able to do more. Right? Like, hasn't the pitch always been your people will be able to focus on the things that matter, I think is usually the language that they use instead of the stuff that doesn't matter. Presumably that makes them more productive. Otherwise why would we do this investment?

00;24;03;28 - 00;24;29;11

Speaker 2

And so yeah, okay, tech companies, it's a little uneven. But like that flat line of productivity, a revenue of margin, whatever that we're seeing. This is all inflation adjusted by the way, that we're seeing from the rest of the S&P 500. I mean, like what is even the point of AI in this case? And I would I would argue that the pitch then anthropic is making is that your people should be more productive and they focus very heavily on their revenue per employee because it's a metric that they care about.

00;24;29;11 - 00;24;46;01

Speaker 2

And presumably they would want to project that out across the broader S&P 500. And I'll tell you what I think anthropic would say in response to this particular piece. I haven't looked to see if they've issued a statement, but I think they'd say two things. First, I think they'd say that our penetration outside of certain areas is quite low.

00;24;46;03 - 00;24;52;29

Speaker 2

So they say that the S&P 500 actually isn't spending that much with us, and they should spend more. And then they would see the real benefit. I think that's what they would say.

00;24;53;04 - 00;25;01;05

Speaker 1

Not well, why do you why do you care what anthropic would say anthropic is going to say? The answer is to buy more AI. It's like asking in video what they think video is going to say. You need to buy more chips.

00;25;01;06 - 00;25;25;00

Speaker 2

Of course they're going to say, but I would go to them first, right? Because it's the first objection in a sales conversation, right? Like other people use you and they don't see a benefit. How do you respond? Right. So okay, I'm not necessarily saying I buy that. The other thing though, and this is where I agree with you, that maybe the charts that that Apollo is giving us don't tell the story that they intend is, well, they talk about how most of the token spend is actually going to VC backed firms.

00;25;25;03 - 00;25;46;01

Speaker 2

But like the S&P 500 comprises, you know, legacy companies that are in the public markets and VCs, I think, have largely exited at that point. And I'm not 100% sure how they do their breakdown, but it looks to me like most of the spend is actually happening outside of the list. That stock is comparing the mag seven two, right?

00;25;46;02 - 00;26;04;20

Speaker 2

So like some manufacturer with loads of revenue that can't figure out how to get off COBOL, which we'll talk about in a bit, and is, you know, puttering around, trying to figure out how to implement clod in a use case or two, but still manages to be in the S&P 500 because, you know, they've been in business for 100 years and have all these relationships and make something that's hard for other people to scale up to.

00;26;04;22 - 00;26;28;11

Speaker 2

I mean, that's not where I would look for the ultimate AI innovation, those VC companies who largely aren't going to be included because many of their revenues aren't public because they haven't exited yet. Right? Like they can release stuff, but they're not required to via traditional, you know, SEC required disclosure mechanisms. We don't know what the impact of AI among those companies is based on this.

00;26;28;14 - 00;26;41;23

Speaker 2

And so I would think like it's a little bit contradictory to say, well, you know, the rest of the S&P 500 isn't gaining any benefit. And by the way, companies outside of the S&P 500 are the ones who are spending all the money on AI. So maybe the diagnosis is not enough AI.

00;26;41;25 - 00;27;04;11

Speaker 1

Well so so that's what I was going to ask you. I mean, you've made kind of an interesting setup there a couple of different ways saying that this is actually not a fair take because you're not even comparing with AI. And without AI you're just basically saying, hey. What? You're basically saying that it's completely possible that everybody just needs more AI, and that's going to be what what gets better here.

00;27;04;11 - 00;27;22;10

Speaker 1

But but I want to ask you, because you can make that statement as kind of a devil's advocate statement. But I want to ask you, like, do you believe that adding more AI is going to juice the productivity numbers? Or do you believe that AI is just not going to have the impact on productivity that the anthropic of the world are telling you?

00;27;22;12 - 00;27;43;05

Speaker 2

Okay, so a couple answers to that and I will tell you my honest personal belief. So everybody watching will know that this is how I actually feel. When I saw this, at first my thought was, I'm not sure how much I buy this, because the use cases that I've seen for AI, at least at a micro level, are actually incredibly compelling.

00;27;43;07 - 00;28;04;27

Speaker 2

Like the amount of time it takes me to create a report now, which I'm required to do by people in my chain of command. This guy, it's fraction and I can use that time to do other things, and I see that myself. Like the Google made its entire name essentially sorting information incredibly effectively for people and then running ads against the results.

00;28;04;29 - 00;28;29;19

Speaker 2

And that's a several trillion dollar business right now in terms of market capitalization. AI is, I think, the same level of utility, if not more useful applied in my context. So when I look at these numbers and people say, well, it doesn't make me more productive, I say the amount of time I spent trying to find something in my email has dropped dramatically, and if other people were using it that way, I mean, that alone would be an incredible business.

00;28;29;19 - 00;28;55;18

Speaker 2

But you can use it to all these other things like solve millennium problems. You know, you can you can use it to, to make advances in biotechnology and yes, adequately place your avocado based spread in your branch line cook situation like you can do that. And so to me the promise of the tool is very real. I think what's happening is a lot of the things that it optimizes don't account for the entire stretch of a process, which we're going to talk about in our Kobol story.

00;28;55;18 - 00;29;08;08

Speaker 2

And so, yeah, I've optimized parts of it, but like, I haven't solved the bottleneck somewhere. And then also these larger companies that are featured in this study, they're not exactly known for being wildly dynamic and jumping on things.

00;29;08;10 - 00;29;34;21

Speaker 1

It's it's interesting. And I, I agree with you on parts of that. But, you know, to me it's still just like comes down to the fact that, as you said right now, AI for the most part is showing up as an individual productivity booster, right? It's actually helping individuals feel more productive. It's not being captured at any sort of like organizational effectiveness layer in a way that's like macro scale impacting the productivity organization.

00;29;34;23 - 00;29;41;13

Speaker 2

From my personal use case, I think that's a fair statement, but I think I'm not a software developer. If I was a software developer, I think my answer to that question might be different.

00;29;41;13 - 00;29;59;11

Speaker 1

Well, I don't I don't know that it would because and we could go deep here, but I don't know that we're going to have the time to do that. But software developers like you can do more stuff. But do you actually have fewer developers? Like, are you actually firing developers because you can do that, maybe. Or are you just launching a bunch more stuff also?

00;29;59;12 - 00;30;18;24

Speaker 1

Maybe. But does that make you more money? Right. And so there's just there's some ifs here that also comes back to like how big is software development into your core business if you're in tech? Yeah, maybe it's really big and that's what you're doing. If you're everybody else, maybe it's just sort of a bit of a sideshow and it's not going to really juice your, your revenue in a big way.

00;30;18;27 - 00;30;37;23

Speaker 1

But to me, like there's just like a bigger belief I have, which is that if you're bringing AI into your organization because you think it's going to save you a big pile of money, like I think you're going to end up the same as if you 15 years ago believe that bringing cloud into your organization was going to save you a big pile of money, right?

00;30;37;25 - 00;30;57;26

Speaker 1

Or bringing like, personal computers into the organization was going to save you a big pile of money. Like, it's just that's not how these waves happen. There's a lot of good reasons to adopt AI. Certainly down the road, I think there could be some sort of cost savings that could transform your organization. But I just don't believe that, like, oh, it's going to save our it's going to save us money.

00;30;57;26 - 00;31;04;15

Speaker 1

It's going to juice our productivity by employee in any reasonable short term time frame. Like I just I don't buy that at all.

00;31;04;15 - 00;31;17;23

Speaker 2

But nobody here is really claiming money savings. Like Torsten Stock isn't saying, oh, this is a disaster because nobody's saving money. What are you saying is our revenue isn't growing per employee, like our productivity isn't growing. All of the spend is concentrated in one place. So I agree with you. Well, so.

00;31;17;23 - 00;31;24;21

Speaker 1

So let me frame it out differently. I don't think we're going to see that productivity gain. I think you can add more AI everywhere. I don't think you're going to see it in the next five years.

00;31;24;21 - 00;31;44;27

Speaker 2

But like if a thresher makes me a more productive farmer and allows farms to get bigger, how is a really effective coworker or a cloud code type agent? Not the equivalent for a knowledge worker. Like that's what I'm struggling with, because when I look at this at sort of the the micro level, as you've already alluded to, I see the potential for this, and I've used these tools to build stuff.

00;31;44;27 - 00;32;01;28

Speaker 2

And like I agree, the future isn't using Claude to generate a 15 slide deck in 30s and then sending it to you and putting my feet up. That's not what it's for. But like Linus Torvalds, who created Linux or was part of the team that did, you know, the eminent Linux developer, he started using it to push code.

00;32;02;00 - 00;32;23;29

Speaker 2

This is like the ultimate open source guy. He's like, wow, I know what I'm doing. This makes me so much more effective. And I know that's just an anecdote, but like, I really do think that if you can effectively insert it into your process, I really think that the potential for change is there. And I think that a lot of the change is happening, but it's probably not happening in the S&P 493, at least at the pace that it's happening elsewhere.

00;32;23;29 - 00;32;28;09

Speaker 2

So I'd love to see some results from VC. Back to companies I agree.

00;32;28;09 - 00;32;53;11

Speaker 1

And I think like I'm not saying there's no change. I think we're seeing process optimizations. I think we're seeing it slowly, and I think there's just a lot less revenue impact on process optimizations here and there. Then from knowledge workers like you're just assuming that there's like a linear relationship between knowledge worker efficiency and you know, the overall productivity like related to revenue.

00;32;53;12 - 00;33;11;14

Speaker 1

Right. And I just think that's totally bunk. Like, yeah, a thresher on the farm. It's like the speed at which you can harvest, you know, and like we can get very deep into farm metaphors. We should next episode. Yeah, but but the point is, it's not 1 to 1. And so, like, you just can't squeeze more and more juice from that.

00;33;11;14 - 00;33;20;21

Speaker 1

And there's a lot of good to be had there. I don't want to come across as, like, AI is bogus, but I mean, let's let's see what happens in the next few years.

00;33;20;21 - 00;33;45;08

Speaker 2

I think in the next few years, we're going to have an epidemic of people who don't know what to do when their AI solution goes down, in much the same way that the internet, while solar paradox very famously applied to the internet, now businesses can't run without it. So maybe a final thought on this. Is this a situation where everybody's running faster to say, in the same place, because Claude is very much nonproprietary in the sense, obviously it's anthropic property, but it runs on every cloud, basically.

00;33;45;08 - 00;34;15;27

Speaker 2

And anthropic is very proud of this. With the relationships with CSPs, you know, they use system integrators. They talked about this. You know, the big for your Accenture's right to distribute it. Open AI I think is following a similar idea. Everyone's using this. And so maybe if the market isn't growing as a whole, but I will fall out of the S&P 500 if I don't use this in much the same way that if you started a company and said, yeah, we don't use PCs and we don't use the internet, I mean, I'm not suggesting that those are fundamentally life altering today, but like, they're kind of table stakes.

00;34;15;27 - 00;34;28;11

Speaker 2

So maybe we enter a situation where AI just becomes a table stake, because you can't be expected to do your own reporting manually. And maybe that's what's being reflected in these numbers, although I think penetration is probably too small to really make that case right now.

00;34;28;15 - 00;34;37;06

Speaker 1

I think there's a lot of merit to that. To that point. I think it's like, I don't I don't think there's a case here where you just completely opt out and things go well for you. Yeah, it's not that way.

00;34;37;07 - 00;34;52;19

Speaker 2

I think we agree on that. I'm sure there's somebody in the YouTube comments who vehemently disagree with that. And so please make yourself known, Jeff. And I love reading your comments. And I think in future episodes of the show we might do a bit of a mailbag. So leave a comment for a chance to be featured on an episode of Digital Disruption.

00;34;52;19 - 00;35;01;00

Speaker 2

Now try and keep it PG though. This is a it's a family show. All right. Let's talk about our third story today. Jeff, have you ever heard of cobalt?

00;35;01;01 - 00;35;03;03

Speaker 1

I'm familiar with cobalt. Tell me the story though.

00;35;03;05 - 00;35;21;26

Speaker 2

I think we've I think we've mentioned it a few times. It stands for common business oriented language. It's from the 1950s, so it's pretty old the early days of electronic computers. And it's the thing holding up the important enterprise systems that you might find at a bank or in a government. In fact, those are two of the biggest use cases.

00;35;21;28 - 00;35;42;27

Speaker 2

The verge wrote a piece on an open AI blog, and the piece was titled can I help solve the COBOL Code Crisis? And I believe there's a lot of headlines that says, if you have to ask, the answer is no. Open AI is claiming in this post that their new models are much better at refactoring legacy code things written in COBOL, these old mainframe systems.

00;35;43;02 - 00;36;02;18

Speaker 2

But the experts The Verge spoke to call into question the degree to which that will actually move the needle. So we should talk about it. This is not the first time we've heard this story, by the way. If you go back earlier this year, IBM had its worst day ever on the stock market, and IBM had been around for a very long time.

00;36;02;18 - 00;36;13;14

Speaker 2

They're like 100 plus years old. Actually. Fun fact they have their roots in the guy who built a machine to conduct the US census in the late 19th century. Like, that's how old IBM is.

00;36;13;15 - 00;36;14;09

Speaker 1

Wow, that is fun.

00;36;14;11 - 00;36;33;14

Speaker 2

Yeah. Thank you. You know, you're the only person I talked to who appreciates electromechanical tabulation anyway. Or were you not being sincere? I should move on. I should move on. They had their worst day. Their stock just plummeted off a cliff because anthropic released a similar blog post, basically saying, hey, you can use Claude to refactor these mainframe systems.

00;36;33;16 - 00;36;59;12

Speaker 2

Some of them are written in COBOL, and you won't need to pay IBM to maintain them anymore. And of course, I mean, I'm said, oh no, and they're not exactly winning the cloud or AI race. So that was quite concerning for them. And so yeah, this is the second time that we're hearing it. And my take on it, honestly, Jeff, is I really don't think that the AI companies are processing the COBOL problem in the right way.

00;36;59;14 - 00;37;22;07

Speaker 2

So yeah, they say you take this code, you run it through our updated model, and it's going to produce something that passes our quality tests pretty significantly. And that's great. And literally rewriting code is obviously going to be part of the process of refactoring it. But there was a quote that I actually wrote down from one of the experts that The Verge talked to here, Vadym Zaitsev, who's a professor.

00;37;22;09 - 00;37;44;11

Speaker 2

He said rewriting code is not the same as replacing system, and every system comprises a whole bunch of different elements. Talk to any systems engineer. We have a bunch here at Infotech, which is pretty amazing to to work with folks like that. And they'll tell you that, yeah, you know, the actual lines of code that you need to maintain, that's part of it.

00;37;44;13 - 00;38;02;18

Speaker 2

But like there's business process, there's like latent knowledge of all sorts of other things. And this doesn't solve that. And so it would be like there's ten steps in a transaction and you solved one of them. But in the same way that two women can't have a baby in four and a half months, you're not necessarily going to massively expedite this transition here.

00;38;02;18 - 00;38;03;15

Speaker 2

So what do you think, Jeff?

00;38;03;16 - 00;38;30;27

Speaker 1

I totally agree. I mean, first of all, we should probably like talk a little bit about the scale of this problem. Like this is I'm tempted to call this like $1 trillion problem, like the amount of massive companies and organizations that are running on these systems and are limited by it, like forget AI adoption in general, like they just can't do a whole slew of, of modern things because they have these really, really old timey, rickety systems.

00;38;30;27 - 00;38;50;00

Speaker 1

It's huge and it's a huge drain on productivity. And migrating them is just like it's like a career ender for a lot of leaders, right? I was actually meeting with an insurance company earlier this year, and they were on year 11 of their mainframe migration strategy. And spoiler, it wasn't going very well. Yeah.

00;38;50;00 - 00;38;50;01

Speaker 2

I.

00;38;50;01 - 00;39;08;03

Speaker 1

Don't like like nobody was like, if you're 11 and this is the year, you know, like it's just they just go on indefinitely. And I think there's a lot of, you know, I'll call it like prayer around the fact that, like, oh, once AI is here, this will like people just tend to talk about AI, like, oh, well, our data quality is a disaster.

00;39;08;04 - 00;39;29;06

Speaker 1

The data is everywhere. It's a mess. But once AI is here, like, you know, pixie dust, it's going to like magically fix it. Everything better. And I think this is the exact same story. Like, are there cases where it can be used to refactor code? Yeah, but I think you're the quote you picked is exactly the one that caught my attention in the article, which nails it, which is, yeah, there's code, but there's also systems.

00;39;29;06 - 00;39;49;11

Speaker 1

Does this help you get away from the system? No. Does it help you actually make the system like more livable? Incrementally. But I don't think it completely changes what you're able to do with this. And until you're able to kind of wrap your arms around the thing and get away from it entirely, I think you're still like, you know, going in the same circle.

00;39;49;12 - 00;40;08;18

Speaker 2

Yeah. And a couple more comments on this because I think it's interesting, like the clue to COBOL is in the name, like it's common business oriented language like previously to COBOL and to other similar programing languages like, you know, you had to speak machine, you know, right, in assembly or whatever COBOL was designed actually make it comprehensible. The issue is, is that we since moved on to other more modern languages.

00;40;08;18 - 00;40;24;08

Speaker 2

And a lot of the people who write COBOL have since retired, or they're very expensive or worse, unfortunately. And also like a lot of the documentation just isn't there. So you look at your 100 million lines of COBOL and you're like, what does this even do? And somebody refactor it for you. That doesn't actually necessarily solve that problem.

00;40;24;09 - 00;40;34;28

Speaker 2

So there's a whole bunch of elements to this, but there's a useful way to think about this that I was sort of playing around with as I was preparing for the show this morning. Have you ever heard of the Doorman fallacy?

00;40;35;00 - 00;40;36;00

Speaker 1

No, I don't think so.

00;40;36;01 - 00;40;59;03

Speaker 2

So this is an interesting one. And it was coined by a guy named, I think, Rory Sutherland, if I'm not mistaken, here. And the idea here basically is if you reduce something that a person does right down to like just a single element. So a doorman, I don't know, opens the door or something, and then you automate that.

00;40;59;06 - 00;41;28;00

Speaker 2

You end up creating a bunch of problems because you didn't realize all the other stuff that the doorman did. You know, your tenants start leaving because, you know, they don't feel welcome. You know, people start congregating outside of the building and, you know, smoking or something and ruin. The doorman would have historically chased them away. Right. And I think that like a blind tech optimism where like, oh, all that's needed to solve this COBOL problem is to just rewrite the code is kind of like a version of the doorman fallacy.

00;41;28;00 - 00;41;48;12

Speaker 2

It's like, well, we'll just automate that one thing that somebody who maintains this system does and then boom. The problem is solved. But you're leaving all this other stuff on the table. And I think that's probably what this is. And so while I'm much more bullish on like the prospects of AI more generally, and eventually we'll get to the point where this stuff is so powerful that it can just barrel through all of this.

00;41;48;12 - 00;42;07;17

Speaker 2

You can feed it your 250 million lines of code, and it can just brute force it. We're not there yet, though. And so I actually, I don't know, I read about this OpenAI blog post, I read this verge piece, and basically every professor that they interviewed said exactly what was in my brain, although they had the audacity to say at first, Jeff, which was neat, might be useful in some situations.

00;42;07;21 - 00;42;09;16

Speaker 2

Probably not solving anything tomorrow.

00;42;09;19 - 00;42;29;06

Speaker 1

Yeah, well, and that's it. Like, it's I mean, I don't know if it's fair to call it like spaghetti code, but so much of the COBOL code is so context specific. It's some of it may have been in place there for 40 or 50 years. And there's so many of these stories where, you know, you've got this like 65 year old guy, usually a guy, frankly.

00;42;29;07 - 00;42;30;02

Speaker 2

Larry.

00;42;30;04 - 00;42;34;22

Speaker 1

Who, who, like, retires. And he was the one writing the code for the last 40 years.

00;42;34;22 - 00;42;36;11

Speaker 2

Or documenting it or.

00;42;36;14 - 00;42;44;29

Speaker 1

Documenting it, but that but that's what I'm getting at, right? Like so. And these companies are begging the Larry's to come back and paying them, like, just astronomical.

00;42;45;00 - 00;42;47;00

Speaker 2

500 bucks an hour for hours.

00;42;47;02 - 00;43;07;03

Speaker 1

Yeah. The lyrics can demand whatever they want because nobody else gets it. And are they paying the Larry's? Because Larry is the only guy who can write in COBOL? No. They're paying Larry for what's in his brain, for understanding the context, for understanding how things work, and actually being the guy who can, you know, refactor the code himself or put in new lines that are going to make things work.

00;43;07;04 - 00;43;23;16

Speaker 1

Right. It's not just all I speak the language like. It's just so difficult for, you know, an AI system to get its arms wrapped around this in a way that makes it comprehensible. So I yeah, I don't think this problem is going away anytime soon, despite what.

00;43;23;16 - 00;43;38;08

Speaker 2

Opening eyes. Okay, so we did three stories today and prospectus, you and I, I think, disagree a little bit on the relative merits of that. We'll see how it does. I don't make predictions because I'm literally always wrong. Maybe it's like that Seinfeld episode where I just do the opposite of, of of what I was planning on doing.

00;43;38;08 - 00;43;56;06

Speaker 2

So I will keep that close to my chest. We talked a little bit about Apollo Global Management and their skepticism here. I think again, a little bit of disagreement, but we landed on a place of positive agreement. So there you go. There you go. That's it for us this week. Anything you want to say to the folks before we sign off?

00;43;56;06 - 00;43;57;07

Speaker 1

Just lay us out, Jerry.

00;43;57;10 - 00;44;20;11

Speaker 2

All right. This show is hosted by Jeff Nielsen and me, Jeremy Roberts. Our producers are Coulibaly and Maria Knight. Tim Ly is our technical director. And, Jeff, he's going to be editing this episode. So any issues that folks have with editing, we can direct those at. Tim, thanks to our studio team of James Groom. He's right there behind the camera, Tim Champion and of course, Brian Mortgage couldn't do it without you.

00;44;20;11 - 00;44;39;12

Speaker 2

And a special thanks to Holly Callanan, who holds down the fort while we're here in the studio processing the week's news. Digital disruption now is an infotech production. If you've got an IT department, you should check us out@infotech.com, especially in this age of AI, and we will see you next week. Thanks, Jeff.

00;44;39;17 - 00;44;40;13

Speaker 1

Thank you for.